Skip to content

Changelog

0:10 - Plugins Run on OWOX AI
0:28 - AI Assistants Get OWOX Rules
0:57 - What’s New, Right in the App
1:16 - Plugins Get Typed Answers from AI
1:39 - And a Few More Things

  • 7eb4159: New features are easier to discover: each release shows what’s new right in the appSection titled “New features are easier to discover: each release shows what’s new right in the app”

    OWOX Data Marts ships a release every week, but the app itself never said so: to learn what changed, you had to open Help → What’s new and read the changelog. A feature that removes a manual step could go unnoticed for weeks while people kept doing that step by hand. Now the app shows each new release in a card in the bottom-left corner: its what’s-new video, the video’s chapters, each of which plays the video from that scene, and a link to the release in the changelog.

    • The card opens every time you open the app until you close it with ×. After that, the release does not come back in that browser; the next release opens its own card.
    • The card waits for the release’s video. If there is still no video three days after the release ships, the card lists the release’s changes instead; once that card is closed, a video added later does not bring it back.
    • In a browser where the card has never been closed, it opens only in a project that already has a Data Mart, so a new user starts with the onboarding.
    • The card does not open over another video. A video can take its place, and the card then comes back on the next visit. The card never opens in view-only sessions.
    • A self-hosted instance announces only releases up to its own version, which /api/flags now returns as APP_OWOX_VERSION. The app reads the list of releases from https://docs.owox.com/docs/releases.json, so a browser that cannot reach docs.owox.com shows no card.
  • 81bd5fc: Backfill a catalog for any period in one run, and follow a long backfill as parts of one operationSection titled “Backfill a catalog for any period in one run, and follow a long backfill as parts of one operation”

    An analyst who reloads a list of ads, campaigns or accounts for the past year used to confirm “Start 12 runs”, then find one entry in Run History labelled “segment 1 of 12” and nothing after it — such a catalog has no date filter, so a single run loads all of it, and the promised runs never happened. Now the Manual Run form says the node loads the whole catalog in one run, whatever period you pick, asks for no confirmation, and Run History shows one plain run.

    Manual Run sheet for a Reddit Ads catalog node with 1 January – 15 April selected: the notice says the backfill loads the whole catalog in one run, whatever period is picked

    For a node fetched by date, a period longer than 31 days loads in consecutive runs, now called parts and cut at calendar months rather than every 31 days: 1 January – 15 April loads as 1–31 January, 1–28 February, 1–31 March and 1–15 April, so each part lines up with the month you will check it against. The form reads “105 days, loaded in 4 parts, one per calendar month. It counts as 4 runs”, the confirmation asks “Start backfill?”, and each entry in Run History reads “part 3 of 4 loaded”. A backfill already running when this version is deployed finishes with the 31-day parts it was planned with. Operators of a self-hosted deployment with several replicas: finish the rollout before anyone starts a backfill longer than a month, because a replica still on the previous version continues a calendar-month backfill with 31-day parts and leaves days unloaded. The connector fields API gains an isFetchedByDate flag per node, so an API client can tell a catalog node from a dated one before it starts a backfill.

    Manual Run sheet for a dated node with 1 January – 15 April selected: 105 days, loaded in 4 parts, one per calendar month, counting as 4 runs

    Start backfill? confirmation naming 4 parts and 4 runs, with Keep editing and Start backfill buttons

    See Backfill a custom period.

  • 8ac4b37: Microsoft Ads imports work for advertisers who sign in with GoogleSection titled “Microsoft Ads imports work for advertisers who sign in with Google”

    Teams that sign in to Microsoft Advertising with Google could set up the Microsoft Ads source, but every run failed with error 126 IdentityTypeMismatch and imported nothing. Microsoft Advertising accepts only a Google sign-in for these users, and the source offered only a Microsoft one. Now the Microsoft Ads Credential offers Sign in with Google next to Sign in with Microsoft, and runs import data with the Google sign-in.

    The Add Credential form with Microsoft Ads selected as the provider: under Connect with Microsoft Ads, a Sign in with Microsoft button, a Sign in with Google button below it, and a Manually link

    • Signing in now checks the user with Microsoft Advertising. A sign-in of the wrong type fails right away and says which button to use. The Credential shows the email of the connected user instead of “Microsoft Ads User”, and a new Credential is named after that email instead of “Microsoft Ads”.
    • A run whose Credential has the wrong sign-in type stops after the first request with the same advice, instead of repeating error 126 for every day of the import.
    • Other Microsoft Ads errors in Run History read as Microsoft Advertising’s own message, such as HTTP 401: Authentication failed. (105 InvalidCredentials), instead of the raw JSON response.
    • Credentials signed in with Microsoft keep working as before.
    • Self-hosted: Sign in with Google appears once you set OAUTH_MICROSOFT_ADS_GOOGLE_CLIENT_ID and OAUTH_MICROSOFT_ADS_GOOGLE_CLIENT_SECRET, with OAUTH_GOOGLE_REDIRECT_URI pointing to the Google callback page, see Environment Variables. A Credential entered by hand can use a Google OAuth client too, see Microsoft Ads credentials.
  • ceb2db1: Plugins that use AI work at once: choose OWOX AI instead of your own AI keySection titled “Plugins that use AI work at once: choose OWOX AI instead of your own AI key”

    A project member who installed a plugin that uses AI had to get an API key from an AI provider, add it to the project as an AI Credential and select it before the plugin worked. That was extra work, and a key to look after. Now the install dialog offers OWOX AI, the AI models that the deployment provides, as the first option and selects it, so the plugin works at once in its pages, backend actions and scheduled executions. Members can still choose their own AI Credential, and can switch between it and OWOX AI later on the plugin’s Credentials tab. Plugin code and plugin.json do not change.

    • On OWOX Data Marts Cloud, each successful OWOX AI call is billed as an AI Process Run under the existing rules and uses the project’s credits. A project that cannot run AI Process Runs gets 403 with the code PROJECT_OPERATION_BLOCKED. On a self-managed deployment, OWOX AI calls are not billed.
    • OWOX AI provides fast and reasoning when the deployment sets its chat configuration. It provides embedding only when AI_PLUGIN_EMBEDDING_MODEL is set. The model IDs depend on the deployment.
    • A streamed AI response no longer includes the provider’s response ID or the model ID, for OWOX AI and for an AI Credential. Plugin code cannot tell which one answered.
    • Self-managed operators turn on the OWOX AI chat models by setting AI_BASE_URL, AI_API_KEY, AI_PLUGIN_FAST_MODEL and AI_PLUGIN_REASONING_MODEL. AI_MODEL is not a fallback for plugins.

    See OWOX AI, Environment variables, Consumption units and the plugin authoring guide.

  • 0b81b46: claude.ai now gets every rule for querying a Data MartSection titled “claude.ai now gets every rule for querying a Data Mart”

    When an analyst asks about their data in claude.ai, the AI assistant builds the query with the query_data_mart MCP tool and follows the rules in the tool description. claude.ai keeps only about the first 4,096 characters of a tool description, and this description had about 9,000. So the assistant did not get the rules for choosing between slices and filters or for reading the results, and it could filter a joined Data Mart after the join and report wrong totals without an error. Now the description is shorter than 4,096 characters, and the rules for reading the results come first. Claude Code has the same limit when it loads tools without tool search.

    • The table of operators and aggregations for each field type is no longer in the description. get_data_mart_details_by_id already returns the operators and aggregation functions that each field accepts, in operators_by_category and allowedAggregations.
    • The description and the query_data_mart reference now say when rows are grouped, and that a filter on a joined field sees one value per join key, not the joined rows.
  • bf173a4: Safer custom connector tests in your AI assistant: API keys stay out of the chatSection titled “Safer custom connector tests in your AI assistant: API keys stay out of the chat”

    A Data Owner who builds a custom connector through MCP tests each node before publishing it. Previously test_custom_connector also took the manifest and temporary API keys in its arguments, so a key that the user gave the assistant went through the chat, and the test could run a manifest that differed from the saved draft. Now keys go only through saved Credentials, and the test runs the saved draft, so it checks the connector that you publish. test_custom_connector no longer accepts manifest or keys in configuration: the assistant saves the draft with save_custom_connector_draft, then tests it with connector_id and, when the API needs keys, a credential_id. ChatGPT keeps the tool list from the time when you added the OWOX app: disconnect and reconnect the app to get these changes.

    • create_custom_connector and save_custom_connector_draft reject a manifest that holds a literal key in authentication, in a parameter with attributes: ["SECRET"], or in a credential header or query parameter, or a [REDACTED] value. Declare each key as a parameter with attributes: ["SECRET"] and select a Credential that holds its value.
    • get_custom_connector_version shows such a literal key in any version as [REDACTED].
    • start_credential_setup takes an optional connector_id. The Credential form then opens with the hosts and key fields of that connector, and the user enters only the key values.
    • In ChatGPT, the assistant does not run the test itself. get_custom_connector_test_link gives you a link that opens the test in the Connector Builder with the node, parameter values and Credential filled in. Click Run test, then Copy for AI, and paste the result into the chat. Other AI clients use test_custom_connector.
    • Published connectors and Data Mart runs work as before.

    See Custom connector authoring and how to load new tools.

  • 44b3e5e: Connector Data Marts set up through MCP no longer fail their first run because of the table pathSection titled “Connector Data Marts set up through MCP no longer fail their first run because of the table path”

    A saved connector Data Mart with a three-part table path now stops its run with a format error. Change the table path to two parts through MCP or the API to run the Data Mart again. The web app cannot change the table path of a saved connector.

    An AI assistant that sets up a connector Data Mart through MCP often gives the table path as project.dataset.table. Previously OWOX saved and published this table path. The first run then failed with Invalid dataset ID. Now OWOX rejects such a table path when you save it or publish the Data Mart. The error message gives the expected format. The MCP tool description gives the format too, so the assistant can use the correct table path from the start.

    • Google BigQuery expects dataset.table. The connector always writes to the project of the Storage. The dataset name can contain only letters, digits, and underscores.
    • AWS Athena expects database.table, without the catalog.

    The web app already writes table paths with two parts. Connector Data Marts that you set up in the web app do not change. See MCP Server.

  • b0b37b8: Connector runs on a Google BigQuery Storage without a location no longer failSection titled “Connector runs on a Google BigQuery Storage without a location no longer fail”

    The web form requires a location for a Google BigQuery Storage. MCP and the API can save a Storage without a location. Previously, a connector run on such a Storage failed with Invalid location 'AUTODETECT' or Invalid location 'undefined'. This happened when the run created the target table or replaced all its data. These runs loaded no data until you set a location on the Storage. Now the run loads data.

    The run writes to the target dataset when this dataset exists. Otherwise, Google BigQuery creates the dataset in its default location. Set a location on the Storage before the first run to create the dataset in another location. A later change of the Storage location does not move the dataset. For the steps, see Google BigQuery. Only connector Data Marts on Google BigQuery change.

  • 11b79d6: claude.ai and ChatGPT now get the OWOX rules for safe setup and deliverySection titled “claude.ai and ChatGPT now get the OWOX rules for safe setup and delivery”

    When a Data Owner or analyst sets up Data Marts, Credentials or reports through an AI assistant, OWOX gives the assistant rules: ask before a deletion, never take provider secrets in chat, and do not send a report message twice. These rules were in one MCP instructions text of about 27,500 characters. claude.ai did not give that text to the model, ChatGPT gave only its start, and Claude Code its first 4,096 characters, so in these clients the assistant often worked without the rules. Now the shared instructions are short and start with the most important rule, each safety rule is also in the description of the tool that it applies to, and the new get_owox_guide tool returns the full workflow for one task. ChatGPT keeps the tool list from the time when you added the OWOX app: disconnect and reconnect the app to get these changes.

    • get_owox_guide topics are reporting, data_mart_setup, data_mart_runs, credentials, storages_destinations and custom_connectors. The tool needs only the mcp:read scope and does not read project data.
    • delete_context now tells the assistant that it can detach Data Marts, Storages and Destinations through MCP before it deletes a Context.

    See get_owox_guide and how to load new tools.

  • 0be60b9: Invited teammates get their email right away, and you know if it was not sentSection titled “Invited teammates get their email right away, and you know if it was not sent”

    When a Project Admin invites a teammate, the teammate learns about it from the invitation email. If that email did not reach them, the admin had no way to know and kept waiting for someone who never got the invitation. Now OWOX Data Marts sends the invitation as soon as you click Send invitation, and tells you when it could not be sent.

    The invitation email from OWOX Data Marts: Hey friend, you have been invited by user maria@example.com to join Marketing Analytics as a Data Owner, with an Accept invitation button and the project link below it

    • The email names who invited the person, the project and the role (Project Admin, Data Owner or Business User). Its Accept invitation button opens the project.
    • If who invited the person or the project name cannot be loaded, the email leaves that detail out. Without the project name, it says “the OWOX project”.
    • If the email cannot be sent, the invitation still stands. Invite member and the invite step after sign-up name who did not get it, so you can let them know yourself. The POST /api/members/invite response then has emailSent: false.
    • Only OWOX-managed projects send this email. Self-hosted installations keep the one-time invitation link.

    See Invite a Member.

  • 51d1018: Access request emails name the project and take Project Admins straight to the requestSection titled “Access request emails name the project and take Project Admins straight to the request”

    When a teammate asks to join a project, every Project Admin gets an email about it, and the teammate gets one once an admin approves. An admin of several projects could not tell from the email which project the request was for, and roles were shown as ADMIN, EDITOR or VIEWER. Now both emails name the project and use the role names you see in the app (Project Admin, Data Owner, Business User). The admin’s Check request button now opens Project Settings → Members of that project, where you approve or decline the request.

    The access request email to a Project Admin: the user with the email alex@example.com has requested a change in permissions for Marketing Analytics to the role of Data Owner, with a link to the Members page and a Check request button

    • If the project name cannot be loaded, the emails say “your project” or “the OWOX project” instead.

    • After approval, the teammate’s Join now button opens the project.

      The approval email to the teammate: you have been granted Data Owner access to Marketing Analytics, with a Join now button that opens the project

    • Both emails are sent even if the recipient turned off email notifications. Asking again for the same role while the request is still pending does not email the admins again.

    • If an email cannot be sent, the access request or the approval still goes through.

    • Only OWOX-managed projects have access requests, so self-hosted installations are not affected.

    See Manual provisioning.

  • dd208f6: Start an OWOX task from the claude.ai connector menu with its rules attachedSection titled “Start an OWOX task from the claude.ai connector menu with its rules attached”

    claude.ai does not give the AI model the instructions of an MCP server. So when an analyst or a Data Owner starts a report or a Data Mart setup in claude.ai, the model gets the OWOX workflow rules only if it calls get_owox_guide itself. Now the connector menu of claude.ai and Claude Desktop has an Add from item for the OWOX connector. Select a task there. For a report, a Data Mart setup or a check of runs, the matching guide goes into the conversation with your request.

    • Prompts: Explore my data, Create a report, Set up a Data Mart and Check Data Mart runs. In Claude Code, they are slash commands.
    • Resources: the six OWOX workflow guides at owox://guides/<topic> and a link to the MCP documentation. In Claude Code, you attach them with @.
    • ChatGPT has no menu for prompts or resources, so nothing changes in ChatGPT.

    See Prompts and guides in the connector menu.

  • 5011294: Plugins get one typed answer from a fixed list, with no chat reply to parseSection titled “Plugins get one typed answer from a fixed list, with no chat reply to parse”

    A plugin author who needed one option from a fixed list, for example the visualization component for a report, had to prompt a chat model and parse its text reply. Now a plugin declares { "id": "ai", "models": ["decision"] } in plugin.json and passes ctx.credentials.ai.decision to experimental_decide from the ai 7.x package. The call answers typed questions (choice, score and boolean) and returns the selected option, with probabilities where the route gives them. Existing OpenAI, Gemini and Anthropic AI Credentials now provide decision with the recommended model, so a plugin that declares it sends its data to that provider with the key of the Credential. To turn it off, a maintainer sets Decision model to Not configured in the Credential form.

    The AI models section of an OpenAI Credential form with Fast, Reasoning, Embedding and Decision model rows, and the Decision model menu open on Use recommended model, Override model ID and Not configured

    • Credentials. OpenRouter and openai-compatible Credentials do not provide decision. Anthropic Credentials return no probabilities for choice and score. Setting Embedding model or Decision model to Not configured now removes the mapping. Before, the Credential kept it. An installed plugin that declares the removed model can no longer use this Credential for any model until the maintainer maps the model again or the member selects another Credential, see AI mappings.
    • OWOX AI. OWOX AI serves decision only when the operator sets AI_PLUGIN_DECISION_PROVIDER, AI_PLUGIN_DECISION_BASE_URL, AI_PLUGIN_DECISION_API_KEY and AI_PLUGIN_DECISION_MODEL. The operator must choose a model with zero data retention where that matters. The state is text or JSON only. On OWOX Data Marts Cloud, each successful OWOX AI decision call is billed as an AI Process Run under the existing rules. A self-managed deployment does not bill it.
    • Limits. A call has 1 to 64 questions, and each question needs instructions. A choice question needs 2 to 255 options, and a score question needs 2 to 10 levels with a description each. The box question type is not supported. A call that breaks one of these rules gets 400 on every route.

    See Choose from a fixed list, AI mappings, OWOX AI and Environment variables.

  • @owox/internal-helpers@0.44.0
  • @owox/idp-protocol@0.44.0
  • @owox/idp-better-auth@0.44.0
  • @owox/idp-owox-better-auth@0.44.0
  • @owox/backend@0.44.0
  • @owox/web@0.44.0

0:16 - Your AI Assistant Sets Up OWOX
1:12 - Connector Setup in Fewer Steps
1:28 - LinkedIn Ads Spend by Location
1:43 - And a Few More Things

  • 70d2ce1: Set up and maintain Data Marts from your AI assistantSection titled “Set up and maintain Data Marts from your AI assistant”

    Data Owners preparing data for their team previously had to leave their connected AI assistant to create and configure each Data Mart in OWOX. Now they can manage Data Marts through MCP, from draft setup and connector selection to publication, relationships, and recurring runs.

    The new tools are available through a custom MCP connection and use your existing project and resource permissions. Connectors use saved Credentials. Authorization and secret entry stay in the OWOX browser flow. See Data Mart setup with MCP for the setup workflow and available tools.

  • 47370ba: Manage integration access with your AI assistant and keep secrets out of chatSection titled “Manage integration access with your AI assistant and keep secrets out of chat”

    Data Owners previously had to manage provider access separately when setting up integrations with an AI assistant. The assistant can now find and reuse Credentials, the access details saved in OWOX, and guide you through adding or updating them.

    To save or reconnect provider access, you enter secrets in a dedicated OWOX form. After you finish, the assistant can check the result and continue setup. You can reuse the saved access for Google Sheets, Google Chat webhooks, and custom connector tests without entering secrets again. The form checks your project and account before you make changes. New Credentials start unshared, and existing permissions still apply.

    These tools are currently available through the custom MCP connection flow. They will become available in public agent directories later. See Credential management for setup details.

  • 90d0bb5: Set up and manage Storages and Destinations with AI agentsSection titled “Set up and manage Storages and Destinations with AI agents”

    Data Owners previously had to open the web UI to configure Storages or change Destination settings while working with an AI agent. AI agents can now create, configure, and manage Storages and Destinations in OWOX Data Marts through MCP.

    The new tools are currently available only through a custom MCP connection. They will become available through AI agents’ public connector directories later.

    See Storage management and Destination management for supported operations and setup details.

  • 8ffe6e3: See which Credentials a plugin uses, and change them on the plugin’s Credentials tabSection titled “See which Credentials a plugin uses, and change them on the plugin’s Credentials tab”

    Members who reopened Configure Credentials on an installed plugin always saw an empty choice, so a saved Credential looked unsaved, and they had no way to check which Credential the plugin was using. Now an installed plugin that needs Credentials has a Credentials tab next to Overview. It shows the Credential granted to each requirement, already selected, and members change it right there and press Save, or Discard to go back. There is no separate dialog. A Credential that has since been disabled or deleted is marked as no longer usable and has to be chosen again. Configure Credentials no longer sits in the page header.

    The video shows the difference: before, the dialog showed nothing selected even after saving; now the tab shows the Credential the plugin uses, keeps a change only on Save, and still shows it after a reload.

    When a plugin’s AI request fails at the provider, the plugin still gets “AI provider request could not be completed”. The server log now also records which Credential and model failed, the provider’s HTTP status, and, for the built-in AI providers, what the provider answered, so the cause can be found. Timeouts and refused connections are recorded the same way. The prompt and the Credential’s secret are never written to the log.

  • a83952a: Parallel reads of a SQL Data Mart on Google BigQuery no longer fail with a rate limit errorSection titled “Parallel reads of a SQL Data Mart on Google BigQuery no longer fail with a rate limit error”

    A plugin page with many cards, such as Gyroscope, reads one SQL Data Mart many times at once. HTTP Data clients and the MCP query_data_mart tool can do the same. Previously some of these reads failed with Exceeded rate limits: too many table update operations for this table (HTTP 424), and others came back without totals. The cause: every read rewrote the Data Mart’s view in Google BigQuery, and Google BigQuery allows 5 updates of one table in 10 seconds. Now a read rewrites the view only after the Data Mart’s SQL changes, so parallel reads return their data and totals.

    Only SQL Data Marts on Google BigQuery and Legacy Google BigQuery change. Snowflake, Databricks, AWS Redshift and AWS Athena are not affected.

  • 160d56a: Plugin AI calls no longer fail with a timeout after 30 secondsSection titled “Plugin AI calls no longer fail with a timeout after 30 seconds”

    A plugin that uses an AI Credential, such as Gyroscope when it rebuilds its model, can wait more than 30 seconds for a reasoning model on a large project. Previously such a call failed with HTTP_ERROR 408 Request timeout after 30 seconds, and the plugin showed that the AI model did not answer. The plugin SDK and the server wait 120 seconds for an AI call, but a general 30-second request limit stopped it first. Now a plugin’s AI call gets the 120 seconds, as Handle request errors describes.

    A Credential fetch still times out after 30 seconds.

  • bd757b0: A new connector Data Mart starts loading data right after setupSection titled “A new connector Data Mart starts loading data right after setup”

    Before this release, Save in the connector sheet only saved the settings. A new Data Mart stayed a draft, and no data loaded until you found Publish & Run Data Mart. Now a draft Data Mart gets a “Save & Run” button as the last button of the Set Up Connector, Edit Connector and Connector Fields sheets. See Set up a connector Data Mart.

    • For a draft Data Mart, “Save & Run” saves the settings, publishes the Data Mart and starts the first run. “Save” is in the menu next to it.
    • For a published Data Mart, “Save” is the last button. “Save & Run” is in its menu. Save & Run saves the settings and starts a run.
    • The sheet shows only “Save” if the Storage is not set up, a run is in progress, or Target Setup shows that another Data Mart writes to the same table.
  • bd757b0: Connect a Google Sheets plan with one Google sign-inSection titled “Connect a Google Sheets plan with one Google sign-in”

    Before this release, Sign in with Google failed on the Credentials page for a Sheets Source Google account. The error was “Failed to open popup window”. When the popup opened, Google asked for access a second time to choose a spreadsheet. Now one sign-in adds the account, and the spreadsheet picker opens immediately.

    The video connects two plans to Data Marts with one Google account. The second plan uses ”+ Add spreadsheet” in the Data Mart form, and the form shows a warning when another Data Mart already writes to the same table.

    • The spreadsheet list in the Data Mart form has a ”+ Add spreadsheet” item at the end. You can add a new plan without leaving the form. A Google window opens and closes by itself, then the spreadsheet picker opens.
    • To add spreadsheets, sign in to Google with the account of the Credential. If the form shows “Sign in to Google as … to add spreadsheets.”, click ”+ Add spreadsheet” again and choose that account.
    • If a Credential shows no Google account, reconnect it to add spreadsheets.
    • When you choose the spreadsheet from the list or with ”+ Add spreadsheet”, the suggested table name has the document name and the sheet name, for example budget_2026_plan. Two documents with a “Plan” sheet then do not get the same table name.
    • When you type the spreadsheet URL, or the Credential is a service account, the suggested table name has the sheet name only, for example plan.
    • The form suggests a table name only when the target table is empty.
    • Ukrainian and Russian names change to Latin letters, for example data_soft_potochni_vytraty. A table name has 100 characters or fewer. Tables of existing Data Marts keep their names.
    • Target Setup shows a warning when another Data Mart writes to the same table. Each run replaces the table, so the two Data Marts overwrite the data of each other.
    • If the spreadsheet list cannot load, the form shows “Could not load the list of spreadsheets.” and the cause from Google Sheets. You can still type the spreadsheet URL manually.
  • 80c62f9: Compare LinkedIn Ads results across markets: spend by country, region, and countySection titled “Compare LinkedIn Ads results across markets: spend by country, region, and county”

    Teams that run LinkedIn campaigns in several countries could not split LinkedIn spend by location in their warehouse: the LinkedIn Ads connector imported metrics only by creative, campaign, and account, so a market-by-market comparison had no LinkedIn numbers. Now three new endpoints, Ad Analytics by Country, Ad Analytics by Region, and Ad Analytics by County, import each campaign’s daily impressions, clicks, spend, and conversions per location, with the location’s English name (such as United States) next to its LinkedIn URN.

    • LinkedIn breaks location down for one campaign at a time, so these endpoints make more API requests than Ad Analytics: at least one for every campaign that ran that day, and one more per campaign for every further 18 metrics you select.
    • LinkedIn drops a location with fewer than 3 events, so totals can be lower than in Ad Analytics, and location data arrives 12–24 hours late; the default Reimport Lookback Window of 2 days picks it up on the next run.

    See spend by country, region, or county in the LinkedIn Ads guide.

  • 9a765e4: Plugins show loading in the same bar as OWOX Data MartsSection titled “Plugins show loading in the same bar as OWOX Data Marts”

    Plugin authors had to build their own loading indicator, so each plugin showed loading in a different way. Now a plugin asks the host to show the loading bar that OWOX Data Marts shows at the top of the page for its own requests.

    • ctx.ui.startLoading() shows the bar and returns a loading with setProgress(fraction) and stop().
    • The bar stays while any started loading runs. When every running loading sets a fraction, the bar shows how much of the work is done.

    The change is backward compatible. To use the new member, update @owox/plugin-sdk to this release. On an older OWOX Data Marts deployment, the bar does not show and the calls do not fail.

    See Show the loading bar.

  • c875200: Set up a connector Data Mart from another one again, with its Credential and accountSection titled “Set up a connector Data Mart from another one again, with its Credential and account”

    Since connector keys moved to Credentials, a new connector Data Mart could take only the Credential: account IDs and the other settings had to be looked up in an existing Data Mart and entered by hand. Reuse from another Data Mart is back in the connector settings. Pick a Data Mart with the same source, and the new configuration uses the same Credential and gets the same settings, such as account IDs. No key is copied: both Data Marts use one Credential, so replacing its values updates both.

    • If the selected Data Mart’s Credential is not available to you, only the settings are copied, and you select a Credential yourself.
    • The list shows only Data Marts you can open.

    See Add access Credentials.

  • @owox/internal-helpers@0.43.0
  • @owox/idp-protocol@0.43.0
  • @owox/idp-better-auth@0.43.0
  • @owox/idp-owox-better-auth@0.43.0
  • @owox/backend@0.43.0
  • @owox/web@0.43.0
  • 1371345: Count unique records when reading a Data Mart over the HTTP Data APISection titled “Count unique records when reading a Data Mart over the HTTP Data API”

    Plugins and scripts that read a Data Mart through the HTTP Data API, or through dataMarts.traverseData() in @owox/api-client and @owox/plugin-sdk, can now ask for a Unique Count, the same one reports already offer. Before, COUNT_DISTINCT worked only on text columns, so a numeric or composite primary key, or a joined Data Mart’s records, could not be counted at all.

    Pass uniqueCount with the sources to count: '' for the Data Mart itself, or the alias path of a joined Data Mart. Each source counts distinct values of its primary key, whatever the key type. Without a column selection you get a single total row; with columns you get a count per group. You can also sort by the Unique Count column, as in reports. See Stream Data Mart rows for details.

  • @owox/internal-helpers@0.42.0
  • @owox/idp-protocol@0.42.0
  • @owox/idp-better-auth@0.42.0
  • @owox/idp-owox-better-auth@0.42.0
  • @owox/backend@0.42.0
  • @owox/web@0.42.0

0:11 - Your AI Assistant Sets Up Contexts and Connectors
0:39 - Build a Data Mart on Top of Another
1:06 - Rotate a Key Once, Everywhere It’s Used
1:27 - And a Few More Things

  • f66ac98: Safer sign-in and request validation: Better Auth and zod security fixesSection titled “Safer sign-in and request validation: Better Auth and zod security fixes”

    Magic links sent before the upgrade and not yet used stop working: send those users a new one. Self-hosted deployments ran dependency versions with published security advisories, among them a Better Auth flaw that let an OAuth state value work as a magic link and sign someone in as another user (GHSA-965c-763c-88jm, critical), and a zod flaw where validating an oversized collection could exhaust memory (GHSA-mrhx-c7xc-m3w5). This release moves to versions with the fixes.

    • Magic links. Better Auth 1.7 stores magic links in a new format, so an unused link sent before the upgrade returns an invalid-token error. Links expire after an hour anyway; an admin can send a new one, see Adding New Users. A primary admin without a password gets a new link in the server logs on the next start.
    • MySQL. On start, the application widens verification.id in its sign-in tables to VARCHAR(255): Better Auth 1.7 writes longer ids there than its own MySQL migrations create, and without the change an invited user’s first sign-in fails. The database user needs the ALTER privilege on that table.
    • API validation errors. 400 responses word default validation messages differently, for example Invalid input: expected string, received undefined instead of Required, and raw issues under errorDetails.errors follow zod 4: nested issues sit under errors instead of unionErrors, and enum and literal mismatches are invalid_value. Custom messages are unchanged. Update any script that matches on this text.
    • zod 4. owox, @owox/idp-protocol, @owox/idp-better-auth and @owox/idp-owox-better-auth now declare zod 4 as a peer dependency.
  • b9c59d7: Dates from Google Sheets arrive as dates, not textSection titled “Dates from Google Sheets arrive as dates, not text”

    A planner who keeps daily targets in Google Sheets got the date column in the warehouse as text, such as 10/1/2026. Date filters and date grouping did not work on it, and setting the field type to DATE in the Output Schema only added a warning: the next upload wrote text again.

    • In a new Data Mart, a column of dates (cells formatted as Date in the sheet) is stored as DATE from the first run, with no setting to turn on. The Output Schema then shows it as DATE.
    • Existing Data Marts keep their columns as they are. To store a column as a date, set its type in the Output Schema to DATE: the next run, or the next Upload from the Extension, writes dates and the warning goes away.
    • In a DATE column, a date-time cell keeps its day, and any other cell that is not a date is imported as an empty value. Run History lists the rows to fix, so one stray value never turns the column back into text. A column with no date at all stays text, and Run History says so.
    • In other columns, dates typed as text, date-times, times, durations, and dates mixed with other values stay text.

    See the Google Sheets source guide.

  • 071320f: Set up Contexts without leaving your AI assistantSection titled “Set up Contexts without leaving your AI assistant”

    Project admins setting up Contexts and the project description previously had to switch from their AI assistant to OWOX Data Marts. They can now make those changes within the conversation through MCP.

    Your assistant can create Contexts, update their names, descriptions and icons, inspect their usage, assign members, and delete detached Contexts. It can also list project members to find the right people for assignments and read project settings. Admins can update or clear the project description.

  • 8cb8346: Build and maintain custom connectors from your AI assistantSection titled “Build and maintain custom connectors from your AI assistant”

    Data Owners previously needed the Connector Builder or HTTP API to save, test and publish custom connectors for HTTP APIs. They can now create drafts, test and publish connectors in their connected AI assistant through MCP. The tools also support metadata updates, version inspection and rollback to an earlier published version.

    These tools are currently available through the custom MCP connection flow. They will become available in public agent directories soon.

    The tools apply project permissions and existing checks for Data Marts that use the connector. See Custom connector authoring for the available tools. Use the manifest reference to write a connector manifest.

  • 1619442: Shared SQL logic stays in one place: build a Data Mart on top of anotherSection titled “Shared SQL logic stays in one place: build a Data Mart on top of another”

    An analyst who needed a narrower version of an existing Data Mart — retail orders out of all orders — had to copy its SQL into a new Data Mart. When the original changed, the copy didn’t, and the two started reporting different numbers. Now a SQL Data Mart can reference another Data Mart of the same Storage in its FROM clause, so the logic lives in one place and every Data Mart built on it follows its changes.

    The video builds a Completed orders Data Mart on top of Shop orders: it writes only the filter, takes the business names from Shop orders, and its report’s SQL still reaches the warehouse table.

    Type @ in the SQL editor to pick a published Data Mart; it’s inserted as [Orders](dm:<id>). When the query runs, OWOX puts that Data Mart’s own SQL in place of the reference, so the warehouse receives plain SQL and Preview SQL of a report shows it down to the warehouse tables. People who read the new Data Mart don’t need access to the one it references.

    The SQL editor with SELECT o.order_id, o.user_id, o.created_at, o.num_of_item FROM @: the @ picker is open, searching "Shop", with the note that only Data Marts from the same Storage can be referenced and the published Data Mart Shop orders offered

    • A column that keeps its name takes the referenced Data Mart’s alias and description, read-only, so business names are written once and stay in step. Primary keys aren’t taken over.
    • Changing the input source of a referenced Data Mart, or deleting it, now lists the Data Marts that reference it, so you can check them first.
    • Available on Google BigQuery, AWS Athena, Snowflake, AWS Redshift and Databricks, not on Legacy Google BigQuery storages. See Reference another Data Mart.
    • API: GET /api/data-marts accepts storageId, and GET /api/data-marts/{id}/input-source-change-impact also returns referencingDataMartsCount and referencingDataMarts.
  • a3766e5: Rotating a connector key is one change: Sources share reusable CredentialsSection titled “Rotating a connector key is one change: Sources share reusable Credentials”

    Before you upgrade, check your custom connector templates: a template path no longer steps into text. {{ parameters.Key.0 }} or {{ record.name.length }} is now unresolved. That fails the run wherever an unresolved path does, such as a request path, a request body or an authentication field; in a query parameter or a header, the parameter is not sent. See Templating scopes.

    When a connector key has to be replaced, for example a token about to expire while the colleague who created it is on vacation, whoever maintains it used to replace the same key in every Connector Data Mart, which can be dozens or even hundreds of them, and a copy they missed broke the next run. Custom connectors had the same problem, and testing one in the builder meant typing its key into the browser. Now a configuration selects a reusable Credential instead of keeping its own key: the Credential lists every Data Mart that uses it, replacing its secret applies to all of them, and switching it off stops them all at once. The video follows this for a Facebook Ads token that is about to expire while its creator is on vacation.

    Existing configurations are converted once, on the first start of this version (or when you run migrations), and keep running. If you run several instances, upgrade all of them together: an instance on an earlier version cannot run a configuration saved on this one, and keeps using a key after its Credential is replaced. The member who connected a connector’s own sign-in owns its Credential, even if they have since left the project; a Project Admin can change its owners. Every other Credential, and a sign-in with no recorded member, is owned by the technical owners of the Data Marts that use it, or by their business owners when a Data Mart has no technical owners. Converted Credentials are Not shared: an owner has to share one for use before others select it elsewhere. A configuration that cannot be converted keeps running on its own key until you select a Credential, and the migration log names it.

    • The Data Mart’s connector configuration has a Credential field in place of the key fields: select a fitting Credential, or click + Create new Credential to add one beside the configuration.
    • The Credentials page gains a provider for each connector with a key, in the groups Ads, Social, E-commerce, Currencies and Custom connectors, and lists the Data Marts each Credential is used by. Sheets Google account is now Sheets Destination Google account, next to the new Sheets Source Google account.
    • Changing an identity field of a Credential (a Shopify shop domain, a Reddit or Criteo client ID, an AWS access key ID, a Snowflake username) asks for its secrets again, so a Credential never mixes an old secret with a new identity.
    • With a Sheets Source Google account Credential, the Data Mart form lists the spreadsheets that Credential can open. Self-managed installs enable the Google Drive API in the Cloud project of the OAuth client for this list.
    • In the builder, Test settings switches between Enter values and Use a Credential; with a Credential, the browser never sends or receives the stored values, and the output masks those values, not what an API returns for them. Keys from a Credential work only in requests, and publishing a version that Credentials no longer fit lists the Data Marts that will stop until someone selects a fitting Credential.

    Copy credentials from another Data Mart is removed: select the same Credential in both Data Marts. API clients: a connector configuration takes _credential_id; new key values, sourceDataMartId and _copiedFrom are rejected with This page is out of date. Reload it. POST /connectors/custom/test also takes credentialId, which supplies the key values.

    See Credentials, Connector Builder and Key parameters.

  • e8eda7f: Rotating a warehouse key is one change: Storages share reusable CredentialsSection titled “Rotating a warehouse key is one change: Storages share reusable Credentials”

    When a warehouse key has to be rotated, for example after someone who had it leaves the company, whoever maintains it used to open every Storage to find the ones holding a copy, and a copy they missed broke once the old key was revoked. Now a Storage selects a reusable Credential instead of keeping its own key: the Credential lists every Storage that uses it, replacing its secret applies to all of them, and switching it off stops them all at once.

    Existing Storages are converted once, on the first start of this version (or when you run migrations), and keep running; no action is needed. Every owner of a Storage owns its Credential, Business Users included: they can edit, share and delete the Credential, but still cannot run or edit the Storage. Converted Credentials are Not shared, whatever the Storage’s sharing: members who run or maintain a Storage without owning it keep doing so, but an owner has to share the Credential for use before they change where the Storage connects (project, region, account or host) or reuse its key elsewhere, and for maintenance before they replace its key. Identical keys of Storages with the same owners become one Credential, which keeps only the Contexts all its Storages shared. A Storage whose key cannot be converted keeps running on its own key until you select a Credential.

    The Storage form now has a Credential field in place of the authentication method and key fields: select a Credential that fits the Storage type, or click + Create new Credential to add one beside the form and select it on save. With exactly one fitting Credential, a new Storage selects it for you. The Credentials page gains Storage providers (BigQuery Service Account, BigQuery Google account, AWS access key, Snowflake password, Snowflake key pair, Databricks token), lists the Storages each Credential is used by, and does not delete a Credential that a Storage uses. The video follows a key pasted into a team chat: the Credential shows which Storages use it, where to replace it once, and where to switch it off.

    Copy credentials from another storage is removed: select the same Credential in both Storages. PUT /data-storages/:id takes reusableCredentialId, and the old credentialId, credentials and sourceStorageId fields are rejected. Storage responses no longer include credentials or credentialId.

    See Credentials for Storages and Credentials.

  • 21a3f01: Reuse one Google Sheets or Google Chat key across DestinationsSection titled “Reuse one Google Sheets or Google Chat key across Destinations”

    Existing Google Sheets and Google Chat Destinations are converted once, on the first start of this version (or when you run migrations), and keep running; no action is needed. A Google account Destination whose stored sign-in lacks a permission the Sheets Google account Credential requires stops until you reconnect its Credential; the migration log names each one. Every owner of a Destination owns its Credential. Converted Credentials are Not shared, whatever the Destination’s sharing: members who maintain a Destination without owning it can still use and edit it, but reuse its key elsewhere only after an owner shares the Credential. Identical keys of Destinations with the same owners become one Credential. A Destination whose key cannot be converted keeps running on its own key until you select a Credential.

    Each Google Sheets or Google Chat Destination used to keep its own service account key, Google account or webhook URL. Using one key in two Destinations meant copying it, and rotating a key, for example after someone who had it leaves the company, meant finding and editing every copy. The Destination form now has a Credential field: select a Credential that fits, or click + Create new Credential to add one beside the form; the form then selects the new Credential. With exactly one fitting Credential, a new Destination selects it for you. Replacing a Credential’s secret updates every Storage and Destination that uses it, as the video shows.

    • Google Sheets Destinations use a Google Service Account or a Sheets Google account Credential. The Google Drive permission (drive.file) is optional for a Sheets Google account: without it, reports still write, and Create document still creates the spreadsheet, but in the account’s Drive root, outside the selected folder and not shared with you.
    • Google Chat Destinations with webhook delivery use a Google Chat webhook Credential. Channel email delivery keeps its addresses on the Destination.
    • The Credentials page gains the Sheets Google account and Google Chat webhook providers, lists the Destinations each Credential is used by, and does not delete a Credential that a Destination uses. The BigQuery Service Account provider is now called Google Service Account, because one key serves a BigQuery Storage and a Google Sheets Destination.

    Copy credentials from another destination is removed: select the same Credential in both Destinations. API clients: POST /data-destinations and PUT /data-destinations/:id take reusableCredentialId for Google Sheets and Google Chat webhook delivery and reject the old credentialId, sourceDestinationId and key credentials fields. The Destination Google OAuth routes are removed: POST /data-destinations/oauth/authorize, POST /data-destinations/oauth/exchange, GET /data-destinations/oauth/credential-status/:credentialId, POST /data-destinations/:id/oauth/authorize, GET /data-destinations/:id/oauth/status and DELETE /data-destinations/:id/oauth; connect a Google account from the Credential instead. Destination responses no longer include credentialId for these types; credentials keeps only the public parts, as before: the service account email and project, the Google account, or that a webhook is set.

    See Google Sheets, Google Chat and Credentials.

  • 9dddd2d: Finding the right Data Mart is quicker: the list keeps your search when you come backSection titled “Finding the right Data Mart is quicker: the list keeps your search when you come back”

    When you look for one Data Mart among many, you type part of its name in the Data Marts list and open a match. If it was the wrong one, for example an SQL Data Mart named like the connector Data Mart you needed, going back showed the whole list again, and you typed the search once more for every wrong guess. Now the list keeps the search when you come back in the same browser tab, through the back arrow, the browser’s Back button or the sidebar.

    The Data Marts list searched for facebook: three Facebook Ads Data Marts are listed, and the search field ends with a × that clears it

    • The search is kept per project and only in the current tab: a new tab opens the full list.
    • A × in the search field clears the search in one click. The Connectors, Destinations, Storages, Contexts and Members lists get it too; their search still starts empty each time.

    See Find a Data Mart in the list.

  • acd2a73: See on one screen who has access to a Data Mart, who doesn’t, and whySection titled “See on one screen who has access to a Data Mart, who doesn’t, and why”

    Project Admins decide which members see each Data Mart and what they can do with it. Now the Data Mart’s Overview tab shows the result in one place: every project member, whether they have access, what it comes through, and what to change to give them more — for example, To see it: add Paid ads in Contexts.

    Watch how a Project Admin sees why a business user has no access to a Data Mart and gives it on the same page.

    • Each member’s row shows their role, what their access comes through — Ownership, Shared for reporting, Shared for maintenance, or a shared context — and what they can do: see the Data Mart and build reports on it, edit it, or change its Sharing and Ownership.
    • Hints link to where the change is made: Project settings → Members opens the member’s details, and a context’s name opens that context.
    • The Sharing and Contexts sections show who each option and each context lets in, the way Ownership shows owners, and warn how many members a setting keeps out, such as a Data Mart without a context that members limited to selected contexts can’t see.
    • The list follows the rules the product enforces and is visible to Project Admins only; see Ownership and Sharing.
  • a2029a1: Deleting a context no longer gets stuck on attachments you cannot seeSection titled “Deleting a context no longer gets stuck on attachments you cannot see”

    A Project Admin cleaning up Contexts could not delete one: the dialog said it was attached to several Members or to a Data Mart, but there was nothing to detach. Those members showed Entire project scope and no contexts in Project Settings → Members, and the Data Marts link opened an empty list. A member could be freed only by switching them to Selected contexts only, unticking the context and switching back; a deleted Data Mart could not be freed at all. Both kept a link the admin could not see: a member switched to Entire project, or made a Project Admin, keeps their context assignments so they come back if the member is switched back, and a deleted Data Mart, Storage or Destination keeps its contexts. Now the delete counts only what the admin can see and detach.

    • Members in the delete dialog, and memberCount in GET /api/contexts/{id}/impact, count only non-admin members with Selected contexts only scope.

    • Data Marts, Storages and Destinations in the dialog, and their counts in the same endpoint, no longer include deleted ones.

    • Assignments kept by Project Admins and Entire project members, and the contexts of deleted Data Marts, Storages and Destinations, are removed together with the context.

    • The Members column in Project Settings → Contexts lists Project Admins and Entire project members on every context, as the context panel does, instead of only on contexts they kept from an earlier scope. In Project Settings → Members, Project Admins show no contexts.

      Project Settings → Contexts with Finance, Marketing and Sales: every row's Members column shows the same admins and Entire project members, and the hover card on Finance's +3 lists Marco Rossi, Olivia Park and Tom Becker

    See Deleting a Context.

  • 4e6ef21: Clearer warning when a report counts a field from a joined Data MartSection titled “Clearer warning when a report counts a field from a joined Data Mart”

    When you tick Count on a field from a joined Data Mart, the report editor warns that the number counts your report’s Data Mart rows, not the joined Data Mart’s. That warning was one dense paragraph, so it was hard to tell what it meant or what to do, and a count of customers that have an order could still pass for a count of orders. Now it opens with what the number counts, such as “Counts Customers rows, not Orders rows”. One sentence with the real Data Mart names says whether it can come out higher or lower than the joined Data Mart’s row count, and the last line says what to pick instead. When a missing primary key is the reason, the warning names the Data Mart that needs one and links to its Output Schema, which opens in a new tab so the report you are editing stays as it is. Once you set the key there and come back, the warning updates without reopening the report.

    The aggregation popover for the column name from the joined Customers Data Mart, with Count ticked and the warning "Counts Companies rows, not Customers rows": Companies has no primary key, so the count can be higher than the number of Customers rows; set a primary key on Companies, or use Count Unique or the Unique Count of Customers

    Data Mart names and aggregation options are in bold, and field names are in monospace as in the column list, so Count Unique, which counts distinct values of this field, no longer blurs with Unique Count, the joined Data Mart’s own row count. When the warning appears is unchanged, and it still blocks nothing. See Counting rows in a group.

  • c2a758b: Correct client address behind a load balancer that adds two forwarding entriesSection titled “Correct client address behind a load balancer that adds two forwarding entries”

    Behind some load balancers, such as a Google Cloud external Application Load Balancer, the application saw the load balancer’s own address as the client address, because it trusted exactly one proxy while the load balancer appends two X-Forwarded-For entries: the client and itself. Set the new TRUST_PROXY_HOPS variable to the number of entries your proxies append, 2 in that case, and the real client address is used.

    The default stays 1, so a deployment that does not set the variable behaves as before. The value is a number from 0 to 10; anything else stops the application at startup, and a value larger than the real number of proxies would let clients choose their own address. 0 is only for clients that connect directly, with no proxy in front. See Reverse proxy.

  • @owox/internal-helpers@0.41.0
  • @owox/idp-protocol@0.41.0
  • @owox/idp-better-auth@0.41.0
  • @owox/idp-owox-better-auth@0.41.0
  • @owox/backend@0.41.0
  • @owox/web@0.41.0

0:10 - Read and Copy Business Descriptions
0:30 - Find Data Marts Missing a Context
0:48 - Plugins Run on a Schedule
1:08 - An Icon for Each Context
1:25 - Readable Diagrams
1:43 - And a Few More Things

  • 9dc6e5f: Warnings when a Count reads a joined Data MartSection titled “Warnings when a Count reads a joined Data Mart”

    Count on a joined field counts the rows of the report’s Data Mart that found a match, not the joined Data Mart’s own rows. The aggregation popover in the report editor and in the Google Sheets extension now says so while Count is ticked: whether the number can come out higher, lower, or cannot be checked because a primary key is missing. It points to Count Unique, and to the joined Data Mart’s Unique Count when that Data Mart offers one. Nothing is blocked, and no warning appears when every join on the way is keyed by a primary key on both sides.

    The aggregation popover for the joined column order_id from Orders, with Count ticked and a warning that several Orders rows can share one join key value, so the count can be lower, pointing to Count Unique and the Unique Count of Orders

    The formula editor gives the same advice for a calculated field that counts a joined field, and now reads the primary key as you have set it in the Output Schema, saved or not. Previously it kept saying the key was missing until you pressed Save, and the save then answered differently. Clearing the key works the same way.

    The Output Schema editor with the primary key ticked on session_id but not saved, and the formula editor for orders_count = COUNT(orders.order_id) warning that several Orders rows can share one join key value

    The video walks through both on a Sessions Data Mart joined to Orders:

    See Counting rows in a group and A joined COUNT counts this Data Mart’s rows, not the joined one’s.

  • 7ab7631: Bring in the colleague who sets up your project right after sign-upSection titled “Bring in the colleague who sets up your project right after sign-up”

    A new project’s admin can now invite the teammate who will connect the data warehouse and create Data Marts, on a step right after the sign-up questionnaire. Add one or more emails, press Enter after each, pick a role and send the invitations in one go, or skip to the app.

    • Data Owner is preselected. Project Admin and Business User are one click away, and every invitation gives access to the entire project.
    • The step shows once. Later invitations go through Project Settings → Members or the Invite teammates step of the setup checklist.
    • When sign-up started from somewhere else, such as connecting an AI assistant, the step is skipped and that flow continues.

    See Invite Teammates After Sign-Up.

  • f83d878: Open the Google Sheet a scheduled report fills, right from TriggersSection titled “Open the Google Sheet a scheduled report fills, right from Triggers”

    Checking the results of a report-run schedule used to mean opening the row’s info card and finding Open in Google Sheets tucked inside it — an action many people never discovered. On the Triggers page under Data Marts, and in each Data Mart’s Triggers tab, hovering a report-run row now shows the same quick actions as the Reports tables: an Open document button that opens the report’s Google Sheet in a new tab, next to the info button.

    The Data Marts Triggers page with a hovered Report Run row showing an info button and an Open document button, with the Open document tooltip

    • The info button now opens its card on click instead of on hover, for both reports and connectors, and each button names itself in a tooltip.
    • Clicking either button, or anything inside the card, no longer opens the trigger’s edit panel.
    • The Run Target column is wider to fit the report name alongside these buttons.
    • Open document shows only for reports whose destination is Google Sheets.
    • See Report Trigger for how report-run schedules work.
  • 5eee385: Fix loading indicator errors when requests failSection titled “Fix loading indicator errors when requests fail”

    Fixed the Cannot read properties of undefined (reading 'skipLoadingIndicator') error on Data Mart pages and other application pages. This secondary error could appear while handling a failed request and hide its original cause.

  • a9faca2: Invalid HTTP Data queries no longer create Run History entriesSection titled “Invalid HTTP Data queries no longer create Run History entries”

    HTTP Data requests to a Data Mart rejected because of invalid query configuration no longer create failed Run History entries.

    See the HTTP Data API for details.

  • 814b882: Plugins do their work on schedule, also when nobody has them openSection titled “Plugins do their work on schedule, also when nobody has them open”

    A plugin’s backend action used to run only while a member had the plugin open, so a check meant for every Monday at nine waited for someone to open it. Now a plugin can create schedules for its installation with ctx.schedules, and OWOX Data Marts starts the action on its cron expression and time zone with the installing member’s access. The plugin’s page lists its schedules, and members pause, resume and delete them under Plugins → Schedules, which also lists their runs and lets them cancel a queued or running one; a project admin sees the whole project there, and the plugin cannot resume a schedule an admin paused.

    The video adds a Google Doc to a weekly schedule of the example plugin Metric Watch, then opens Plugins → Schedules, the schedule’s run in Run History, and the check that run added to the Doc:

    OWOX Data Marts disables a schedule, with the reason, when the action is no longer declared or the parameters no longer match its schema, and skips its runs while the member is out of the project or their account is locked. A scheduled run that meets the installation’s action limits waits, up to 24 hours, and starts once they allow. In the SDK, PluginActionStatus now includes QUEUED, and an execution’s trigger can be schedule. See Run actions on a schedule. A self-managed deployment can change the schedule limits with five PLUGIN_HOST_* variables, such as PLUGIN_HOST_MAX_SCHEDULES_PER_INSTALLATION; see Environment variables.

    In the SDK, BackendActionContext.schedules, PluginActionExecution.scheduleId, PluginActionExecutionContext.scheduleId and PluginContext.schedules are new members: update test doubles and exhaustive switches.

  • 7d5b78e: Read and copy business descriptions without opening the editorSection titled “Read and copy business descriptions without opening the editor”

    When analysts need to read a business definition or reuse it in an AI assistant, switching to an editor adds an unnecessary step. Project, Context, and Data Mart descriptions now open as formatted text, and Copy as Markdown preserves the definition with its exact Data Mart and field references.

    Expand descriptions to full screen for reading or editing. Return to the inline editor with your draft and Undo history intact. Copy a permanent Context link to share the same business definition with other project members; the link keeps existing access permissions. Existing edit permissions and Save/Cancel actions still apply. See Business descriptions and Data Mart links.

    The demo shows how to keep a business definition and its exact metric reference together when sharing them.

  • 27b4843: Tell contexts apart at a glance: pick an icon for each contextSection titled “Tell contexts apart at a glance: pick an icon for each context”

    A project that groups its Data Marts, Storages and Destinations by many contexts showed every context as a plain-text badge, so finding the Marketing or the Finance resources in a long list meant reading each label. Now a Project Admin picks an icon for each context, from the same set as Data Mart icons, and it appears before the context name: in the contexts list, on the context badges of Data Marts, Storages, Destinations and members, and in every Contexts picker. A context without a picked icon shows the default tag icon.

    Watch a Project Admin give the Marketing and Finance contexts their icons, and the Data Marts list pick them up.

    Scripts that set up contexts through the API can set the icon too. PUT /api/contexts/{id}/icon sets it, or resets it with null, and POST /api/contexts and PUT /api/contexts/{id} accept an optional icon. The contexts list and the contexts listed on Data Marts, Storages, Destinations and Credentials now return icon. MCP tools are unchanged. See Contexts.

  • 91c287a: Organizing a project is faster: find Data Marts missing a context or a relationshipSection titled “Organizing a project is faster: find Data Marts missing a context or a relationship”

    A project admin who groups Data Marts into contexts needs to find the joined Data Marts that are still outside every context: members whose access is limited to selected contexts do not see them. Before, that took two screens and a manual comparison. The Models canvas showed which Data Marts have relationships, one storage at a time and without contexts, and the Data Marts list could show “no context” only by excluding every context one by one. Now the Data Marts list answers it with two filters: Relationships With relationships and Contexts No context.

    The Data Marts list Filters panel with two conditions, Relationships is any of With relationships and Contexts is any of No context, and the list behind it narrowed to 2 rows

    • Relationships filter: With relationships or Without relationships, across all storages. A relationship counts whether the Data Mart defines it or another Data Mart points to it; a relationship with a Data Mart you cannot see does not count.
    • Contexts filter: a new No context option, which you can combine with real contexts.
    • GET /api/data-marts returns relationshipsCount on every list item: the same number as on the Data Mart’s Models canvas card.

    See Find Data Marts With or Without Relationships and Contexts.

  • 78bf519: Connecting a plan from Google Sheets takes fewer guessesSection titled “Connecting a plan from Google Sheets takes fewer guesses”

    A marketer or product manager who keeps targets in a spreadsheet and wants plan-vs-fact answers had to guess their way through the Google Sheets connector: which definition type hides the connectors, whether the Google sign-in actually worked, why the file picker opened without its search box, and why the suggested table name was the same one their previous Data Mart already used. Each guess cost a retry or a second sign-in. This release removes the ones that did not need a bigger change.

    • The Connector definition type shows the logos of the most used sources (Facebook Ads, Google Ads, TikTok Ads, LinkedIn Ads, Google Sheets) instead of a generic description, so a spreadsheet owner sees where to go.

      The Definition Type list with the Connector option showing the logos of Facebook Ads, Google Ads, TikTok Ads, LinkedIn Ads and Google Sheets, followed by "and others"

    • The Input Source card shows the connector by its name from the catalog (“Google Sheets”, not “GoogleSheets”) and, for Google Sheets, the selected sheet with a link to the spreadsheet, so several plan Data Marts can be told apart without opening their settings.

    • After Sign in with Google a confirmation names the connected account and the next step, and the button shows a connected state, so nobody has to spot a changed label.

    • Copy from… is now Reuse from another Data Mart for every connector, with a hint that says what it reuses: the credentials and connection settings of another Data Mart with the same source, so access is not entered or granted twice. For Google Sheets that includes the spreadsheet and sheet, which can still be changed afterwards.

      The Reuse from another Data Mart list open above the connector settings, with the hint on what is copied and a Data Mart to copy from

    • The Google file picker is sized to the browser window (down to Google’s own 566×350 minimum), so its title and search field stay visible without resizing the window.

    • The suggested table name for a Google Sheets source comes from the selected sheet (for example product_engineering) instead of the same google_sheets for every Data Mart, so differently named sheets get different defaults. The dataset stays google_sheets_owox, and a sheet name that leaves no Latin letters or digits still falls back to google_sheets, so check the suggestion before saving: nothing stops two Data Marts from writing to one table.

    See Connector Data Mart and the Google Sheets source guide.

  • 7a942d8: Large relationship diagrams open readable, on the Data Mart you are working onSection titled “Large relationship diagrams open readable, on the Data Mart you are working on”

    When you open a Data Mart, the Graph view of its Joinable Data Marts block shows how it joins to other Data Marts. It used to fit the whole diagram into the block, so on a model with many relationships the cards shrank until their names could not be read, and you had to zoom in and find the Data Mart you had opened. Now a diagram too large to read that way opens on the current Data Mart at a readable size, with the Data Marts it joins right next to it.

    The Joinable Data Marts block in the Graph view, opened on the Sessions Data Mart at a readable size, with Campaigns, Orders and the Data Marts they join next to it and the minimap showing the rest of the diagram

    A diagram that fits at a readable size still opens whole, and Fit to view shows the full diagram, as before. See Transitive Joins.

  • 02550f2: Plugin backend actions can use AI CredentialsSection titled “Plugin backend actions can use AI Credentials”

    A plugin’s backend action now gets the AI Credential through ctx.credentials.ai, as the plugin page does: the models the plugin declares, as the project maps them. Previously only exact Credentials reached a backend, and the AI Credential routes refused its execution token with EXECUTION_TOKEN_NOT_ALLOWED, so a scheduled action that needed AI had to declare an exact provider Credential and call that provider’s API directly.

    • The backend needs @owox/plugin-sdk 0.40.0. A backend on an older SDK keeps getting only exact Credentials, as before.
    • An AI Credential that is not usable is left out of ctx.credentials and does not stop the action.

    See Run code on your backend with actions.

  • @owox/internal-helpers@0.40.0
  • @owox/idp-protocol@0.40.0
  • @owox/idp-better-auth@0.40.0
  • @owox/idp-owox-better-auth@0.40.0
  • @owox/backend@0.40.0
  • @owox/web@0.40.0
  • 9e9168f: See who published a plugin in Plugin historySection titled “See who published a plugin in Plugin history”

    Installation history is now Plugin history, one list, newest first, of what happened to the plugins you can reach in a project: your own installs, removals and restores, every publication for the project with the member who made it, your personal publications, Verified listing changes, version updates and suspensions. All, Installations and Publications filter the list, and Restore stays within reach for every plugin you removed.

    Plugin history with the All, Installations and Publications filters, a Restore button on a removed plugin, and install and publish entries below it

    • Installations stay personal: other members do not see yours.
    • Open it from Plugins → ⋮ → Plugin history. The API is GET /api/plugins/history.

    The video opens Plugin history from Plugins, filters the list, and restores a removed plugin:

    See Plugin history.

  • 57ddb22: Install a plugin as you publish itSection titled “Install a plugin as you publish it”

    Publish & Install in the Publish a plugin sheet lists the plugin, asks you to confirm the installation, and opens the plugin, so it is in your menu right away. Previously a published plugin was missing from your menu until you found it in the Gallery and installed it.

    • Publish only, in the button’s menu, lists the plugin without installing it.
    • Publish & Install on a plugin you already have opens it.

    The video publishes and installs a plugin, publishes it again when it is already installed, and lists a second one with Publish only:

    See Publish in OWOX Data Marts.

  • b23b1c5: Reload a year of history in one backfill, not twelveSection titled “Reload a year of history in one backfill, not twelve”

    An analyst who adds an ad account, or fixes a field that started collecting late, needs the Data Mart’s whole history back — and used to get it a month at a time: a manual backfill rejected any period longer than 31 days, so a year meant starting twelve backfills by hand and waiting for each one to finish before queueing the next. Now you pick any past start and end date in Manual Run, and OWOX runs the period as a chain of runs, one segment of up to 31 days each.

    The video picks a 90-day period, confirms the three runs it costs, and follows the segments through Run History:

    Each segment counts as one run. Before the backfill starts, the form shows how many days the period covers and how many runs that is, and asks you to confirm from two segments up; a period over a year also notes that it can take hours. A segment is counted only once one of its accounts loads data — a segment that loads nothing, or is cancelled or interrupted, is not.

    • Every segment is its own run in Run History, the log of every run of a Data Mart, showing its period and position, such as “segment 3 of 18”.
    • A failed segment never stops the rest of the backfill: OWOX attempts the whole period, and the failed run names the last day it loaded.
    • Cancelling the active run stops the whole backfill. Finished segments stay loaded and keep their history.
    • While a backfill is running, OWOX starts no other run of the same Data Mart, including its scheduled runs.
    • If the project becomes restricted, the backfill stops; the last run names the segments that did not load, and OWOX does not resume them when the restriction lifts.
    • The start date cannot be earlier than 2000-01-01, which catches a typo such as a wrong year.

    See Backfill a custom period.

  • 0cd64e4: Long Facebook Ads imports no longer fail on a corrupted responseSection titled “Long Facebook Ads imports no longer fail on a corrupted response”

    A Facebook Ads import of a large account could fail near its end with SyntaxError: Unexpected non-whitespace character after JSON at position 88743, after it had already loaded thousands of rows. Facebook had answered one page request with a damaged body, and the connector gave up on that page immediately: Max Fetch Retries never applied to this error, so the only way forward was to start the import again by hand and hope the same page came back clean.

    A damaged body now counts as a retriable failure, like a dropped connection. The connector re-requests the page, waits between attempts exactly as your retry settings say, and the import finishes. A page fails the run only when every attempt in the budget comes back damaged, and the error then names the endpoint that sent it and how large the body was.

    • Nothing to change in your connector setup: Max Fetch Retries and Initial Retry Delay (ms) already cover this case.
    • Re-run any Facebook Ads import that failed with a JSON parsing error.
    • This covers the paginated data requests of the Facebook Marketing source; other sources are unchanged.
  • 3d82afb: Tell your AI assistant which Data Mart field to useSection titled “Tell your AI assistant which Data Mart field to use”

    Project, Context, and Data Mart descriptions can now link to a specific output-schema field, including calculated fields, rather than naming a metric in plain text. Type @, select a published Data Mart by title and icon, and search its fields by alias or technical name. The link gives your AI assistant an exact field reference to verify against the schema.

    The video shows a business description linked to a metric, navigation to that field, and the editor in all three description locations:

    • Opening a field link expands Output Schema and any parent records, scrolls to the exact field, and highlights it. If the field was renamed or removed, the link still opens the Data Mart normally. Saved labels and technical names remain as inserted.
    • Escape returns from fields to the Data Mart list, then closes the picker while preserving the description draft. You can continue typing literal @ text after dismissal; Tab also closes the picker when focus leaves it.
    • Description editors show Data Mart icons in Preview, more space above the first line, and a single Markdown label. Links do not change context assignments or access permissions.

    See Business descriptions and Data Mart links.

  • 43e82de: Developing plugins is easier: find release problems before you publishSection titled “Developing plugins is easier: find release problems before you publish”

    A plugin author used to learn that a release broke a rule only after creating the GitHub Release, when release sync rejected it. Now you check a branch, tag, or commit first and fix the problems it finds in one pass. owox-ctl plugins check OWNER/PLUGIN_NAME --ref <branch, tag, or commit SHA> runs the release rules a GitHub Release from that commit would face: manifest, collection compatibility, delivery URL, and Credential definitions. It lists every failed rule it can evaluate, with release sync’s rejection codes, and exits with code 1 when there are issues. The API is POST /api/plugins/check.

    • Nothing is recorded: no version, no change to the current version, no sync report. The check does not delay Check and Update or the daily check.
    • Without --version, the candidate is the next version in the current line: the next patch below 1.0.0, the next minor from 1.0.0. Without --version, a plugin with no current version gets no collection compatibility check.
    • Only deployment publishers and members who manage a publication of the plugin can run it. Checks are rate-limited.

    See Update or roll back.

  • 73e1410: Clear errors when a plugin’s parentId does not match the collection’s entityBindingSection titled “Clear errors when a plugin’s parentId does not match the collection’s entityBinding”

    Previously, a plugin’s collection put whose parentId did not match the collection’s entityBinding could fail with 403 and a message about the member’s permissions. It now fails with status 400 and a message naming the collection and the plugin’s current version: OWOX reads entityBinding from that version’s plugin.json, not from the author’s working copy.

    • parentId sent to a collection without entityBinding: API code PLUGIN_COLLECTION_NOT_ENTITY_BOUND.
    • parentId missing on a collection with entityBinding: API code PLUGIN_COLLECTION_PARENT_ID_REQUIRED.

    A plugin receives both as HTTP_ERROR with status 400 and the message. See Bind collection documents to OWOX entities.

  • b81425c: Enable the Save button when the Create destination form opens pre-filledSection titled “Enable the Save button when the Create destination form opens pre-filled”

    Previously the Save button on the Create destination form stayed disabled until you edited at least one field, even when the form opened with valid pre-filled values. Now a valid pre-filled create form saves right away. In edit mode, Save still stays disabled until you change something.

  • 7622312: Plugins can run actions on their own backendSection titled “Plugins can run actions on their own backend”

    A plugin can declare actions in a backend block of plugin.json and run them on its own backend, such as a Cloudflare Worker or a Cloud Run service. The page calls ctx.actions.invoke(name, input) and waits for the result, or ctx.actions.start to let the action finish after the plugin closes. OWOX Data Marts validates the input against the action’s JSON Schema, calls the backend, and gives it an execution token with the installing member’s access that works only while the action runs. Credential secrets stay in OWOX Data Marts.

    The video starts a check in an example plugin, leaves the plugin while the check runs on the plugin’s backend, and comes back to the saved result:

    • @owox/plugin-sdk/server provides createBackend. It checks the execution token with the calling host before a handler runs.
    • createBackend runs only executions of its own plugin, named by its GitHub repository id. Without hosts it calls back only public HTTPS hosts, and on Node @owox/plugin-sdk/server/node also refuses names that resolve to private addresses.
    • The release check rejects an input schema with pattern, patternProperties, format, uniqueItems, $async or a recursive or non-local $ref (ACTION_INPUT_UNSUPPORTED).
    • Within a compatibility line, a release cannot remove an action (ACTIONS_INCOMPATIBLE), add a backend or move its origin (BACKEND_ORIGIN_CHANGED).
    • Input and results nested deeper than 100 levels are refused (INVALID_INPUT, BACKEND_INVALID_RESPONSE). OWOX clears the input, result and a plugin-authored error message 24 hours after an execution finishes.
    • At most 5 actions run at once per installation, each for up to 300 seconds; an installation also starts at most 60 per minute and 5,000 per 24 hours, and new starts are refused once 100 MiB of uncleared inputs and results are stored.
    • The install dialog names the origin the plugin runs code on.

    See Run code on your backend with actions.

  • 584a3de: Plugins can call Google APIs with a service accountSection titled “Plugins can call Google APIs with a service account”

    A new built-in Google Service Account Credential stores a service account JSON key. A plugin that declares google-service-account with OAuth scopes in plugin.json calls Google APIs, such as Google Docs and Google Sheets, through the Credential: OWOX Data Marts mints the access token and adds it to each request, so neither the plugin page nor its backend sees the key or the token.

    The video saves a service account key, installs a plugin that asks for Google Docs and Google Sheets scopes, and has the plugin add a ROAS check to the end of a Google Doc and as a row of a Google Sheet:

    • The Credential form takes the key as pasted JSON or a dropped file and shows the service account address to share documents with.
    • A key Google rejects is refused on save, and a replacement key must belong to the same service account.
    • The install dialog lists the scopes a plugin asks for.
    • When Google refuses a stored key during a request, the plugin gets CREDENTIAL_REJECTED and the Credential is marked as rejected, and the Credentials page’s Validation filter lists it.
    • Once a project saves a Google Service Account Credential, an earlier version cannot list that project’s Credentials, so roll back only after deleting such Credentials.

    See Use project Credentials.

  • @owox/internal-helpers@0.39.0
  • @owox/idp-protocol@0.39.0
  • @owox/idp-better-auth@0.39.0
  • @owox/idp-owox-better-auth@0.39.0
  • @owox/backend@0.39.0
  • @owox/web@0.39.0
  • ce4310f: Import from Google BigQuery blocks tables stored in another locationSection titled “Import from Google BigQuery blocks tables stored in another location”

    Previously, Import data marts from storage let you pick a table or view from a dataset stored in a different location than the Google BigQuery storage, for example a US dataset with an EU storage. The data mart was created, but it could not read its data. The picker now shows each dataset’s location and greys out the tables and views outside the storage’s location, so they can’t be picked.

    The Select… picker for Table, View and Pattern Data Marts follows the same rule.

  • 5e885c5: Publishing a plugin you already have now says it is installedSection titled “Publishing a plugin you already have now says it is installed”

    Publishing a repository whose plugin you have already installed used to confirm only Plugin published, leaving it unclear whether anything had been installed. The confirmation now names the plugin and says you already have it installed. Publishing still changes only who can find a plugin; it installs it for nobody and leaves existing installations as they are.

    Publish a plugin panel with the repository of an already installed plugin entered

    Confirmation after publishing: Plugin published. You already have Import Model installed.

    See Publish in OWOX Data Marts.

  • 4382602: Edit a relationship from its arrow on the Models canvasSection titled “Edit a relationship from its arrow on the Models canvas”

    Click an arrow on the Models canvas, or its join fields label, to open that relationship in a panel on the right. Previously a click only highlighted the arrow, and changing a join meant opening the source Data Mart. The panel edits the same Report Fields, Join Settings and Description settings as the relationship’s row in Joinable Data Marts, in collapsible sections like the app’s other side panels, and changes save as you type.

    The video shows a relationship opened from its arrow, edited in place and switched to another one:

    • The arrow stays highlighted while the panel is open. Click another arrow to switch the panel to it.
    • A two-headed arrow stands for two mirrored relationships; pick the direction to edit at the top.
    • Allow for reporting is under General, and Delete Relationship under Danger zone. Open in Data Setup at the top opens the relationship’s row in its source Data Mart.
    • A row in a card’s N relationships list opens the same panel. From the keyboard, Enter moves focus into the panel and Esc returns it to the row.
    • After a join fields change, the canvas redraws the arrow and keeps your zoom and position.

    See Edit a relationship.

  • 582661c: Preview SQL from an open reportSection titled “Preview SQL from an open report”

    The report side panel now has a Preview SQL button in its header, next to Copy link. When a teammate shares a link to a report, you can read the query the report runs without going back to the reports list. The button is available for Google Sheets, Microsoft Excel, Data Studio, Email, Slack, Microsoft Teams and Google Chat reports. As in the reports list, it is not shown for Data Marts with a Pattern definition on storages other than Google BigQuery.

    Watch a report opened from a shared link show its SQL, and the note that appears when Report Columns has unsaved edits.

    The SQL comes from the saved report. If Report Columns has unsaved edits, the Report SQL dialog says they are not included. In Data Studio reports, the button moved from under Report Columns to the header. See Joinable Data Marts.

  • 5c38b50: Storage and Destination settings point to the field you need to fixSection titled “Storage and Destination settings point to the field you need to fix”

    Previously, when a Storage or a Destination could not be saved because of one of its values, the only explanation was a toast at the top of the page, such as Invalid config — projectId: Invalid GCP project ID…, and nothing in the form showed which field it meant. A Destination’s panel even closed, discarding what you had entered. Now the panel stays open and Save marks that field in red, opens its section if it was collapsed, moves the cursor to it, and shows the reason under the field. The mark clears when you correct the value.

    The video shows a Storage and a Destination that each point to the field to fix:

    • Google BigQuery: the Project ID format is checked before saving, so a project name entered instead of the ID is flagged at once. Spaces around a pasted ID are removed.
    • Service Account in a Google BigQuery Storage or a Google Sheets Destination: a key that is not valid JSON or has no client_email is flagged on the field. Before, Save in a Storage did nothing in this case.
    • Any storage type: when the server rejects a value in the connection settings or credentials, the form marks that field the same way. Connection errors reported by the warehouse itself, such as a wrong password, still appear as a message only.
    • Google Sheets Destination: a Drive folder the service account can’t use is marked on the folder field, with the reason and the fix. Email, Slack, Microsoft Teams and Google Chat Destinations mark a recipient or webhook URL the server refuses.
    • A save that fails for another reason, such as a lost connection, now says so instead of doing nothing. A change to the owners is kept when you fix a field and save again. A save refused after your session had to be renewed no longer signs you out.

    Also, a calculated field’s formula editor no longer loses characters typed quickly while the page is busy, which also closed the field suggestions mid-word.

    See Google BigQuery, Storage Management and Destination Management.

  • 76696b7: Plugin page drops Reinstall and renames Check now to Check and UpdateSection titled “Plugin page drops Reinstall and renames Check now to Check and Update”

    An installed plugin’s page no longer offers Reinstall. OWOX does not host or package plugins, so installing an installed plugin again rebuilt, replaced or reset nothing — it only looked like a lifecycle action. Install for a plugin you have not installed, Restore in Installation history after an uninstall (with its confirmation), and Configure Credentials for a plugin that asks for Credentials work as before.

    The button next to the plugin’s version is now called Check and Update instead of Check now. It does the same thing: asks for a newer release right away instead of waiting for the daily check — a check that has just run answers for it — and a newer valid release it finds becomes current for everyone using the plugin. Updates and versioning are otherwise unchanged.

    See Update or roll back.

  • 9f81f2d: Google Sheets errors in Run History explain the cause and the fixSection titled “Google Sheets errors in Run History explain the cause and the fix”

    Previously, when Google Sheets failed to apply a report’s changes, Run History showed only Google’s text — Internal error encountered., The service is currently unavailable. or Requested entity was not found. — and the same report could fail with a different message every day. These failures now show one message in plain words: Google Sheets couldn’t finish updating the spreadsheet. It says what to do: run the report again after a single failure; if it fails on every run, limit the formulas that read whole columns of the report’s sheet or send the report to a separate spreadsheet, and fix circular-reference errors — Iterative calculation is only for circular references made on purpose. Google’s text stays at the end as details. See When Google Sheets can’t finish updating the spreadsheet.

    Other errors from Google Sheets now name the step that failed and Google’s reason.

  • 16d8a24: Shopify connector: more orders fields, including line-item discountsSection titled “Shopify connector: more orders fields, including line-item discounts”

    The orders lineItems JSON now includes discountedUnitPriceAfterAllDiscountsSet (unit price after all discounts), totalDiscountSet (the line’s discount total, excluding order-level discounts), currentQuantity (units net of refunds and removals), and discountAllocations with the allocated amount and the discount’s type, code or title, and index. The index joins each allocation to its entry in discountApplications, which now also exports index and __typename. See the Discount Fields section in the connector guide for how the amounts differ.

    Data marts that already select lineItems or discountApplications get the new JSON keys on their next run; previously imported rows keep the old shape until a backfill.

    The orders field list also gains 47 previously unavailable scalar and money fields (order number, confirmation number, test/edited flags, additional payment and duties totals, fulfillment and tax flags, and more). These are opt-in via the existing Fields picker and aren’t added to any data mart automatically.

    See the Shopify connector guide for setup.

  • e42c3cf: Explore business contexts with your AI assistantSection titled “Explore business contexts with your AI assistant”

    Connected AI assistants can read complete context descriptions and discover Data Marts within selected contexts. Search, catalog, and Data Mart details now include attached context names and IDs, while preserving each user’s existing access. Empty context results no longer suggest creating a first Data Mart when accessible published Data Marts exist elsewhere.

    In Project, Context, and Data Mart descriptions, use Markdown and type @ to find a published Data Mart and insert a link. Preview the result before saving. If saving a Data Mart description fails, your draft stays in the editor for retry. Links explain which data supports a goal; they do not change context assignments or permissions.

    See MCP and Contexts.

  • 5717428: SQL validation no longer fails for large queriesSection titled “SQL validation no longer fails for large queries”

    On MySQL-backed deployments, SQL validation (dry run) failed with a server error for queries over 64 KB. Large queries now validate normally.

  • b2f580c: Athena query results are cleaned up when the Output Bucket names a folderSection titled “Athena query results are cleaned up when the Output Bucket names a folder”

    When an AWS Athena storage’s Output Bucket included a folder, such as my-athena-results/owox/, OWOX Data Marts could not delete the results of the queries that read your data, and they stayed in your bucket. They are now deleted once read, as with a plain bucket. The bucket link on the Data Mart page opens that folder.

  • 60644c8: Inspect executed SQL for HTTP Data streaming runsSection titled “Inspect executed SQL for HTTP Data streaming runs”

    Run History now shows the SQL used by the Data Mart and report HTTP Data streaming endpoints with output controls or joined fields in the Executed SQL block, with parameter values inlined and a copy button. Previously, these streaming API runs had no dedicated SQL block in Run History.

    See the HTTP Data API for how to locate the run using the x-owox-run-id response header.

  • cdd9424: Enable the Create trigger button on a freshly opened formSection titled “Enable the Create trigger button on a freshly opened form”

    The trigger creation form opens with valid defaults, but the Create trigger button stayed disabled until the user changed any field. The button is now enabled immediately in create mode; submitting an incomplete form shows a validation error instead (for example, a missing report). Editing an existing trigger still requires a change before Save changes enables.

  • @owox/internal-helpers@0.38.0
  • @owox/idp-protocol@0.38.0
  • @owox/idp-better-auth@0.38.0
  • @owox/idp-owox-better-auth@0.38.0
  • @owox/backend@0.38.0
  • @owox/web@0.38.0
  • 4444a95: Google and Microsoft sign-in works after a page is left openSection titled “Google and Microsoft sign-in works after a page is left open”

    You can choose Google or Microsoft on a sign-in or sign-up page left open for a while without an expired session interrupting the sign-in. If a sign-in session does expire, the page now shows “Your sign-in session expired. Please try again.” so you know to retry.

  • e339903: Clearer connector errors, and custom connectors that handle more APIsSection titled “Clearer connector errors, and custom connectors that handle more APIs”

    • When a connector’s request to its API fails, Run History shows the provider’s own message instead of the raw response body, and a skipped account names what was being imported.
    • A run that loaded nothing shows Loaded 0 rows in Run History, and for a custom connector the log names each node that got no records and what to check.
    • The Connector Builder checks node and parameter names as you type, explains the rule and suggests a name that fits, such as v1_balance_history for v1/balance/history.
    • Test runs the right node after you rename it in Code, instead of failing with Unknown node.
    • When a rate-limited API says in a Retry-After header when to try again, a custom connector waits that long, up to 5 minutes, instead of using up its retries inside the limit and failing the run.
    • Errors from APIs whose error bodies follow JSON:API, Klaviyo among them, show the API’s own explanation instead of the raw response body.
    • Publish and Test refuse a node that puts {{ dateWindow.start }} or {{ dateWindow.end }} into its request without fetching by date window, and name the node and the fix, where every run of it used to fail with Template path "dateWindow.start" is unresolved.
    • A run that needs a field the connector no longer provides names that field and points to Edit Fields, which now drops such a field when you save.
    • A node can end its date window a number of days before today, with Skip the last days under Incremental or incremental.endLagDays in the manifest, for APIs such as OpenAI Ads that refuse a window ending today.
  • cb5a9c1: Unchecking a column in the OWOX Extension no longer fails Save & RunSection titled “Unchecking a column in the OWOX Extension no longer fails Save & Run”

    Previously, unchecking a column in the Google Sheets Extension or the Excel add-in left its aggregation, date bucket and sort behind, and Save & Run failed with Output controls validation failed. Sort column not selected: … Aggregation column not selected: …. The report saved only after those rules were removed one by one in Output controls.

    Now the Extension removes the rules that hang on an unchecked column:

    • The aggregation and date bucket on the column go with it, together with a metric filter bound to that aggregation.
    • The sort on the column goes too, whether or not the report still aggregates. A report sorted by a column it does not print skips the automatic collapse and delivers raw rows, so the Extension no longer offers such a sort either. One saved before stays and is not marked as broken, so such a report stays uncollapsed until you remove it.
    • Reset above the column list returns the columns together with their aggregations, date buckets, metric filters and sorts to the last saved state. A saved sort the report can no longer sort by, for example after you add a Unique Count, stays removed.

    This covers the row checkboxes, the checkbox that selects or clears all columns, and unchecking a disconnected column. See Report Output Controls.

    Also fixed in the Extension:

    • Editing a metric filter, a filter on an aggregated value such as the Sum of Revenue, keeps it a metric filter. Before, the edit turned it into a filter on the rows. The filter lists now name the aggregate, for example Sum greater than: 100.
    • Unchecking every column shows Select at least one column and does not save. Before, an empty selection delivered every column, or failed on the server when the report had slices, a sort or a filter on an aggregating calculated field. A Unique Count on its own still saves, as a metrics-only report.
    • Saving from the Extension keeps the column list a report was saved with. A report made in the web app or through MCP with every native column selected used to lose its automatic duplicate collapse the first time it was saved from the Extension.
    • Adding a slice to a report that lists no columns now shows the automatic aggregation that the report gets on delivery.
  • 0afa9d1: Technical Owner and Technical User are now called Data OwnerSection titled “Technical Owner and Technical User are now called Data Owner”

    The owner field on a Data Mart, formerly Technical Owner, is now Data Owner. The project role formerly called Technical User is now also Data Owner. Both are named after the person who is responsible for the data. Access and permissions do not change. Business Owner, Business User and Project Admin keep their names.

    The video shows the new name in the app:

    Ownership section of a Data Mart Overview tab with the Data Owner and Business Owner fields

    Configure member panel with the Data Owner role selected for a project member

    The new name is used on the Data Mart page, in the Data Marts list column and filter, in the member and access-request role pickers, and in sharing hints, error messages and the MCP guidance. Invitation emails in self-hosted deployments use it too. The API keeps its field names, such as technicalOwnerIds and technicalOwnerUsers, and the role value stays editor, so existing integrations keep working. See Roles and Permissions and Ownership and Sharing.

  • 8a7ddfd: Preview data no longer fails on out-of-range datesSection titled “Preview data no longer fails on out-of-range dates”

    On a Databricks or Snowflake Data Mart, Preview data failed with a server error when a DATE or TIMESTAMP column held a value outside the range the app can display (before year −271821 or after 275760), and one such cell broke the whole preview. The preview now loads and shows that cell as [Invalid Date], with the other rows and columns as usual.

  • dbfb24c: Filters with “contains” and “does not contain” can be appliedSection titled “Filters with “contains” and “does not contain” can be applied”

    In Filters on the Data Marts, Storages, Destinations, Reports, Insights and Triggers lists, switching a row’s condition from is to contains or does not contain left Apply filters disabled whatever value you typed. The button now turns on as soon as the row has a value, and the filter applies.

    The conditions that take several values are renamed from is and is not to is any of and is none of, as in report filters. They work as before.

  • f344499: Pick any Lucide icon for a Data Mart, with searchSection titled “Pick any Lucide icon for a Data Mart, with search”

    The Data Mart icon picker now offers the whole Lucide icon library, more than 1,500 icons, not only the recommended set. The recommended icons stay at the top. All icons lists the rest below them. See Data Mart icons.

    The video shows the picker, a search and the picked icon on the Models canvas:

    • A search box filters both lists by English name, for example cart or arrow up. A recommended icon also matches the name of its picture, so cart finds Purchases. Enter picks the first match.
    • The picked icon shows on the Data Mart page, on the Models canvas card and in the PNG and SVG exports, like the recommended ones.
    • In the API, icon on POST /api/data-marts and PUT /api/data-marts/{id}/icon also accepts lucide:<icon-name>, for example lucide:rocket. Detail, list and Models canvas responses return the same value. The recommended keys such as purchases are unchanged, so icons picked before keep working.
  • 0cf0911: Clearer MCP labels for external systemsSection titled “Clearer MCP labels for external systems”

    OWOX Data Marts MCP now identifies when queries, report schema checks, and report schedules can interact with a connected data warehouse or report destination outside OWOX.

  • a003a9c: Clearer MCP report actions and confirmationsSection titled “Clearer MCP report actions and confirmations”

    OWOX Data Marts MCP now describes its report export options without directing assistants away from other integrations. Report creation, delivery, and updates declare their possible irreversible effects, and report schema lookup indicates when it may refresh a warehouse view.

  • @owox/internal-helpers@0.37.0
  • @owox/idp-protocol@0.37.0
  • @owox/idp-better-auth@0.37.0
  • @owox/idp-owox-better-auth@0.37.0
  • @owox/backend@0.37.0
  • @owox/web@0.37.0
  • 42c797b: Preview Data Mart rows from Data SetupSection titled “Preview Data Mart rows from Data Setup”

    The Data Setup tab now has a Preview data button under the Output Schema. It runs a query in your data warehouse and shows the first 10 rows of every visible field, so you can check the Input Source and schema before building reports. Draft Data Marts can be previewed too.

    • Change Limit (1–1000) and click Update to fetch more rows; page through them without another query.
    • Click a column name to sort by it (ascending, descending, off); the sort runs in the warehouse, so you see the real top rows.
    • Filter a column from its header: the condition runs in the warehouse as a WHERE clause, and active filters show as chips above the table.
    • Each preview, including Re-run and each limit, sort or filter change, is a new warehouse query. It is not a Data Mart run: it does not appear in Run History and does not consume credits.

    See Data Preview.

  • 9455d36: Upload a Google Sheet to your storage from the OWOX ExtensionSection titled “Upload a Google Sheet to your storage from the OWOX Extension”

    A Data Mart on the Google Sheets connector imports one tab of a spreadsheet into your storage. Previously, refreshing that table after editing the sheet meant opening OWOX Data Marts and starting a manual run, or waiting for the next scheduled one. Now the person working in the sheet can do it right there: open the OWOX Extension for Google Sheets on that tab and click Upload.

    • The extension finds the published Data Mart that imports the open tab and shows its Upload card right away, with the date and status of the last import. No report is created, and no Data Mart has to be picked.
    • Upload starts an ordinary manual run: the storage table is fully replaced with the current sheet data, the run appears in the Data Mart’s Run History and counts toward consumption. The button stays locked until the run finishes. If a run is already in progress, the extension says so instead of starting another one.
    • Upload needs the project Editor role and edit access to the Data Mart, the same as a manual run in the web app.

    On other tabs and in other spreadsheets, the extension works as before, and the Upload card also offers Create a report on this sheet instead. The Excel add-in is not affected.

    See Upload from the OWOX Extension.

  • e1b5816: Hiding a calculated metric no longer fails the save on a Data Mart with a joined formulaSection titled “Hiding a calculated metric no longer fails the save on a Data Mart with a joined formula”

    Previously, on a Data Mart with a calculated field that reads a joined Data Mart — roas over SUM(orders.amount), for example — Hide from reports on any calculated metric, that one included, made the Output Schema save fail with “Cannot build report SQL. Disconnected columns” naming the hidden metric. A metric that was already hidden could block adding such a formula in the same way, with “Hidden columns” naming a field nobody had touched. The save now succeeds: every formula is still checked against your warehouse, and hidden metrics leave the report column picker as usual.

    Data Marts whose calculated fields read only their own columns were not affected. See Hidden columns warning.

  • c5d7108: Warn when a join changes how a calculated field’s number readsSection titled “Warn when a join changes how a calculated field’s number reads”

    A calculated field that reads a joined Data Mart now tells you — while you write the formula and when you save it — how the join shapes its number. Nothing is blocked: every field still saves as written, and an AI assistant reading the field through the MCP server gets the same advice as a caveat to pass on, without naming a Data Mart or field it cannot report on. The walkthrough below opens four such fields and shows that each still saves.

    • A joined COUNT without DISTINCT counts the rows of this Data Mart that found a match, not the joined Data Mart’s own rows. It comes out higher when one joined row matches several rows here (FORMULA_JOINED_MEASURE_MULTIPLIED), and lower when several joined rows share one key — a customer with three orders counts once (FORMULA_JOINED_MEASURE_COLLAPSED). The warning names the join at fault and points at COUNT(DISTINCT …) over a column of the joined Data Mart, or at its Unique Count as a report column when it offers one. Where no primary key is declared to check against, it says the count cannot be checked either way (FORMULA_JOINED_MEASURE_GRAIN_UNPROVEN) — and a save that sets that key is judged against it.
    • Any joined measure leaves out the joined Data Mart’s rows that match nothing here, so the result may not reconcile with that Data Mart’s own totals (FORMULA_JOINED_ROWS_EXCLUDED). You see this when you write or change the formula, not again after every unrelated save.

    The formula editor of the orders_counted field, COUNT(orders.amount), with two amber warnings: the join on user_id can match several rows so COUNT counts matches — use COUNT(DISTINCT ...) or pick its Unique Count measure in a report — and the joined Orders rows that match nothing are dropped

    SUM, AVG, MIN, MAX and COUNT(DISTINCT …) over a joined Data Mart are already computed set-based and are not flagged. See Calculated Fields Across a Join.

  • 64e2741: Pick the right Dedup for a joined field, with the guide one click awaySection titled “Pick the right Dedup for a joined field, with the guide one click away”

    The joined Data Marts guide now explains what a joined field’s Dedup does: the value a report column shows, the type it gives the field and with it the aggregations a report can apply, and what a report’s Sum, Average, Count Unique and Count read after the join. A table of common cases shows which Dedup to pick — a lookup attribute, revenue per customer, individual orders, hits per session, first and last dates, text that varies per key — and where each one stops being exact.

    The tooltip on the Dedup column header in a joined Data Mart’s Report Fields tab now links straight to that section.

    The Dedup column header tooltip in the Report Fields tab of a joined Orders Data Mart, with a Learn more link, above order_id set to COUNT_DISTINCT, session_id to ANY_VALUE and amount to SUM

    See Choosing a Dedup.

  • 5062ee8: MCP clients can recover from expired OAuth grantsSection titled “MCP clients can recover from expired OAuth grants”

    MCP clients that submit an expired or invalid refresh token or authorization code now receive the standard invalid_grant OAuth error and can restart authorization automatically, instead of failing with an unexpected server error. C2C authentication failures and Identity Provider outages remain server errors and are not reported as invalid user grants.

    See Connect AI assistants with MCP for setup and reconnection guidance.

  • 84cab72: Support MCP protocol 2026-07-28 and alert on compatibility failuresSection titled “Support MCP protocol 2026-07-28 and alert on compatibility failures”

    The MCP endpoint now understands the 2026-07-28 protocol revision, including the server/discover method — Claude clients that probe for it before connecting no longer get rejected. Existing clients on 2024–2025 protocol revisions remain supported through the compatibility fallback. Rejections of the modern revision ODM promises to serve are now logged at error level so production alerting catches them, while malformed claims for a legacy revision remain warnings.

    During owox serve shutdown, ordinary MCP calls drain before long-lived subscription streams are closed, so an open subscriptions/listen response cannot stall a rolling update. See MCP Server for connection and troubleshooting guidance.

  • a3e027e: Connect Codex to project-specific MCP URLsSection titled “Connect Codex to project-specific MCP URLs”

    OAuth clients that verify issuer-bound authorization responses (RFC 9207), including Codex, can now complete sign-in against project-specific MCP URLs (https://<project>.mcp.owox.com). Both successful and error callbacks from a project-specific URL carry its iss value.

    After a client’s client_id and redirect_uri are trusted, authorization failures are now sent back to every MCP client through that redirect URI with the standard error, error_description, and state parameters. Unexpected internal failures use server_error; the underlying cause stays in the server logs.

    See Connect AI assistants with MCP for setup instructions.

  • 4936268: Output Schema rows are separated again in the dark themeSection titled “Output Schema rows are separated again in the dark theme”

    In the dark theme the Output Schema table of a Data Mart showed its fields as one solid block: the line between rows was drawn in the same shade as the row background, so it could not be seen. Rows are now separated by a visible line in the dark theme, as they always were in the light theme, for every storage type. The divider between the Add Field and Add Calculated Field buttons under the table is visible again for the same reason.

    The Output Schema of a Data Mart in the dark theme, each field row separated from the next by a thin line, with a visible divider between the Add Field and Add Calculated Field buttons below the table

    The light theme is unchanged, and no field, alias, or description is affected.

  • aae9afa: Long field descriptions wrap in the Output SchemaSection titled “Long field descriptions wrap in the Output Schema”

    A field’s description in a Data Mart’s Output Schema now wraps inside its column instead of running on as one line. Previously a long description, including the ones the AI helper generates for a nested record, stretched the Description column to the length of its longest line, and the whole table had to be scrolled sideways to read it. Line breaks written into a description are kept, so a STRING: … / BOOLEAN: … breakdown still reads as separate lines.

    The Output Schema table with a RECORD field whose long description wraps inside the Description column, its STRING and BOOLEAN sub-field lists kept on separate lines

    • A description longer than eight lines ends in an ellipsis; click it to read and edit the full text.
    • The description editor opens wide enough for a few sentences and still fits a narrow screen.
    • A schema with no descriptions yet keeps the width it had before.

    Applies to every storage the Output Schema is edited for, nested BigQuery RECORD fields included. See Table-based Data Mart and SQL-based Data Mart for where field descriptions are written.

  • 77dd97d: Find unconnected Data Marts and read field aliases and descriptions on the Models canvasSection titled “Find unconnected Data Marts and read field aliases and descriptions on the Models canvas”

    The relationships filter on Data Marts → Models has a third option, Without relationships only: it leaves only the Data Marts that no other visible Data Mart joins to, so the ones still waiting to be connected — or to be cleaned up — stand out instead of hiding among the joined cards. It respects the status filter, and it is part of the page URL (rel=unconnected) like the other filters.

    The Models canvas relationships filter open with All Data Marts, With relationships only and Without relationships only, the last one selected and only the two unconnected Data Marts left on the canvas

    In the ERD view, each field row now shows the Output Schema description under the field, and two new entries in the Field rows section of the canvas settings control the rows: Field descriptions switches the description line off, and Field aliases switches the row text between the Output Schema alias (as before) and the technical field name. Whichever of the two is not shown is available on hover. Both labels are on by default. Long descriptions are cut to one line — hover to read the whole text. The Joinable Data Marts diagram gets the same labels.

    The Models canvas in the ERD view with the canvas settings open: Field aliases and Field descriptions are ticked under Field rows, and the Orders card lists each field by its alias with its description underneath

    See Models Canvas.

  • d9d236d: Editing a relationship description no longer collapses the joined Data MartSection titled “Editing a relationship description no longer collapses the joined Data Mart”

    On Data Setup → Joinable Data Marts, the Description tab of a joined Data Mart autosaves while you type. Previously every autosave reloaded the whole list of joined Data Marts: the expanded card collapsed, the Description tab closed and the cursor left the field, so a longer sentence could only be entered a few words at a time. Now the saved text is applied to the card in place — the card stays expanded, the tab stays open and the cursor stays where it was, while the description still reaches the AI assistant (MCP) and the report column picker as before. The “Relationship updated” notification no longer pops up after every pause: the description saves silently, like the other fields on the card, and only a failed save is reported.

    Saving Join Settings is unchanged and still refreshes the list. See Describe the Relationship.

  • d77c01b: Custom no-code connectorsSection titled “Custom no-code connectors”

    Build a connector to any REST API without writing code. A manifest describes the API — authentication, pagination, nodes and fields — and a three-pane web builder edits it with live testing against the real endpoint. Connectors are versioned: publish, roll back, and bind them to Data Marts alongside the built-in ones. Title, description and documentation link stay editable; the name is fixed once the connector exists, because a Data Mart references its connector by name — deleting a connector frees its name to be used again. A connector’s manifest is readable by editors only — it is author-written JSON that can hold a credential typed straight into the builder — while the connector list, its configuration form and its field schema stay open to viewers. Publishing a version, or making another version active, changes what runs in every Data Mart that follows the connector’s active version, so it needs edit access to each of those Data Marts; otherwise a project admin can do it.

    See Custom Connectors and Connector Builder.

  • 89fbd10: Automatic aggregations you can changeSection titled “Automatic aggregations you can change”

    When a report with a selected column list has no aggregation, OWOX picks one for each metric so the report does not return duplicate rows. Removing those aggregations did not stick: after Save the panel was empty, but the next run or reopen of the report applied them again, so the report could not return its rows as stored. Now OWOX remembers every column you remove an aggregation from:

    • Remove some of the automatic aggregations and the report runs with the ones you kept.
    • Remove all of them and the report returns every row as stored, with no aggregation and no DISTINCT, on every run and every reopen.
    • Add an aggregation back to a column and the report runs with the aggregation you chose. Take the column out of the report and OWOX forgets the removal.

    This covers every destination that collapses on delivery, including the Microsoft Excel add-in’s own fetch. See Automatic Aggregation.

  • 0a963d8: Start a connector run without expanding the Input Source cardSection titled “Start a connector run without expanding the Input Source card”

    The Manual Run button of a connector Data Mart now sits in the header of the Input Source card on the Data Setup tab, next to the collapse arrow. Previously it was inside the card body, so a collapsed card hid it and you had to expand the card first, or find the run in the Data Mart’s ⋮ menu. Now you can open a Data Mart from a failed-run email and start a new run straight away.

    The video collapses the Input Source card and starts a run from its header:

    • The button appears once the connector is configured.
    • It stays disabled, with the reason in a tooltip, while the Data Mart is a draft or a run is in progress, as before.

    See Connector-based Data Mart for running a connector manually.

  • 9455d36: Clearer Data Mart cards and canvas settings on the Models canvasSection titled “Clearer Data Mart cards and canvas settings on the Models canvas”

    Cards on Data Marts → Models now match the Data Mart cards on the OWOX website. Each card leads with an icon and the title, then short badges for the input source, the field count, and the number of triggers, reports and relationships. The badges share a line while they fit, and a count of zero shows no badge. The footer keeps the Data Quality and Data Last Updated indicators and shows when the Data Mart is shared for reporting or for maintenance. Hover a sharing icon to see what it allows.

    • Click N relationships to list the Data Marts this one joins or is joined by, with the join fields. The count and the list cover the whole storage, whatever the canvas filters hide.
    • Click N fields in Compact mode to list the Output Schema fields under the card.
    • Only unpublished Data Marts carry a status badge, Draft, next to the title. Published ones no longer show a Published label.
    • The colored Data Quality bar on the left edge of the card is gone. The Data Quality shield in the footer shows the status, and hovering it explains the result.

    The canvas settings (the gear button) have a new design, and every checkbox hides exactly what it names:

    • View is picked from two cards with a small preview of each. Horizontal / Vertical and Show join fields sit on one row below them.
    • Card content has a checkbox for each part of a card: Input source, Fields, Triggers, Reports, Relationships, Draft badge, and Quality and sharing. Title only unticks every checkbox, Field rows included, and leaves only the titles. Show all ticks everything back on.
    • Hover an option to see an info icon with a tooltip that explains it.
    • Field rows holds Field aliases and Field descriptions, next to Card content. It works in the ERD view, where cards list their fields, and stays greyed out in Compact mode.
    • Ticking a checkbox or the Show join fields switch keeps your zoom and position. Picking another view or layout algorithm still fits the whole graph.
    • A title-only preference saved earlier still shows only the titles.
    • The settings scroll when the window is too short to fit them.

    See Models Canvas.

  • 9455d36: Pick an icon for each Data MartSection titled “Pick an icon for each Data Mart”

    Click the icon next to a Data Mart’s title to give it one that matches its subject. Everyday icons cover purchases, orders, sessions, customers, countries, ad spend and traffic sources. Data-stack icons cover data sources, pipelines, SQL, reports, dashboards, joins, metrics, UTM tags, attribution and more. The icon shows on the Data Mart page, on its card on Data Marts → Models and in the canvas PNG and SVG exports. Reset to default brings back the plain box that Data Marts show until an icon is picked.

    • Changing the icon needs edit access to the Data Mart.
    • The API returns the icon as icon on a Data Mart, in the Data Mart list and in GET /api/model-canvas/data-marts. Set it with PUT /api/data-marts/{id}/icon or when creating a Data Mart; an unknown icon key is rejected with 400, and null resets it.

    See Models Canvas.

  • @owox/internal-helpers@0.36.0
  • @owox/idp-protocol@0.36.0
  • @owox/idp-better-auth@0.36.0
  • @owox/idp-owox-better-auth@0.36.0
  • @owox/backend@0.36.0
  • @owox/web@0.36.0
  • 6a5d52e: Deleted reports are retainedSection titled “Deleted reports are retained”

    This update includes a database migration. Older backend versions do not hide deleted reports; reverting the migration makes those reports visible again.

    Report deletion now uses soft delete: reports are removed from lists while their settings and run history remain stored.

  • 0d63382: Handle array fields safely in reportsSection titled “Handle array fields safely in reports”

    Reports can display repeated fields and fields whose existing schema type is ARRAY as columns. Filters, slices, sorting, aggregations, and date buckets are unavailable for these fields, including count, any-value, and blank checks. Reports with stored array controls must have those controls removed or be recreated before saving or running.

    Joined arrays are returned as JSON arrays that preserve each source row’s array; each ancestor join adds another array level. Existing deduplication overrides on joined array fields are ignored in favor of this JSON-array rollup. Ordinary nested scalar fields and opaque JSON, VARIANT, and SUPER fields keep their existing report controls.

    Whole records selected as columns are exported to Google Sheets as JSON text. Selecting their nested scalar fields continues to work as before.

  • 8c5f21b: Destinations tab now promotes connecting an AI assistant alongside your existing destinationsSection titled “Destinations tab now promotes connecting an AI assistant alongside your existing destinations”

    A published Data Mart’s Destinations tab now shows a card inviting you to connect Claude or ChatGPT even when you already have other destinations configured — previously this offer only appeared before any destination existed.

    The card sits in the destinations list like any other destination, offers the same Claude/ChatGPT choices used elsewhere in the product, and a link to the MCP setup guide. It can be collapsed, like any other destination card, if you’re not interested.

  • f44a455: Resume an interrupted manual backfill from the last fully loaded dateSection titled “Resume an interrupted manual backfill from the last fully loaded date”

    When a deploy or a restart interrupts a manual backfill, the automatic retry now continues from the day after the last one it fully loaded, instead of starting the whole period again. It keeps the days it already imported and requests only the rest, which matters most on a long backfill.

    Starting a backfill yourself always loads the whole period you choose, including after one that failed. Only the automatic retry of an interrupted run resumes.

    Run History names the day that retry starts from, so you can see why it covers a shorter period than the one you chose.

    Scheduled and incremental runs are unaffected. A backfill still never moves the incremental load position, so the next scheduled run continues to pick up where regular loading left off.

    Shopify and TikTok Ads backfills still reload the whole period. Shopify imports one data type at a time across the full range rather than one day at a time across all of them. TikTok Ads reports a failed day only after it has walked the whole range, so it cannot yet tell which days it truly loaded.

  • 3f00542: Select conversions and conversion values by default in Facebook Marketing insightsSection titled “Select conversions and conversion values by default in Facebook Marketing insights”

    • The conversions and conversion_values fields are now selected by default for the Facebook Marketing insights endpoints, including the breakdowns by age and gender, country, device platform, product ID, publisher platform and position, and region, plus the ad set and campaign endpoints. New Data Marts built on these endpoints include conversion data without manual field selection.
    • Ad Account Insights by Link URL Asset keeps its previous defaults. Meta supports only impressions, clicks, spend, reach, actions, and action_values with Dynamic Creative asset breakdowns, so the conversion fields stay available but unselected there.
    • Both fields now carry real descriptions instead of placeholders, so the field picker, the Output Schema, and MCP clients can tell conversions apart from actions.
    • The connector documentation now lists how Meta limits conversion metrics on the link URL asset, product ID, and region breakdowns, and the troubleshooting guide names these fields as the first ones to clear when Meta asks you to reduce the amount of data.
  • 4b18771: Find reports from searchSection titled “Find reports from search”

    Self-hosted upgrades: stop the previous application and its background workers before starting this version. Running both versions together can leave report search results outdated.

    API client users: update the OWOX API client or explicitly filter by supported entityTypes. Unfiltered search results now include REPORT.

    Find reports alongside Data Marts, storages, and destinations. Search by full words or the beginning of a word, with report names ranked highest. Each result shows its Data Mart and destination and opens the report directly. Results respect access to both the Data Mart and destination. Watch the search-to-report flow:

    Existing reports become searchable automatically, and results stay up to date as reports, Data Marts, and destinations change. Report search is also available through the API client and MCP for AI assistants.

  • 34ec11a: A hidden column no longer reads as a broken schemaSection titled “A hidden column no longer reads as a broken schema”

    Hiding a field from reporting takes it off the reporting menu — it does not remove it from the Data Mart. A report that already selected that field said otherwise: it listed the column under Disconnected columns and refused to build with “They are missing from the current Data Mart output schema… contact your analyst to restore the schema.” Nothing was missing, and the analyst being contacted was usually the person who had just hidden it.

    The report editor and the error now tell the two states apart:

    • Hidden columns — still in the Data Mart, taken off the menu. Uncheck them, or ask your analyst to show them in reports again.
    • Disconnected columns — genuinely gone from the schema. Unchanged: uncheck them, or have the schema restored.

    The Report Columns list of a report that selected a since-hidden calculated field. An amber "Hidden columns" group at the top holds the checked ROAS column, with an eye-off icon whose tooltip reads "They are still in the Data Mart, but hidden from reporting. Uncheck them and remove any filter, sort, aggregation or date bucket rule that references them, or ask your analyst to show them in reports again." The remaining columns — date, source, medium, campaign, revenue, adCost, Unique Count — are listed below.

    This covers every switch that hides a column — a Data Mart’s own field, a calculated field, a joined Data Mart’s field, and Hide from reports on a single joined field in the join’s Report Fields tab — and nothing about what hiding does has changed: the column stays out of the picker and the MCP tools, a report that selects one still cannot run until it is unchecked, and a formula that references one keeps computing. A report carrying one column of each kind is told about both, and keeps the advice to restore the schema. See Report Output Controls.

  • ba59eee: Reports no longer deliver duplicate rows when no aggregation was chosenSection titled “Reports no longer deliver duplicate rows when no aggregation was chosen”

    A report with an explicit column selection and no aggregation, date bucket, or Unique Count set anywhere used to deliver every underlying row — most visibly from a join that fans one source row out across many. OWOX now collapses it on delivery.

    Two things to check before upgrading. An automatically aggregated column is relabelled exactly as a manually chosen one always has been — sessions becomes sessions | SUM — so a Google Sheets formula or Looker Studio binding that reads the plain sessions header stops resolving. And a Data Mart defined by SQL now counts as having output controls, which materialises its definition as a view (CREATE OR REPLACE VIEW) on every run and on every Generated SQL preview: a read-only storage credential that worked before will now fail.

    Watch a report that sets no aggregation get one applied, and the query it will actually run.

    What a report gets depends on what it selected:

    • Dimensions only — returned distinct. Nothing is renamed and no value can change.
    • A metric among them — the first aggregation that Data Mart’s governance allows, in priority order: SUM → AVG → MIN → MAX for numbers, MIN → MAX for dates and times. A boolean metric is left alone, and so is a text metric under its default governance; explicitly allowing MIN or MAX on a text field opts it back in.
    • A row-level calculated formula — recomputed at group level when OWOX can prove that returns the same value, so {{revenue}} / {{cost}} becomes the ratio of the totals rather than an average of per-row ratios. One it cannot prove stays a grouping key rather than being guessed at.
    • A column from a joined Data Mart — left alone entirely, because grouping by a joined metric would move its total.

    The report editor fills the choice in for you rather than only predicting it: open a report that sets no aggregation and the function is already ticked on the column, counted on the Aggregations button, and listed in the panel — exactly as one you picked yourself, and changed or removed the same way. While the choice is fresh, a note names the columns OWOX chose for; once the report is saved the rule is simply yours, like any other. Pressing Save is enough to store it, and removing it leaves a note saying what delivery will still do — an unaggregated report on these destinations is collapsed either way. Run history records what was applied.

    The Aggregations panel of a report that set none, reading "Applied automatically because this report set none: cost — Sum. Change or remove it below." above a cost rule aggregated by Sum, with the Aggregations button counting it

    Ad-hoc reads are unchanged: HTTP Data, the MCP query_data_mart tool, apps/ctl, the Looker Studio cache-fill query, “copy as Data Mart”, and a report’s save-time dry run all keep returning exactly what was asked for, duplicates included — including pulling somebody’s Google Sheets report over HTTP Data, where the caller is a third party rather than the report’s reader. The two pull destinations part ways: a Microsoft Excel report collapses like every other report, because the add-in fetches the very rows it writes into the workbook, while a Looker Studio report stays uncollapsed, because its connector reads the report the way any other ad-hoc caller does. See Report Aggregations for the full list of cases where a report is deliberately left uncollapsed.

  • ccff2c9: Google Sheets source: pick the sheet tab from a list and a leaner setup formSection titled “Google Sheets source: pick the sheet tab from a list and a leaner setup form”

    The Google Sheets source setup now lists the tabs of the selected spreadsheet so the Sheet Name can be picked instead of typed, for both Google OAuth and Service Account authentication. With Google OAuth the spreadsheet is chosen with Google Picker only, so the manual “Spreadsheet ID or URL” input no longer appears in that mode. Header Row and Range moved to Advanced Settings because most sheets keep their headers in the first row and import the whole tab.

  • 0e60d60: Limit manual backfill to 31 days per runSection titled “Limit manual backfill to 31 days per run”

    A manual backfill run now covers at most 31 days, so a full calendar month always fits in one run. The Manual Run form explains this limit, shows how many days the chosen period covers, and rejects longer periods before the run starts. The manual-run API rejects such requests with a clear error as well.

    To reload a longer history, start several backfills with consecutive periods, one after another.

    If a backfill longer than 31 days was running when this version was deployed, the resumed run now stops at the limit. Start it again as several shorter backfills.

  • bad3158: Google Sheets: the first sheet of an auto-created document is named after the reportSection titled “Google Sheets: the first sheet of an auto-created document is named after the report”

    When a Google Sheet is created for a report — by add_report through the assistant without a spreadsheet_id, or with the “Create document” button in the report form — its single sheet (tab) is now named after the document title, like the Drive file already was: the report name for add_report, and for the “Create document” button the report title or, while the report has no name yet, the data mart title. Previously that sheet kept Google’s default “Sheet1”, so a document holding several related exports read “Sheet1 / <second report>”: the first report’s tab was the only one not carrying its name. Sheets added to an existing spreadsheet and sheets restored by “Reconnect” were already named after their report; documents created before this change are not renamed. Because sheet names are unique within a spreadsheet, every report added to the same document needs a distinct name, including the first report’s.

  • ca8ac65: AI assistants now use the metrics your Data Mart already definesSection titled “AI assistants now use the metrics your Data Mart already defines”

    An assistant connected through the MCP server now selects a Calculated Field whose formula aggregates — a ROAS written as SUM(revenue) / NULLIF(SUM(adCost), 0), a CTR, a conversion rate — instead of pulling the columns behind it and doing the division in its reply.

    That matters for two reasons. OWOX computes the field in your warehouse over every row matching the question, so the answer stays correct even when the row limit truncates what the assistant sees; arithmetic performed on truncated rows looked like a normal number, and nothing on the number itself showed it was incomplete. And the definition stays the one you wrote — any weighting, exclusion or NULLIF guard inside the formula is applied, rather than quietly lost in a plain division.

    get_data_mart_details_by_id now marks such a field with a usage note, and states that an empty allowedAggregations on it means the value is already computed rather than unavailable.

    Nothing in your Data Marts changes, and no field needs to be redefined.

  • 6b58dc1: Current vendor API versions for the Criteo, LinkedIn, and Facebook connectorsSection titled “Current vendor API versions for the Criteo, LinkedIn, and Facebook connectors”

    The Criteo Ads, LinkedIn Ads, LinkedIn Pages, and Facebook Ads connectors now use current vendor API versions. Criteo moves to 2026-07, LinkedIn to 202609, and Facebook Graph to v26.0. The previous versions were approaching their vendor end-of-support dates, with Criteo expiring first in February 2027. Existing data marts keep working, and you do not need to change any settings.

  • 81559d7: Create Microsoft Excel destinations from the web appSection titled “Create Microsoft Excel destinations from the web app”

    Microsoft Excel is back in the destination type list. Previously the only way to get an Excel destination was to let the OWOX add-in for Excel create one on your first report run. Now you can also add one yourself from the Destinations page.

    The Destination Type list in the new destination panel, with Microsoft Excel highlighted below Google Sheets

    Give it a title and save — there are no credentials to fill in. The automatic path is unchanged: if your project has no Excel destination you can use, the add-in still creates one the first time you build a report. See Microsoft Excel for installing the add-in and working with reports.

    The new destination panel with Microsoft Excel selected, showing only a Title field and the Save button

    Also fixed: picking a “Copy credentials from” source and then switching the destination type no longer sends that source along with the new destination. Previously this failed with an error for types that do not use credentials.

  • d53402e: Snapshot builds are now distributed as container images onlySection titled “Snapshot builds are now distributed as container images only”

    owox@next is no longer published to npm. To run a pre-release build, pull ghcr.io/owox/owox-data-marts:next, or an exact 0.x.0-next-<timestamp> tag to pin one.

    This applies to every package the repository releases, not just owox. In particular @owox/plugin-sdk, @owox/api-client and @owox/ctl no longer get snapshots either, and they are not part of the container image, so changes to them now reach consumers only in a release. Existing snapshot versions stay installable; no new ones appear.

    Releases are unchanged: npm install -g owox still installs the newest release, npm install -g owox@1.8.0 still installs an exact one, and every release still ships a latest container image alongside it.

  • @owox/internal-helpers@0.35.0
  • @owox/idp-protocol@0.35.0
  • @owox/idp-better-auth@0.35.0
  • @owox/idp-owox-better-auth@0.35.0
  • @owox/backend@0.35.0
  • @owox/web@0.35.0

OWOX Data Marts – v0.34.0

  • 75f9d94: MCP report tools: update the report you have instead of creating anotherSection titled “MCP report tools: update the report you have instead of creating another”

    When a report is created through the assistant and the next request changes it — “add a filter”, “sort by revenue”, “rename it” — the assistant now updates that report instead of creating a second one.

    • add_report refuses a report whose fields duplicate one the same user already created on the same data mart and destination, returning error_code: similar_report_exists with the existing report’s definition, so the assistant can switch to update_report. Pass allow_similar: true to create a separate report when that is what the user wants.
    • update_report returns the report as it is after the update — fields, filters, slices, aggregations, date buckets, sort, limit, and for Google Sheets the spreadsheet and sheet. By default it runs a Google Sheets report again when the export changed, so the sheet reflects the new definition. Email, Slack, Microsoft Teams, and Google Chat reports are not re-sent by an update unless run_immediately: true is passed, since a run delivers the message to every recipient or channel; a name-only or message-only change never triggers a run.
    • get_data_mart_reports lists the same definition for every report — including Unique Count metrics and the report-only STRING_AGG / ANY_VALUE aggregations, which the report tools now accept so a UI-created report round-trips — plus created_by_current_user and created_at, so the assistant can recognize an existing report before creating one.
    • Related Google Sheets exports can share one document: add_report accepts spreadsheet_id and adds the report as a new sheet of that spreadsheet instead of creating another file, after checking that the spreadsheet is shared with the requesting user.
    • Rules created in the OWOX UI that the assistant cannot express (post-aggregation constraints, regex, calendar presets) are listed as ui_only_filters and are never dropped by an update.

    The MCP instructions steer the assistant to the report tools for any export or delivery, rather than copying query rows into a file or document through another integration.

  • 305a184: Configure joins before publishing Data MartsSection titled “Configure joins before publishing Data Marts”

    Draft Data Mart joins can now be configured before publication, including their output fields, output aliases, and join-specific descriptions.

  • 4e792b6: Connect Claude or ChatGPT to your Data MartsSection titled “Connect Claude or ChatGPT to your Data Marts”

    Ask questions about your data in plain language and get answers drawn straight from your published Data Marts and their analyst-approved definitions, instead of numbers an assistant guessed on its own. Two equal choices are offered — Claude and ChatGPT. A link to the MCP setup guide walks through it end to end.

    • Setup Checklist — the onboarding checklist has a new step for connecting an AI assistant. Opening the connector does not tick the step off: like running a report, this step is per person and completes on real usage — the first time you ask a question in Claude or ChatGPT and it successfully queries one of the project’s Data Marts, the step flips to done for you. A query that fails does not count, and the checklist can take a few minutes, or a reopen, to catch up.
  • bb0ff9c: Destinations tab now promotes Google Sheets, Excel, and MCP togetherSection titled “Destinations tab now promotes Google Sheets, Excel, and MCP together”

    A published Data Mart with no destinations yet now offers three ways to get value from your data as switchable tabs — Google Sheets, Microsoft Excel, and connecting an AI assistant (Claude or ChatGPT) via MCP — instead of only Google Sheets.

    The active tab cycles automatically every few seconds until you hover, focus, or pick one yourself, so all three options stay visible without cluttering the page with three promo blocks at once.

  • 93ed7ea: MCP: guidance for projects without Data MartsSection titled “MCP: guidance for projects without Data Marts”

    When a project has no published Data Mart visible to the connected user, the MCP discovery tools (list_data_marts, get_relevant_data_marts_by_prompt, and summarize_data_catalog) now return getting_started instead of a bare empty list: a link to create the first Data Mart in OWOX Data Marts, the setup guides, the user’s draft Data Marts that still need publishing, and instructions the assistant relays to the user. The guidance depends on the user’s role — Project Admins and Technical Users are walked through creating and publishing a Data Mart, while Business Users are advised to ask them for one.

    The MCP system instructions tell the assistant to stop rephrasing searches or querying data in that case and to explain what to do next in the web app instead, so a new user hears where to start rather than “no data found”.

  • a00d329: Configure disconnected joined fieldsSection titled “Configure disconnected joined fields”

    The Report Fields tab now keeps disconnected Output Schema fields available for alias, visibility, and aggregation configuration. Reports continue to exclude these fields while they are disconnected.

  • d433bec: Keep report output controls consistent with the column selection and the Data Mart schemaSection titled “Keep report output controls consistent with the column selection and the Data Mart schema”

    • Unchecking a column in the report editor now removes the aggregation and date bucket set on it, a metric filter bound to that aggregation, and its sort rule when the sort can no longer resolve: the report aggregates, the column is a calculated field, or the column is missing from the Data Mart schema. Row filters and slices stay.
    • A report that does not aggregate can sort by any column of the Data Mart, selected or not, the same way a filter works. Once the report aggregates, sorting is limited to the selected columns: adding an aggregation, a date bucket or a Unique Count removes a sort rule on an unselected column. A calculated field is sortable only while selected.
    • An aggregation or date bucket on a column that is missing from the Data Mart schema is reported as a disconnected column, with the column named, instead of a misleading “not selected” or “unknown type” error.
    • A scheduled run, and a Looker Studio data pull, drops a sort rule on a column that is missing from the schema and continues, with a warning in the logs, instead of failing. The limit is kept, so under a limit the delivered rows may differ from the ones the sort used to pick. Saving the report, the Generated SQL preview and ad-hoc queries still report the missing column.
    • The “Output controls validation failed” message now names the rules that failed and the columns they reference, so run history, MCP clients, and the Google Sheets extension show the reason.
  • bb52739: Keep output schema field connection statuses server-controlledSection titled “Keep output schema field connection statuses server-controlled”

    Output schema updates no longer accept a client-provided connection status as the field’s actual state. API users and plugin developers could previously encounter misleading connected fields when saving a schema before the storage had verified them; the backend now derives the status, and the web app omits it from update requests.

  • cce4c7e: Model Canvas and Data Mart navigation polishSection titled “Model Canvas and Data Mart navigation polish”

    The Model Canvas page (formerly “Models”) now opens straight into your data model: with a single storage, it is selected automatically instead of asking you to pick one first. When you do need to switch, the storage picker moved up into the page header as a clear “Model for [Storage]” title, and a “Data Marts” breadcrumb takes you back to the list in one click — the same breadcrumb now appears on the Reports, Triggers, Run History, and Insights pages too.

    On the canvas itself, each Data Mart card now shows a full-height color stripe reflecting its Data Quality status, so you can spot cards with warnings or errors at a glance instead of opening each one. Downloading the canvas (as an image, JSON, or OKF Markdown) is now a dedicated button in the toolbar instead of being tucked inside the Actions menu, and bulk actions like Publish and Delete now state exactly how many Data Marts they will affect.

    Opening the canvas with no filters applied now shows all Data Marts, published or draft, connected or not — previously it defaulted to published-only, connected-only.

    Long Data Mart names in tables no longer wrap and push rows taller — they truncate with the full name available on hover. In the Reports table, the Report column now leads, with Data Mart moved to second place.

  • 29a9def: Fix the Model Canvas failing with a full-page error on larger data modelsSection titled “Fix the Model Canvas failing with a full-page error on larger data models”

    Opening the Model Canvas — or the Joinable Data Marts diagram — for a storage with many Data Marts and relationships could replace the whole page with a “Something went wrong” error screen right after the loading skeleton. Storages with only a few Data Marts were unaffected, so the failure was intermittent and hard to reproduce.

    The automatic layout step throws on some otherwise-valid relationship graphs — typically when a model has several relationships between the same pair of Data Marts, a Data Mart joined to itself, or circular joins — and that error was not handled, so it propagated to the route error boundary and took down the entire page instead of just the canvas.

    The canvas now recovers instead of crashing. When the layout step fails it retries with a simplified view of the relationships, which resolves the vast majority of these cases and still produces a proper layout. If that also fails, it falls back to a plain grid you can rearrange by dragging. Node positions you set are still saved per storage.

  • 5960148: Open the OWOX add-in for Excel straight from the Destinations tabSection titled “Open the OWOX add-in for Excel straight from the Destinations tab”

    A Data Mart that has no Microsoft Excel reports yet now links to the OWOX add-in for Excel, so a business user reaches the place an Excel report is actually created in a single step instead of being told where to look. The Excel destination’s description also links to the installation guide, for organizations that install the add-in manually.

  • c31c75f: Data Studio: complete responses for large reports, and real query errorsSection titled “Data Studio: complete responses for large reports, and real query errors”

    • Large reports (c31c75f) — Data Studio reports now return as many complete rows as fit within Apps Script response limits instead of ending with malformed JSON.
    • Query errors (3e7b118) — when loading a Data Mart schema or its data fails, Data Studio now shows troubleshooting guidance and the underlying query error instead of suggesting that the deployment URL is incorrect.
  • 3f5b6e6: Fix LinkedIn Ads adAnalytics silently dropping data when the response exceeds 15,000 elementsSection titled “Fix LinkedIn Ads adAnalytics silently dropping data when the response exceeds 15,000 elements”

    Previously, an adAnalytics export over a large date range could silently lose data: the endpoint does not support pagination and caps its response at 15,000 elements, so campaigns from the beginning of the period were missing from the result. The connector now fetches, saves, and checkpoints analytics one day at a time, so a single day cannot exceed the limit, and an interrupted run resumes from the last completed day instead of restarting the whole range.

    If a daily response still reaches 15,000 elements, the import finishes with a Warning status that lists the affected days instead of losing data silently.

    The connector also refreshes the LinkedIn access token once per run instead of before every request. LinkedIn API errors now fail the run instead of silently returning no rows; rate-limit (429) and server (5xx) errors are retried first. Analytics days are now computed in UTC, so the day the connector requests always matches the day it logs and checkpoints, regardless of the runner’s time zone.

  • 6457d34: Temporary BigQuery faults no longer fail the whole importSection titled “Temporary BigQuery faults no longer fail the whole import”

    Previously, a single batch that BigQuery failed with a temporary fault ended the entire run, even after hundreds of thousands of rows had been stored. The storage now waits and saves that batch again, using the existing Max Fetch Retries and Initial Retry Delay settings. This covers both a batch BigQuery rejects outright and one it accepts and then fails while running. Errors that another attempt cannot fix, such as an invalid query or a denied permission, still stop the run immediately.

  • 6d12fdf: Keep automatic project redirects scoped to the signed-in userSection titled “Keep automatic project redirects scoped to the signed-in user”

    OWOX Data Marts no longer carries an automatically remembered project into a session for a different user. After switching accounts during sign-in, users continue with a project available to the newly authenticated account instead of being sent to request access for the previous account’s project.

  • 6449cb1: Projects in the switcher are sorted alphabeticallySection titled “Projects in the switcher are sorted alphabetically”

    The Switch project menu now lists your projects in alphabetical order instead of the order the server happened to return them in, so a project always sits where you expect it. Sorting ignores letter case, so alpha and Alpha are ordered together, and it reads numbers as numbers: Project 2 comes before Project 10.

  • @owox/internal-helpers@0.34.0
  • @owox/idp-protocol@0.34.0
  • @owox/idp-better-auth@0.34.0
  • @owox/idp-owox-better-auth@0.34.0
  • @owox/backend@0.34.0
  • @owox/web@0.34.0

OWOX Data Marts – v0.33.0

  • b3d008d: Calculated Fields in the Output SchemaSection titled “Calculated Fields in the Output Schema”

    An analyst can now define a Calculated Field directly in a Data Mart’s Output Schema — a formula written in that Data Mart’s own warehouse SQL, such as SUM(clicks) * 1.0 / NULLIF(SUM(impressions), 0) for a click-through rate. Unlike a ratio hard-coded into a Data Mart’s SQL, it recomputes at whatever grain the request asks for, so a report broken down by day, a Google Sheet, a Looker Studio page, an MCP agent, an HTTP Data caller and a report’s Totals row all get the true ratio of the sums rather than an average of per-row ratios.

    Whether the field is a metric or a dimension is read from the formula rather than chosen, and the formula is checked when the schema is saved — every problem names the field it belongs to, and the resulting query is test-run against the warehouse once per save, so a broken formula is caught before anyone builds a report on it. Available on Google BigQuery, AWS Athena, Snowflake, AWS Redshift and Databricks; see Calculated Fields for what each dialect supports and the limitations to know about.

    Six changes to the Calculated Field editor (ad7d572), from feedback on the released version:

    • A row-level field keeps its “Σ available” control while its formula is edited. Touching the formula of a calculated field that is a dimension took that control off the row until the schema was saved, which read as the setting having been lost.
    • Unsaved edits survive a background refresh. The Output Schema editor reset itself whenever the Data Mart was re-read — which happens on its own, without anyone asking — and a formula typed but not yet saved went with it, while a failed save’s errors stayed on screen naming a field no longer in the table. It now resets only when the saved schema has actually changed.
    • A rejected formula is named where the refusal appears. Choosing “Save & leave” on the unsaved-changes prompt reported Request failed with status code 400, which names neither the field nor the fix. It now names the field the warehouse rejected.
    • A calculated field no longer stretches its row. A formula written over several lines made its row that many lines tall. The row now shows at most two lines — the most it can show without growing — and hovering it shows the whole formula with the author’s own line breaks.
    • The division warning is about the formula in front of you. It quoted NULLIF(SUM(impressions), 0) whether or not impressions appeared anywhere in the formula, and read like a refusal although it blocks nothing. It now quotes the denominator this formula actually divides by, underlines it in the editor, and says plainly that it is advice. It quotes the whole denominator or nothing at all — never a piece of one, since guarding a piece leaves the division just as broken and silences the warning.
    • COALESCE no longer counts as a guard against that division. What fails a query is a zero denominator, not a null one — and COALESCE(SUM(x), 0) produces exactly the zero that fails it. The message and the setup guide now say zero rather than “zero or empty”, and the guide explains why COALESCE does not count. The guide also gains a worked example for rates over a condition, such as an activation rate.
  • a4dc4ff: Send Reports to Microsoft Excel DestinationSection titled “Send Reports to Microsoft Excel Destination”

    Open the OWOX add-in in a workbook, pick a published Data Mart, and the report’s rows land in the worksheet. The report then appears in OWOX Data Marts with its title, its run history and its last-run status, and its columns, filters and sorting can be edited there afterwards.

    A workbook sits on your machine, so nothing can be pushed into it — the add-in reads the report itself. Four things follow from that:

    • You are not asked to name a document in advance. The worksheet you opened the add-in from is where the rows go.
    • Such a report is created in the add-in, which is what binds it to a worksheet. The web app offers neither the report nor the destination for creation: the add-in sets both up on first use, and a report made in the web app would be one no workbook points at.
    • It cannot be run or scheduled from the web app. The web app says so instead of offering a button that would fail; refresh it from the add-in.
    • Refreshing from the add-in is the report running, so the status shown in Excel and in the web app is the same one.

    Runs from Excel are billed as Excel Report Run, a sub-consumption unit of Report Run.

  • 47a3dc3: Reuse project Credentials across pluginsSection titled “Reuse project Credentials across plugins”

    Projects can now store and manage reusable Credentials with owners, sharing, Contexts, operational state, and dependency visibility. Plugins declare exact provider or logical AI requirements and members explicitly select eligible Credentials during installation, while raw secrets remain in the host. External Credential definitions can be added from public GitHub repositories or private repositories available through the deployment’s configured GitHub access. Plugin authors use the same typed exactCredential helper for built-in and external definitions.

    See Credentials and Credential definitions for setup and authoring details.

  • cb22b51: Report output column schemaSection titled “Report output column schema”

    GET /reports/:id/output-schema returns the columns a report’s rows will carry — name, title, description and type — including aggregated columns (revenue | SUM), Unique Count and calculated fields, which do not exist on the Data Mart schema.

    Joined fields in Excel reports now use the same header form as Google Sheets: Field name (Data Mart name), so the field name stays visible in a narrow spreadsheet cell.

  • 23f0d09: Plugin release rejections are now visible, and collection compatibility follows SemVerSection titled “Plugin release rejections are now visible, and collection compatibility follows SemVer”

    When a plugin’s GitHub release was rejected during sync, the reason was stored in the database and communicated nowhere: no log line, no audit detail, nothing in the UI. From the outside it looked like “Check now did nothing” — the plugin stayed on an old version with no explanation. Now every publisher-fixable rejection is visible in three places:

    • Application logs — both the daily check and manual “Check now” log a warning per rejected release, e.g. owner/plugin: v0.1.2 rejected (COLLECTIONS_INCOMPATIBLE) — Collection "dashboards" cannot change entity binding. Rejections that are permanent by design (drafts, prereleases, non-SemVer tags) stay out of the log so they cannot bury the one line that matters.
    • Audit trail — the UPDATE_CHECK audit record now carries the rejection summary, not only thrown transport errors.
    • Plugin page — publishers see a “Release issues” card listing each rejected release with its code, reason, and the time of the check, and “Check now” tells a publisher directly when the fresh check rejected a release instead of reporting “up to date”.

    The collection-compatibility safeguard is also back, now gated on SemVer: within a compatibility line a release may add collections and change action mappings, but cannot remove a collection or change its name, scope, or entity binding — the previous version stays current and the reason says how to proceed. Opening a new line waives the check, so a breaking collection change ships by declaring itself as one: a major version bump, or a minor bump while the plugin is still below 1.0.0 (where SemVer promises no stability).

  • 885b1ba: Report filters: one “is blank” pair instead of four null/empty operatorsSection titled “Report filters: one “is blank” pair instead of four null/empty operators”

    A blank spreadsheet cell renders identically whether the warehouse holds NULL, an empty string, or whitespace — so offering “is empty” and “is null” side by side forced users to guess, and the wrong guess silently returned 0 rows. The filter picker now offers a single pair, is blank / is not blank, defined by what the cell shows rather than by warehouse state.

    The pair is type-aware on every supported storage (BigQuery, Athena, Redshift, Snowflake, Databricks): a string column is blank when it is NULL, '', or whitespace-only; a number, date, time, or boolean column is blank only when NULL. The MCP query_data_mart and report tools advertise the same single pair, and the Data API accepts is_blank / is_not_blank in filter rules.

    Nothing saved breaks: reports and API calls that use the previous is_empty, is_not_empty, is_null, or is_not_null operators keep working with their exact previous semantics, and saved rules still display under their original names. Only the menus stop offering them.

  • dc3b2ad, 4bfd303: Join descriptions: shown in the column pickers, and editable per join path

    • Shown in the pickers (dc3b2ad) — report column pickers in the web app and the Google Sheets extension now show each joined Data Mart’s description and full join path on hover, using the same user-facing titles as the picker.

    • Editable per join path (4bfd303) — a join description explains what pulling in another data mart actually means: business users read it in the report column picker, and AI assistants read it over MCP. It was written once on the relationship itself, so every data mart that reached that join showed the same sentence. “Orders placed by this customer” is right in a Customers data mart, but in a Companies data mart that reaches Orders through Customers it explains nothing.

      The Description tab of an inherited join is now editable. The inherited text stays visible as a placeholder, and whatever you type applies to that join path only — the relationship itself, and every other data mart that inherits it, keep their own wording. Clearing the field, or the new Reset to inherited button, falls back to the inherited description. Nothing changes until someone types: a join with no description of its own behaves exactly as before. Wherever a description is shown — the join-path tooltip in the report column picker, and joins[] in the MCP data mart details — it is the one that applies to that particular path.

  • d13a90d: Schedule Data Mart triggers in UTCSection titled “Schedule Data Mart triggers in UTC”

    Scheduled triggers now offer UTC as an explicit fixed-offset timezone with no daylight saving time changes. Existing UTC aliases remain compatible, and legacy Europe/Kiev identifiers are shown as Europe/Kyiv without rewriting saved schedules.

  • 67f5afa: Cheaper incremental imports into Google BigQuerySection titled “Cheaper incremental imports into Google BigQuery”

    Incremental connector runs into Google BigQuery now read only the days they update. Before, every incremental run scanned the whole destination table, including all of its history, so as a table grew each daily import cost more even when it brought the same amount of new data. Now the import touches only the partitions for the imported dates, so the cost stays flat over time. This applies to time-series tables partitioned by date, such as ad performance reports. Entity tables without a date column, such as campaign lists, keep their current behavior. Users see lower BigQuery query costs and fewer daily quota errors.

  • 9375f07, de49dce: Resumable incremental imports for Microsoft Ads, Google Ads, Reddit Ads, Criteo, and X Ads

    Previously, an interrupted incremental run restarted the whole date range from the beginning, so long imports could retry endlessly without ever finishing. These connectors now save their progress as the import advances — Microsoft Ads after each imported day (9375f07), and Google Ads, Reddit Ads, Criteo and X Ads as the import advances (de49dce) — so an interrupted run resumes from the last completed date instead of starting over.

    Microsoft Ads also stops early with a clear message when the Account IDs field holds no usable ID, or when Fields references a report the source no longer provides.

  • da26589: Connector runs interrupted by a restart are reported as a warningSection titled “Connector runs interrupted by a restart are reported as a warning”

    When a server restart stops a connector run, the run resumes by itself and continues from the last day it finished. Run history now shows this as a warning instead of an error, so a restart no longer looks like a failed import. Runs that stop for any other reason are still reported as errors.

  • d013726: Checkout and cart tokens for Shopify ordersSection titled “Checkout and cart tokens for Shopify orders”

    The Shopify connector now offers checkoutToken and cartToken fields on the orders node. Select them under orders in the connector’s field picker to import them. checkoutToken matches checkout.token in Shopify Web Pixel events, so you can join orders with checkout_completed events; cartToken identifies the cart that produced the order. Orders with no associated checkout or cart return NULL. Orders imported before you enable the fields stay NULL until you re-import them with a backfill run.

  • a25b495: Connect Google Sheets sources with Google OAuthSection titled “Connect Google Sheets sources with Google OAuth”

    Users can authorize a Google account and choose a spreadsheet with Google Picker when configuring a Google Sheets source. Service account authentication remains available.

  • 94c23bc: Labeled retry settings in connector Advanced settingsSection titled “Labeled retry settings in connector Advanced settings”

    Every connector’s Advanced settings form now shows Max Fetch Retries and Initial Retry Delay (ms) instead of raw keys. The hints state that the delay is in milliseconds and that the retries value counts total attempts, including the first request. This prevents a delay of 5 being read as five seconds when it means five milliseconds.

    The Facebook Ads troubleshooting guide now explains Meta’s Ads Insights rate limit (code 4, subcode 1504022), the ad-account score limit (code 17, subcode 2446079), and how to adjust these settings after a rate limit error.

  • f7b099a, 7667281: Joinable Data Marts diagram opens fitted to the whole graph, and its zoom controls keep working

    • Opens fitted (f7b099a) — switching to the Graph view could land on a viewport zoomed in on the root card, as if the fit had never run, until “Fit to view” was pressed. The automatic first fit relied on fitView, which only accounts for nodes whose DOM dimensions are already measured — on first mount it ran against a half-measured subset. The initial viewport is now derived from the layout geometry itself, so the Graph view always opens showing the entire diagram, exactly as the “Fit to view” button leaves it.
    • Zoom controls keep working (7667281) — the +/- zoom buttons could stop responding until “Fit to view” was pressed, depending on how the page was opened. The allowed zoom range was captured once from a completed fit, so a fit that ran under transient conditions (a still-loading graph or a settling pane) froze the range in an unusable state. The range is now derived from the live graph and pane geometry, small graphs that fit at the maximum zoom keep a usable zoom-out floor, and a corrupted viewport recovers with a full fit instead of ignoring clicks.
  • 3c16cd6: Help videos now stream from Cloudflare instead of GitHub attachmentsSection titled “Help videos now stream from Cloudflare instead of GitHub attachments”

    Three in-product help videos — “SQL to Google Sheets in Minutes”, “Data Studio Setup”, and “Getting Started with Data Marts” — used to load from GitHub issue-attachment storage. Those URLs redirect through short-lived signed links and fail entirely for clients whose networks block github.com, so the videos could show up broken.

    Now all six help videos use the same Cloudflare Stream player. The three migrated videos keep their previous behavior (autoplay, muted, loop) and their exact aspect ratios.

    The documentation site picks up the same fix: pages that embedded the GitHub-hosted files now render the Cloudflare player, while the repository markdown keeps the GitHub links so videos still play inline on github.com. The Google Sheets destination page also gains the setup walkthrough video.

  • @owox/internal-helpers@0.33.0
  • @owox/idp-protocol@0.33.0
  • @owox/idp-better-auth@0.33.0
  • @owox/idp-owox-better-auth@0.33.0
  • @owox/backend@0.33.0
  • @owox/web@0.33.0

OWOX Data Marts – v0.32.0

  • 3674b59: Count unique records from any joined Data MartSection titled “Count unique records from any joined Data Mart”

    Every joined Data Mart now offers a Unique Count field in the report column picker, counting distinct records by that Data Mart’s primary key. It answers “how many orders per customer” without adding an order column to the report. Composite keys are supported. The field can be selected and sorted by, but not filtered or fed into another aggregation.

    In the picker it is simply Unique Count, listed under the joined Data Mart it belongs to, with a tooltip naming that Data Mart and the key columns being counted. In the produced file it is named like every other joined field: Unique Count (Orders) in Google Sheets, Orders Unique Count everywhere else.

    With no usable primary key the field is shown disabled, and its tooltip says which Data Mart and why — no key set, part of the key disconnected, a nested key, or both of the last two. The report’s own Data Mart says instead that no primary key is available for reporting.

    A key column marked Hidden for Report still counts, on the report’s own Data Mart as well as on a joined one. Hiding a column takes it off the list of fields a report can show; counting distinct values of it puts nothing in the output, so there is nothing to hide. On the report’s own Data Mart this is new — such a key used to leave the row disabled.

    Some existing reports will show different numbers, and the new ones are correct. Rows whose declared primary key is empty used to be treated as one single record. That was wrong in three ways, and all three are fixed. A Data Mart with no declared key, or one whose key is always filled, is untouched.

    • Unique Count on a composite key was too high and now goes down. Every row with an empty key component counted as one extra record — while a single-column key had always ignored such rows, so the same declaration behaved differently depending only on how many columns it had. Applies to any Data Mart with a composite key, joined or not.
    • Sum and Average on a joined Data Mart were too low and now go up. All rows sharing an empty key collapsed into one, so the metric read a single row instead of all of them — verified on BigQuery, where twenty such rows summed to the value of one row before the fix and to all twenty after. Min, Max, the percentiles and Combined can shift for the same reason. The two metrics part ways on such a row on purpose: Unique Count does not count a record with no identity, while Sum still reads its value, so one report can legitimately say 0 orders, worth 2000.
    • A key unique only within what it is joined on is now counted per join key, and the number rises sharply. A line number restarting at 1 in every order reported 12 — the largest order’s line count — for a book of 4,300 line items. It now reports 4,300, which is what Sum and Average on that same key had always meant by it.

    Cached report data is cleared on upgrade, so an unedited report shows the corrected numbers on its next run rather than serving what it had cached.

    Two smaller corrections in the same direction: two different composite keys that ran together into the same text now count as two records rather than one, and on Amazon Redshift a composite key holding a text part longer than 256 characters is no longer cut short and merged.

    One correction goes the other way, and removes a column rather than changing a number. On the report’s own Data Mart, a composite key one of whose columns had disconnected from the source was quietly counted by the columns that remained — which merges records the full key keeps apart, so the count read low with nothing to say so. Such a key now withholds the metric entirely, exactly as a joined Data Mart’s already did. Reconnect the missing column, or drop it from the key, and the count comes back.

    Primary-key declarations worth a second look. Only where the report’s own Data Mart is counted on a composite key, whose columns are joined into one value and read as text:

    • a floating-point part can make two records count as one, when their values differ only in digits beyond what the text form shows;
    • a date-time part can do the same — a warehouse writes it at whatever precision it is configured to show, thousandths of a second on Snowflake by default — so two records moments apart count as one;
    • a binary part can stop the report with the warehouse’s own error message rather than an OWOX one.

    Prefer whole-number or text columns, or dates without a time part. A joined Data Mart’s Unique Count reads every part of its key as it stands, so none of this applies there. A key declared on a structured column — JSON, geographic, variant — is not one a warehouse can group by, and stops the report either way.

  • d9356e1: Require an OWOX Data Marts Cloud license for Report RunsSection titled “Require an OWOX Data Marts Cloud license for Report Runs”

    Every Report Run — Google Sheets, Looker Studio, Email, Slack, Google Chat, MS Teams, HTTP Data, and MCP Query — is now authorized with OWOX Data Marts Cloud before it executes, and its consumption is billed to the project signed into the license key. Without a valid key the deployment runs as Community and Report Runs finish as Restricted with an actionable message; configuration is never hidden or deleted, and Process Runs are unaffected.

    Community deployments can now configure every destination type, including Email, Slack, Google Chat, and MS Teams — the license gates only Report Run execution, not configuration.

    Project administrators can create, rename, and revoke license keys in Project Settings → License keys. A key is bound to one deployment origin, is shown once on creation, and expires after 365 days. To rotate, create a new key, update the deployment, then revoke the old one.

    Legacy offline Enterprise license keys are no longer supported and fall back to Community.

  • a30ab25: Plugin collections: host-stored JSON documents for installed pluginsSection titled “Plugin collections: host-stored JSON documents for installed plugins”

    Plugins can declare project- or member-scoped collections in plugin.json and access them through ctx.collections(name). The host persists JSON documents, applies the collection’s entity authorization rules, and keeps credentials out of the plugin frame.

    The SDK collection facade supports paginated list, nullable get, put, and delete operations. Entity-bound writes include an immutable parentId, allowing the host to authorize every operation against the current member’s access to the parent entity.

  • 918026f: Data Last Updated Date for AWS Redshift, AWS Athena, Snowflake, and DatabricksSection titled “Data Last Updated Date for AWS Redshift, AWS Athena, Snowflake, and Databricks”

    Data Last Updated now works for four more storages, on every surface where it already worked for BigQuery: the Data Mart page, the Data Marts list, the model canvas, and MCP query_data_mart responses. In each case OWOX asks the storage which tables the executed query reads — views and SQL Data Marts resolve through to their underlying base tables — and then reads each table’s last data change. When a table that cannot be measured sits alongside measurable ones, the coverage becomes partial, so the reported value stays a lower bound: the real time can only be more recent.

    • AWS Redshift (918026f) — reads each table’s last data modification time from Redshift’s own metadata; materialized views report the time of their last refresh. No setup is needed. Expect the value to trail reality slightly: Redshift refreshes this metadata with a delay of up to about five minutes, so a load that just finished may take a moment to show. Spectrum external tables, whose data lives outside Redshift, and tables Redshift reports no modification time for show as unknown.
    • AWS Athena (53e2b84) — what can be reported depends on the table format. Iceberg tables answer exactly: the value is the time of the last commit to the data, visible right after a write. Classic Hive tables show Unknown instead — their catalog stores no record of when the data changed, and the timestamps it does keep also move when someone only edits the table definition, so reporting one could present stale data as freshly updated. Tables from federated catalogs are not measured either. No setup is needed.
    • Snowflake (b81cdb2) — reads each table’s last data change from the account’s DML history. The connection role needs access to the SNOWFLAKE database (account usage) — without it, Data Marts show Unknown. The value is the start of the hour in which the data last changed, and it can trail reality by several hours, because Snowflake publishes this history with a delay. Schema changes and Snowflake’s own background maintenance do not move the value, so the reported time never overstates how fresh the data is. A table with no recorded data changes in the last year shows Unknown with a note, and a dropped-and-recreated table answers for its current generation only, never for its predecessor’s history. Materialized views, Iceberg tables (whose data can change outside Snowflake), and other objects the history cannot answer for appear as unknown sources.
    • Databricks (15166aa) — reads each Delta table’s history for the last commit that actually changed data; the answer is exact and immediate. Maintenance operations do not count: a table that was only compacted by OPTIMIZE or cleaned by VACUUM keeps the timestamp of its last real write. Delta table history is kept for a limited period (30 days by default), so a table whose data last changed earlier shows Unknown with a note. Non-Delta and external tables have no history and show Unknown too. No setup is needed.
    • In MCP (e387fee) — the AI assistant is now asked to present the data_last_updated timestamp in a business-friendly form (e.g. “August 4, 2026 at 09:46 UTC”, converted to the user’s time zone when known) instead of echoing the raw ISO-8601 value. The structured response still carries the ISO-8601 timestamp for machines.
  • 7edd34f: The “Row Count” column is removed from aggregated reportsSection titled “The “Row Count” column is removed from aggregated reports”

    Previously, every aggregated report automatically included a Row Count (COUNT(*)) column, even though it was never selected. The column has been removed: reports — and the MCP and HTTP data endpoints — now return only the columns you selected. If you need the number of underlying rows per group, apply the Count aggregation to a column that is always filled (an ID column works well); to count unique entities, use a Unique Count. A column of your own named “Row Count” is also no longer rejected in aggregated reports.

    What changes in existing reports:

    • Google Sheets — on its next run, an aggregated report removes the Row Count column from the sheet structurally, shifting everything to its right one column left. Formulas and pivots referencing that range move with it — repoint anything that read Row Count to an explicit Count column first.
    • Data Studio — Row Count was a schema field of the data source, and Looker Studio keeps requesting fields by their stored names. A chart built only on Row Count starts failing with an error naming the missing field; a chart that also uses other fields keeps working but silently drops the Row Count column. Open the data source and use Refresh fields, then rebuild the affected metric on an explicit Count field. Because report data is cached, individual reports switch to the new column set gradually as their cache entries expire (from minutes up to the cache lifetime configured on the report).
  • 94e32a1: Optional description for Data Mart relationshipsSection titled “Optional description for Data Mart relationships”

    Each join between Data Marts can now carry an optional free-text description explaining the business meaning of the relationship (e.g. “Visitors from the website sign up for the product and convert into users”).

    • Data Setup UI — a joined Data Mart’s card gets a new Description tab next to Join Settings, with an autosaving text area. Inherited (transitive) joins show where to edit the description instead.
    • API — POST/PATCH /data-marts/:id/relationships accept an optional description, and relationship responses return it. Sending null or an empty string clears it.
    • MCP — get_data_mart_details_by_id with detail_level=with_joined_fields now returns a joins array — one entry per join edge with the joined data marts, the join key fields, and the relationship description — so AI assistants understand not only which fields a join contributes but how and why the data marts relate.
  • e387fee: Joinable Data Marts diagram: filters on the toolbar and the same view settings as the Models canvasSection titled “Joinable Data Marts diagram: filters on the toolbar and the same view settings as the Models canvas”

    The Status and looped data marts filters moved out of the diagram’s gear popover onto the toolbar, right next to search — the same place the Models canvas keeps its filters. They are now dropdowns (All statuses / Published only / Draft only, and Hide looped data marts / Show looped data marts). More importantly, the filters now cover both views: previously they only trimmed the Graph, so switching to List silently showed a different set of relationships. The List now applies the same rules as the diagram — including subtree semantics, where hiding a data mart also hides everything joined through it. Saved filter preferences carry over unchanged.

    The gear is back, and now holds the same view settings as the Models canvas, shared component and all:

    • View — switch cards between Compact and Detailed; Detailed shows the joined mart’s field rows (name + type, hidden-for-reporting fields dimmed), collapsed to 4 rows with an in-place “+N more” toggle.
    • Layout algorithm — Horizontal or Vertical; the diagram now uses the same dagre layout engine as the Models canvas.
    • Show join fields — join conditions (source = target) rendered as labels on the relationship lines.
    • Object labels — toggle the input-source badge and accent stripe, the field count, and the status dot per card, with “Check all” / “Uncheck all” shortcuts.

    All settings are remembered per browser, and the inline and fullscreen diagrams stay in sync.

  • e387fee: Include field descriptions in the Models canvas exportsSection titled “Include field descriptions in the Models canvas exports”

    The Models canvas export dropped the business descriptions of Output Schema fields: in the OKF bundle the schema table’s Description column carried only the PK/FK notes, and the JSON model graph omitted the field descriptions entirely. Both formats now carry each field’s description as stored in the Data Mart:

    • OKF (Markdown) — the Description cell reads PK. → description. Multi-line descriptions collapse to a single line so they cannot break the Markdown table. The FK to [Target] notes left the schema table entirely: relationships live only in the Joins section, which already lists every join with its condition.
    • JSON — every schema field gains a description property when one is set, staying compatible with the OWOX Model Canvas graph format.
  • cf830e3: Give MCP assistants the complete project contextSection titled “Give MCP assistants the complete project context”

    MCP clients now receive the full project description through get_project_context, together with the current project metadata. Assistants are instructed to call the tool before their first project-specific operation, while the always-on MCP initialization instructions remain focused on OWOX workflows, security, and tool usage.

    Project descriptions continue to support up to 10,000 characters and are returned without truncation. After an admin updates the description, the new value is available on the next get_project_context call without reconnecting the MCP client.

  • 08ebbda: Run newly created push-destination reports immediately through MCPSection titled “Run newly created push-destination reports immediately through MCP”

    Newly created push-destination reports now run immediately by default when created through MCP add_report, and the response returns the initial run_id for status polling. Use run_immediately: false for configuration-only creation; Looker Studio remains pull-based. If queueing fails after creation, the response preserves the report id and directs the assistant to retry with run_report instead of creating a duplicate report or Google Sheet.

  • 0301c9e: Google Chat: direct webhook delivery, clearer setup, and delivery logsSection titled “Google Chat: direct webhook delivery, clearer setup, and delivery logs”

    • Direct delivery (0301c9e) — Google Chat Destinations now offer Incoming Webhook and Channel Email delivery methods. Existing email-based destinations keep their current method, while new destinations default to direct delivery through a space-specific incoming webhook. Webhook messages are formatted cards containing the subject, Data Mart name, rendered report, and a link back to OWOX Data Marts.
    • Setup guidance and logs (953f107) — destination settings now provide contextual instructions for choosing a delivery method and obtaining an incoming webhook URL or space email address. Report run logs identify the delivery method and destination without exposing webhook credentials, with per-part entries shown only when a report is split across multiple Google Chat messages.
  • 2549692: Creating a report no longer stops you at “publish the Data Mart first”Section titled “Creating a report no longer stops you at “publish the Data Mart first””

    Before, if your Data Mart was still a draft, the New Report button was simply greyed out with a tooltip saying “publish the Data Mart first” — you had to leave the page, find the publish button, publish, and come back to try again. Now clicking New Report on a draft Data Mart takes you straight to the next step:

    • If the Data Mart is ready to publish, you’ll see “Publish and create report” — one click publishes it and opens the new report form right away.
    • If something is still missing (like a Storage or an Input Source), you’ll see exactly what’s missing and a “Review Data Setup” button that takes you right there.
  • 3f94296: Reliable bulk publishing for Data Mart draftsSection titled “Reliable bulk publishing for Data Mart drafts”

    Previously, publishing Data Mart drafts in bulk from a Storage’s menu could fail for every draft with a generic “please check them independently” message, even when some drafts were ready to publish. This is now fixed, so drafts you have access to publish succeed as expected. When a draft still cannot be published, the notification names the reason — that it has no definition yet, for example, or that its table name is not in the expected format — so you know what to correct before trying again.

  • 6f2364e: AI metadata generation for SQL data marts: no dataset-creation permission, and visible failuresSection titled “AI metadata generation for SQL data marts: no dataset-creation permission, and visible failures”

    • No bigquery.datasets.create required (6f2364e) — generating AI metadata (field aliases and descriptions, data mart title and description) for a SQL-based data mart previously materialized a technical view in the owox_internal_<location> dataset, and creating that dataset on first use required the project-level bigquery.datasets.create permission — so users with data-only access (e.g. BigQuery Job User plus BigQuery Data Viewer) could not use the AI helper at all. Sample rows are now fetched through an inline derived table built from the data mart’s own SQL (SELECT <columns> FROM (<data mart sql>) LIMIT <n>), so the AI helper needs exactly the permissions a regular report run needs:
      • Plain SELECT definitions are inlined on every storage type except legacy Google BigQuery.
      • WITH/CTE definitions are additionally inlined on Google BigQuery storage, where the SQL dialect accepts a CTE inside a parenthesized derived table; other storages keep the previous behavior for CTEs.
      • Legacy Google BigQuery data marts, SQL that cannot be inlined, explicit table references, and non-SQL definitions (table, view, pattern, connector) continue to use the technical-view path, with the existing human-readable error if permissions are missing there.
    • Failures are now visible and understandable (24932f6) — a missing BigQuery permission used to fail the run before the AI step, and the error toast auto-dismissed in seconds, so users never learned why nothing was generated. Generation errors now show as persistent, dismissible notifications instead of an auto-dismissing toast; leaving the page mid-generation leaves a notice that the run was cancelled. BigQuery permission errors are rewritten into a human-readable message naming the project and the missing permission, with the raw error expandable for support. AI helper trigger logs now include dataMartId and projectId, and the AI insights facade logs the caught error for failed metadata generation, so production incidents can be found by filtering logs on the data mart id.
  • 03172e5: Fix: “Copy from…” no longer shares or deletes another Data Mart’s credentialsSection titled “Fix: “Copy from…” no longer shares or deletes another Data Mart’s credentials”

    Using “Copy from…” to reuse a connector configuration from another Data Mart made both Data Marts point at the same stored credentials instead of copying the values. Saving the new Data Mart could then delete those shared credentials, wiping them on both.

    Copying a configuration that holds manually entered credentials now always creates its own credentials record, and a Data Mart can no longer write to or delete credentials that belong to another one. Data Marts that already share a record are separated automatically the next time each of them is saved.

    A few details worth knowing:

    • Connections authorized through OAuth are still shared between Data Marts, as they were before — such a credential belongs to the project rather than to a single Data Mart.
    • Credentials already wiped by this bug are not restored; those Data Marts need their credentials entered again.
    • A copied Microsoft Ads configuration starts from the source’s current (rotated) refresh token, so the copy keeps working even when the originally entered token has already expired. From there each Data Mart rotates its own token independently.
    • Deleting a Data Mart no longer removes credentials that another Data Mart still references: the record is handed over to the referencing Data Mart instead, so a not-yet-separated pair cannot lose data through deletion.
    • Copying a configuration with “Copy from…” now requires edit access to the source Data Mart for every connector: you must be its technical owner, or it must be shared with you for maintenance.

    The error shown without that access now explains how to get it (8b29435): ask an owner to add you as a Technical Owner or to turn on “Shared for maintenance”, or enter the credentials manually. The connector setup guide now documents the “Copy from…” option — who can copy credentials, the exact error text, and how to resolve it.

  • 6e6bac8: Facebook imports survive an inaccessible ad accountSection titled “Facebook imports survive an inaccessible ad account”

    Previously, the Facebook Marketing connector aborted the whole import as soon as a single ad account returned a permission error. Because accounts are fetched one after another, one account the access token could no longer reach — typically a client that stopped sharing it — discarded the accounts already fetched and every account still queued behind it. The failure repeated on each following run, since the import never got far enough to record its progress, so the data stayed frozen until someone removed that account from the configuration by hand.

    Now, an account that returns a permission error is skipped and the import carries on with the remaining ones, for catalog and time-series data alike. Each skip is raised as a warning on the run, so an account that is quietly missing from the destination is visible without reading the run log. Only permission failures are skipped: a storage write or an exhausted transient error still fails the run, which keeps the import from moving past a day whose data was never stored. If every account fails, the run stops with an error naming each one, because that points to a global cause such as an expired access token.

  • 064557c: Memory-safe Keywords import for Microsoft AdsSection titled “Memory-safe Keywords import for Microsoft Ads”

    Previously, importing Keywords from a large Microsoft Ads account crashed the run with an out-of-memory error. The connector now streams records to storage in small chunks instead of loading the whole download at once. Imports of any account size now complete within normal memory limits.

  • 8976dca: OpenHolidays connector no longer fails when saving fetched dataSection titled “OpenHolidays connector no longer fails when saving fetched data”

    Every OpenHolidays run failed with TypeError: storage.saveData is not a function right after the API returned holiday rows, so no data ever reached the storage table. The connector requested its storage instance without waiting for the async initialization to finish, then tried to save data into the still-pending result instead of the ready storage. It now waits for the storage to initialize before saving, and imports complete successfully.

  • 04f822f: Editing the table name no longer clears the other storage fieldsSection titled “Editing the table name no longer clears the other storage fields”

    On the connector setup step “Choose where to store your data”, clearing the table name also wiped the dataset (or database) name. Typing a dataset name was impossible while the table field stayed empty — each keystroke disappeared.

    The step now stops re-reading its own emitted value once you edit any field. Defaults still fill in on first open, and an existing target still loads when you edit a saved connector. The fix applies to every storage type: Google BigQuery, AWS Athena, Snowflake, AWS Redshift, and Databricks.

  • a4c8c35: Google Sheets reports recover from deleted sheets with one clickSection titled “Google Sheets reports recover from deleted sheets with one click”

    A Google Sheets report breaks when someone deletes its destination sheet, and an import that replaces the spreadsheet’s sheets breaks it the same way. The report then failed every run with a cryptic “Failed to find sheet 0” error.

    Now the error explains what happened and names the fix. A failed report offers a Reconnect & run action in its row menu. One click rebinds the report to a sheet named after it — reusing the sheet if it exists, creating it otherwise — and reruns the report. Renaming the sheet afterwards is safe: the report tracks the sheet by its ID, not its name.

  • aa314fd: Values starting with ”+” no longer break in Google Sheets reportsSection titled “Values starting with ”+” no longer break in Google Sheets reports”

    Google Sheets treats a cell value that starts with + as the beginning of a formula, so exported values like session identifiers of the form +Dri7… used to show up as #ERROR! Formula parse error instead of the data.

    Now the Google Sheets destination prefixes such values with an apostrophe (') — the standard Sheets escape symbol. The cell displays the original value exactly as stored; the apostrophe is not part of the cell content. Only values starting with + are escaped: everything else, including formulas inserted as data (values starting with =), is exported unchanged.

  • e387fee: Typing in the SQL editor now works reliably on saved Data MartsSection titled “Typing in the SQL editor now works reliably on saved Data Marts”

    On any saved Data Mart, the SQL Query editor could refuse to insert spaces: you would type select * from, press the spacebar, and nothing would appear. This happened whenever the Joinable Data Marts section was set to the diagram view — the diagram was silently capturing the spacebar for itself. The editor now always receives everything you type, regardless of how the Joinable Data Marts section is displayed.

    Two more editing annoyances are gone as well:

    • A query you were still writing no longer disappears when the page refreshes its data in the background (for example, right after saving, publishing, or updating the output schema).
    • After picking a different input source type, the Save button no longer stays greyed out while a valid query is on screen.
  • 58d4891: Connecting a Google account no longer fails intermittentlySection titled “Connecting a Google account no longer fails intermittently”

    Authorizing Google BigQuery storage or a Google Sheets destination could fail at random — the popup would show a sign-in screen or an authorization error, and the connection would only succeed after several attempts.

    The cause: the OAuth popup used to boot the entire application and sign in on its own. That sign-in competed with the main tab for the same single-use session credentials, and whichever window lost the race failed. The popup could also end up in a different project context than the tab that started the connection, which made the authorization attempt be rejected.

    Now the popup only hands the authorization result back to the tab where you clicked “Connect with Google”, and that tab completes the connection using its own session. One attempt is all it takes.

  • e387fee: api-client: expose in/not_in filters and the new relative-date presets in the traversal typesSection titled “api-client: expose in/not_in filters and the new relative-date presets in the traversal types”

    Release 0.31.0 taught the Data Mart traversal API the in/not_in operators and the this_week/last_week/this_quarter/last_quarter/next_n_days relative-date presets, but the public @owox/api-client types were not updated: a TypeScript caller writing operator: 'in' got a compile error even though the request itself works.

    • TraverseDataFilterRule gains the in/not_in branch: value is an array of 1..500 same-type values (all strings or all numbers) — never a comma-separated string, and never booleans (use is_true/is_false).
    • TraverseDataRelativeDatePreset gains this_week, last_week, this_quarter, last_quarter, and next_n_days.
    • The @owox/api-client guide now documents every filter operator’s value shape (scalar / array / { from, to } / preset), the in/not_in constraints, the full relative-date preset list, and the NULL-inclusive semantics of negative operators; the OpenAPI description of the filter query parameter shows an in example.
  • ec28b5e: API surface maintenanceSection titled “API surface maintenance”

    • Low-level API client transport methods — OWOX API clients can now use patchJson() and deleteJson() alongside getJson(), postJson(), putJson(), and getStream() for /api/... endpoints that do not yet have typed resource abstractions. Authenticated low-level requests accept only root-relative /api/... paths up to 2,048 characters and refuse unsafe paths and redirects, preventing credentials from being sent to an unintended destination. Existing custom transports remain source-compatible and can add PATCH and DELETE support independently; calling an unsupported new method rejects with OWOXConfigError. The existing plugin protocol adds PATCH and DELETE through ctx.owox as an additive capability while preserving compatibility with existing plugins. Low-level JSON return values are caller-typed and are not runtime-validated; consumers should prefer typed resource abstractions when available.
    • Manage Data Mart run lifecycles through the API client — @owox/api-client now supports starting, listing, inspecting, and cancelling Data Mart runs through the Data-Mart-scoped runs.forDataMart(id) client and its start(), list(), get(), and cancel() methods. Starting and cancelling require Technical User access; listing and inspecting require Business User access. start() accepts typed incremental or manual-backfill options, including connector-specific fields; manual-backfill data is optional for connectors without backfill fields. The client requires an explicit MANUAL_BACKFILL run type whenever data is supplied and rejects data on implicit or explicit incremental runs before sending a request, while the backend HTTP endpoint separately accepts retained object-valued data on incremental requests so existing run forms remain compatible. The client also rejects empty, dot-segment, or separator-bearing Data Mart and run IDs before sending a request. Serialized manual-run options are limited to 1 MiB by both the client and HTTP API; requests above the HTTP transport ceiling return 413 instead of an internal-server error. Packaged authentication middleware now limits its body parsers to /auth, so it no longer overrides the backend API’s 2 MiB transport ceiling.
    • Scoped list pagination — defaults to 100 items when omitted and preserves valid caller-provided limits and offsets without silently capping them. The scoped list method rejects unknown, zero or negative limits, negative offsets, non-integer values, non-finite values, and integers outside JavaScript’s safe range before authentication or network access. Project-wide runs.list() pagination remains unchanged and relies on server normalization. Project-wide and Data-Mart-scoped run methods remain compatible with older self-hosted deployments that omit Data Quality fields, and Data Quality response validation tolerates additive server fields while still checking known values. Existing typed integrations need no migration.
  • 715c6f7: Plugin publishing no longer fails during a safe synchronization cooldownSection titled “Plugin publishing no longer fails during a safe synchronization cooldown”

    Repeated publishing now reuses the plugin’s last validated version while a GitHub synchronization is cooling down. Authenticated GitHub App and server-token reads use a 30-second default cooldown; anonymous reads retain the safer 300-second default. A separate in-progress error is returned while another synchronization is still running.

    • @owox/internal-helpers@0.32.0
    • @owox/idp-protocol@0.32.0
    • @owox/idp-better-auth@0.32.0
    • @owox/idp-owox-better-auth@0.32.0
    • @owox/backend@0.32.0
    • @owox/web@0.32.0

OWOX Data Marts – v0.30.0

  • 61fc1bc: Add Data Quality checksSection titled “Add Data Quality checks”

    Configurable manual and scheduled Data Quality runs for BigQuery, Legacy BigQuery, Athena, Snowflake, Redshift, and Databricks Data Marts. The new Data Quality workspace supports 11 table, field, and relationship checks, stores reproducible SQL and examples, and surfaces compact results in Output Schema, Models, and Run History.

  • e8ae62c: Data Last Updated DateSection titled “Data Last Updated Date”

    You can now see when the source tables behind a Data Mart last changed in the warehouse — in the UI and through MCP.

    • In the UI — the Data Mart page shows a Data Last Updated tile with a Check now button, the Data Marts list gains a sortable Data Last Updated column, and the model canvas shows the value on every node with an Actions → Check Data Last Updated entry that measures everything currently visible in one go. Checking reads warehouse metadata only: it registers no consumption and records no run.
    • In MCP — query_data_mart returns a data_last_updated block alongside the rows and totals, reporting when the source tables behind the result last changed, which tables were checked, and how complete that picture is. It is measured live as part of the same call — never cached, never charged separately — and against the fully composed query, so a blended result covers every joined Data Mart’s tables, not just the primary one. Each MCP query run also records the block in Run History.
    • On every run — every run that delivers data measures the value on the way out: report runs (manual and scheduled), HTTP Data requests, and MCP queries all record it in Run History. When a run reads only the Data Mart’s own sources, the measurement also becomes the Data Mart’s saved value, so the list, canvas, and Data Mart page stay current without anyone pressing a button. A run spanning several joined Data Marts records the measurement in its history entry only, because a blended reading would overstate any single Data Mart.

    Google BigQuery is supported first: TABLE, VIEW, SQL, TABLE_PATTERN, and connector-backed Data Marts all resolve through to their underlying base tables, including nested views and sharded/wildcard table sets. Other storages report unavailable until their support lands.

    Read the timestamp precisely: it says when the source tables were last written to, not which period the data covers — which is why the field is called data last updated rather than freshness, and the tool instructs the assistant to phrase it that way. A null timestamp means unknown, neither fresh nor stale; coverage: "partial" means some sources could not be read, so the real time can only be more recent than reported. Views are excluded from the per-table detail, because a view’s modification time reflects a change to its definition rather than to any data.

  • db60962: Google Sheets source connectorSection titled “Google Sheets source connector”

    A new Google Sheets source connector imports one sheet into a table in the user’s Storage and makes it available as a Data Mart. It supports Service Account authentication, dynamic header-based field inference, and full refreshes across supported storages, so the materialized table reflects the current sheet contents, including schema changes.

  • c6c0d7b: Models canvas redesigned as an ERD, Joinable Data Marts diagram to matchSection titled “Models canvas redesigned as an ERD, Joinable Data Marts diagram to match”

    The Data Marts Models canvas now renders ERD cards instead of plain title boxes. Each card shows a definition-type badge and accent color (SQL / View / Table / Pattern / Connector), a published/draft status dot, and — in the new Detailed view — typed field rows with a primary-key indicator, primary keys first, and hidden-for-reporting fields dimmed. Long field lists collapse to 4 rows with an in-place “+N more fields” toggle. Switch between Compact (default) and Detailed in the canvas settings; the choice is remembered per browser.

    Cards are draggable and their positions persist per storage in your browser across reloads (picking a layout algorithm re-flows from scratch). A new Object labels setting toggles what every card shows — input-source badge, field count, status dot — with “Check all” / “Uncheck all” shortcuts. Relationship lines are smooth bezier curves, neutral gray at rest and brand-blue when selected: click a single line, or click a card to light up all of its connections. The minimap colors nodes by definition type. The canvas renders immediately from the list data and enriches cards with fields as details load, so large models appear without delay.

    The Joinable Data Marts diagram on the Data Mart page uses the same ERD card design, and its gear popover gains two filters: Show looped Data Marts (off by default — self-referencing loop stubs no longer blow up the graph) and a Status filter (All / Published / Draft).

    Thanks to @ikrasovytskyi for contributing the redesign!

  • be313d1: Export the Models canvas as SVG, PNG, JSON, or an OKF bundleSection titled “Export the Models canvas as SVG, PNG, JSON, or an OKF bundle”

    The Models canvas now exports the data model through Actions → Export, in four formats:

    • Image (SVG) — a vector snapshot of the whole visible model, crisp at any zoom.
    • Image (PNG) — the same snapshot rasterized at 2× scale, for chats and tools that do not render SVG.
    • JSON — the model graph (Data Marts, schemas, joins, canvas positions) in the OWOX Model Canvas format, sanitized of project identifiers.
    • OKF (Markdown) — a zip with one cross-linked Markdown document per Data Mart plus an index: an overview, the schema table, and the join list per mart. Works well as context for AI assistants.

    The export covers exactly what the canvas shows — the same filtered set the other Actions target — and captures the whole model regardless of the current pan and zoom. Image backgrounds follow the active theme, so dark-theme exports stay readable outside the app.

    The canvas also got leaner: the Actions menu moved from a canvas overlay into the toolbar above the canvas, so it no longer covers content; the field count now shares a row with the Data Quality and Data Last Updated indicators, making every card one row shorter; and the Object labels settings list dropped its decorative glyphs.

  • 071b429: Change the Input Source of an existing Data MartSection titled “Change the Input Source of an existing Data Mart”

    The Input Source type of a Data Mart can now be changed after setup — a table can become a SQL query, a view can be repointed at a real table, and so on, within the same storage. Previously the type was frozen once saved, so the only way to switch was to rebuild the Data Mart from scratch and recreate everything attached to it.

    The Data Mart keeps its identity through the change: relationships in both directions, reports and their column configuration, and field-level metadata such as aliases, descriptions and aggregation roles all stay in place. Fields missing from the new source are marked as disconnected and fields whose type changed are flagged, the same way they already are when a source of the same type is edited.

    Because a change can disconnect fields that reports and relationships rely on, it is confirmed in a dialog stating what is preserved, what may break, and how many relationships and reports depend on this Data Mart. A published Data Mart stays published. The new source is checked against the storage before the change is saved, so an unreachable table or an invalid query is refused instead of replacing a working definition and failing later. The schema is refreshed automatically right after the change, and refreshing the output schema now reports what it found — how many fields are connected, and how many came back disconnected or with a type mismatch — instead of only confirming that it ran.

    Data Marts based on a connector are not affected: their Input Source type stays fixed, and no other type can be switched to a connector.

  • 66ac952: HTTP Data on report levelSection titled “HTTP Data on report level”

    Adds GET /api/external/http-data/reports/{reportId}.ndjson — stream an existing report’s data as NDJSON, applying the report’s saved filters, aggregations, date buckets, unique count, and sort, without reconstructing query parameters by hand. An optional limit query parameter overrides the report’s saved limit. The @owox/api-client reports.traverseData() method exposes the same endpoint.

    The response carries an x-owox-run-id header (consistent with the Data Mart HTTP Data endpoint); on a successful stream the referenced run is a HTTP_DATA run tagged with the reportId. When the report applies a filter, aggregation, sort, date bucket, row limit, or unique count — or blends fields across Data Marts — the run also records the exact executed SQL via additionalParams.httpData.executionSqlQuery, discoverable through the run history endpoint. (A plain column selection with none of those controls streams the Data Mart’s own query and records no separate executed SQL.) Grand totals — one per selected metric, aggregated over the full result — are recorded in the run history whenever the report selects an eligible metric, computed as a separate best-effort query.

    Failed MCP query_data_mart calls create no run-history entry; both HTTP Data endpoints (Data Mart and report level) record a FAILED run for any failure encountered during query execution.

  • ace4697: Guide AI queries with a field type → operators/aggregations matrixSection titled “Guide AI queries with a field type → operators/aggregations matrix”

    The MCP tools now tell an AI assistant up front which filter operators and aggregation functions each field supports, instead of letting it find out through failed queries:

    • get_data_mart_details_by_id enriches every native and joined field with its type category and the effective allowedAggregations a query may apply to it, and returns an operators_by_category matrix for the filter/slice operators each field-type category accepts.
    • The query_data_mart tool description embeds a field-type matrix generated from the same constants the validator enforces.
    • Operator/type mismatches (for example contains on a number) now return a targeted invalid_operator_for_type error listing the operators the field does accept, instead of a generic failure that sent assistants re-fetching the schema in a loop; date_buckets misuse gets equally specific invalid_date_bucket errors.
    • Boolean fields are now filterable: eq/neq with a boolean true/false value translate to the internal is_true/is_false operators.
    • New in/not_in filter operators (match any of / none of a value list, up to 500 values) are supported natively across BigQuery, Athena, Redshift, Snowflake, and Databricks. They are available in the report filter/slice UI as “is any of” / “is none of” (comma-separated values) for string, number, date, and time columns, and in query_data_mart filters/slices; is_empty/is_not_empty are now also accepted by query_data_mart for string fields.
    • Malformed operands for supported operators (empty in list, bad between shape) now return a precise invalid_filter_value error instead of a generic failure.
    • New relative-date presets on every storage, in the report UI, and in query_data_mart: This week / Last week (ISO weeks, Monday-based on all storages — BigQuery uses ISOWEEK, Snowflake computes the Monday independently of the session WEEK_START), This quarter / Last quarter, and Next N days (includes today, mirroring Last N days). With these, every operator advertised by query_data_mart is now executable — nothing maps to unsupported_operator anymore.
  • edb1a40: List draft Data Marts through MCPSection titled “List draft Data Marts through MCP”

    MCP clients can now explicitly list draft Data Marts with list_data_marts, while published Data Marts remain the default catalog state.

  • be2e29f: Combined and Sample are now on by default for text fieldsSection titled “Combined and Sample are now on by default for text fields”

    Text columns now offer Combined (STRING_AGG) and Sample (ANY_VALUE) out of the box, alongside Count and Count Unique. Previously both were supported but switched off, so you had to enable them field by field before a report could use them.

    Existing Data Marts pick this up automatically — a field only keeps a narrower list if you explicitly chose one for it. Numeric, date, time, boolean, and complex-type fields are unchanged, and neither function is ever included in Totals.

  • 5e9b076: Sort by Unique CountSection titled “Sort by Unique Count”

    Reports can now sort by the Unique Count metric. Previously, enabling Unique Count added the column to a report but the Sort section had no way to reference it. Once Unique Count is enabled, it now shows up as a sort option like any other column, ascending or descending.

  • 04bf74d: Joined field names read better in Google SheetsSection titled “Joined field names read better in Google Sheets”

    Columns from a joined Data Mart used to lead with the Data Mart name — RFM_SEGMENT Recency Score — pushing the part you actually look for out of view in every header cell. Google Sheets now writes the Data Mart name after the field name, in parentheses: Recency Score (RFM_SEGMENT). For a chain of joins the name is the Data Mart the field actually comes from, not the whole path. Fields from the report’s own Data Mart are unchanged, and a blank Output Alias still produces a bare field name.

    Existing sheets pick the new headers up on their next refresh. Row 1 is rewritten in place — no column is added, removed or reordered, and nothing shifts under your own content to the right of the imported range.

    Data Studio, email-based destinations, the HTTP data endpoint and MCP field metadata keep the Data Mart name as a prefix — the position follows the surface that renders the label, not the report, so reading a Google Sheets report through the HTTP data endpoint still returns the prefixed form. Match columns on the technical field name, which is identical everywhere.

    Labels are also no longer padded: an Output Alias or field alias typed with a leading or trailing space used to leak that space into the column name, and a whitespace-only alias produced a column that looked unnamed. And a Google Sheets report that would write two columns under the same header now fails with a clear message naming the header, instead of writing the duplicate silently — previously the refresh after it blanked one column, inserted a spare, and pushed your own content further right. Rename one of the two columns, or the Output Alias of the joined Data Mart one of them comes from.

  • ce982a5: New Data Marts are shared for maintenance by defaultSection titled “New Data Marts are shared for maintenance by default”

    A newly created Data Mart now starts with both sharing toggles on: Shared for reporting and Shared for maintenance. Other Technical Users can join it to their own Data Marts straight away, instead of first having to find the owner and ask them to share it.

    Data Marts that already exist are not changed. The owner can still turn Shared for maintenance off on any Data Mart in one click — non-owners then keep reporting access and lose the ability to edit or join it. Storages and Destinations keep their current defaults: shared for use, not shared for maintenance.

  • 9e0434a: Cleaner connector run logs: warnings for issues you can fix yourselfSection titled “Cleaner connector run logs: warnings for issues you can fix yourself”

    Failures now split into warnings and errors, so you can tell at a glance what needs you to act. Warnings mark issues you can fix yourself — like expired credentials or a canceled run. Errors mark issues our support team can help you resolve. Logs are also easier to scan: each crash appears as a single entry instead of several, and duplicate entries are gone.

  • 562bbe2: Clearer Data Warehouse selection during onboardingSection titled “Clearer Data Warehouse selection during onboarding”

    The Data Warehouse step in onboarding now shows an interactive preview for the selected supported warehouse, illustrating how OWOX takes data from input sources, processes it in the chosen warehouse, and delivers it to destinations such as Google Sheets, Excel, email, Slack, and Looker Studio. The step also gains a responsive two-column layout, updated selection states, and a short animation. The preview is hidden when you select Other or I don’t know.

  • c97264b: Easier access to service account detailsSection titled “Easier access to service account details”

    When viewing Google service account credentials, you can now copy the service account email with one click or open its details in Google Cloud Console. The email remains visible and can still be selected manually when you only need part of it.

  • 970bcee: Correct COUNT DISTINCT, SUM, AVG, percentiles, and Combined across joined Data Marts (fan-out fix)Section titled “Correct COUNT DISTINCT, SUM, AVG, percentiles, and Combined across joined Data Marts (fan-out fix)”

    COUNT DISTINCT, SUM, AVG, the percentile metrics (P25, Median, P75, P95), and Combined on a joined Data Mart are now computed correctly at the report grain, including across multi-hop bridges (a joined Data Mart reached through another joined Data Mart). Previously each was pre-aggregated on the join key and then re-aggregated, which over- or under-counted (COUNT DISTINCT), over-summed (SUM), produced an unweighted average-of-averages (AVG), weighted a percentile by how many times the join repeated each value, or repeated the same text in Combined once per matching row, whenever the same entity was reachable through more than one path. Each metric now uses a dedicated “sleeve” pass that carries the raw identity to the report grain. These are symmetric (non-additive) aggregates: an entity reachable under two dimension values is counted once in each group but once overall, so a column’s per-group values need not sum to its Totals value — that is correct, not a discrepancy. Applies to BigQuery, Athena, Redshift, Snowflake, and Databricks. Numbers will change for any saved report using one of these metrics on a joined Data Mart whose join fans out.

    • Deduplication key. For SUM, AVG, and the percentiles, the value is de-duplicated by the joined field’s pre-join group key, or — when the field is a raw passthrough (pre-join ANY_VALUE) — by the joined Data Mart’s own declared primary key. Two rows agreeing on that key and on the value are the same row the join returned twice and are counted once; two rows agreeing on the key but not on the value stay separate, so contradictory data is never silently resolved to one version. A joined Data Mart with no declared primary key falls back to a per-row surrogate, which cannot tell a genuine duplicate from two distinct rows. A declared key is trusted: a key that is not in fact unique will merge rows — and so will one whose column is empty on more than one row. Declare a key only on a column that is genuinely unique and always filled.
    • MIN and MAX on a joined Data Mart are computed the same way, over the same rows. They used to read the value the pre-join roll-up had already collapsed, which measured a different set from AVG on the very same column and could leave MIN above the average. COUNT is deliberately unchanged: it counts the rows that survive the join, which is a different question rather than a distorted one. The main-side Unique Count (by primary key) is unchanged.
    • COUNT and COUNT DISTINCT are no longer shown side by side on a JOINED field. They are computed at different grains, so together they can invert COUNT DISTINCT <= COUNT. A report that selects both on a joined field now shows only COUNT DISTINCT — the report is not rejected, and COUNT on its own keeps working exactly as before. Totals stop offering the pair for the same reason. A non-joined field is unaffected.
    • Deterministic “any value” roll-up. A joined field whose pre-join roll-up is “any value” is now rolled up deterministically. That roll-up is read twice — once for the report’s grouping and again inside the fan-out correction — and an engine free to evaluate the two separately could return different values, which made the correction silently produce empty results. Applies to text, number, date and time fields; other types are unchanged.
    • Joined COUNT DISTINCT on a pre-rolled-up field (e.g. a text field pre-aggregated with STRING_AGG before the join) now counts distinct raw values at the report grain, not distinct rolled-up values. This is more correct — the previous behavior could conflate different raw values that rolled up into the same combined result — but the number can change for a saved report using such a field.
    • Known limitation: a metric (HAVING) filter on any of these joined metrics — COUNT DISTINCT, SUM, AVG, MIN, MAX, Combined, or a percentile — is not yet supported and is rejected with a clear error; filter on a main-side column or a different metric instead. A previously-saved report carrying such a filter will be rejected on its next run.
    • Totals now honour metric (HAVING) filters. A Totals row has no grouping, so a per-group constraint used to be dropped, and Totals then summarised rows the report itself hides. Totals are now computed over the rows of the groups that pass the filter, and stay symmetric while doing so. Totals will change for any saved report that carries a metric filter.
    • Totals that cannot be computed are now reported instead of silently omitted. They are still best-effort — a failure never costs the query its rows — but an empty Totals block used to be indistinguishable from a report with no totals at all, which invites summing the returned rows instead (wrong for any non-additive metric). The reason travels with the response and is recorded in Run History; AI callers are told only that totals failed, since a warehouse message quotes the query and its table and column names.
    • Incomplete or unusable relationships are rejected with an explanation instead of failing at the warehouse. That covers a relationship saved with no join conditions (a supported draft state — it is refused when a report tries to use it, not when you save it), and one joining on a nested field such as address.city, which the query engine cannot express and which was previously accepted and then failed with a raw database error. Editing a relationship no longer requires filling in the join conditions, matching what creating one already allowed.
    • Redshift identifier collisions are caught at save time. Redshift lower-cases and truncates long identifiers, so two report columns differing only in case, or only past 127 bytes, arrive as one. Such a pair is now rejected when the report is saved, on every warehouse. An already-saved report carrying such a pair is rejected on its next run — on Redshift it was returning data from the wrong column.
    • Reserved internal names. A joined Data Mart with a source column named __owox_rid, or a report that groups by a field (or output alias) named _oid/_oid_<n>, _oid_key_<n>, _val/_val_<n>, or _dedup alongside a joined SUM/AVG/percentile metric, is now rejected with an error naming the column instead of silently returning a corrupted value — rename the field or its alias.
    • Rejected requests now say what was actually wrong. Every failed validation rule — across the whole API, not only reports — reached the caller as the single word “Bad Request Exception”, with each per-field reason dropped before the response left the server. Each reason now travels with the response and is shown in the interface. A server error also shows a message identifying the request instead of failing silently.
  • 2680781: Keep NULL rows for “is not” / “is none of” / “does not contain” report filtersSection titled “Keep NULL rows for “is not” / “is none of” / “does not contain” report filters”

    Negative filter operators (is not, is none of, does not contain, does not match regex) no longer drop rows where the filtered column is NULL. SQL three-valued logic previously excluded those rows; filters and slices now keep them so missing values behave like “not equal to X”. Add is not null or is not empty when you want to exclude NULLs. Applies to BigQuery, Athena, Redshift, Snowflake, and Databricks.

    In blended reports this also changes post-join negative filters on LEFT JOINed columns: a filter like crm.status is not Churned now keeps source rows with no join match at all, where it previously acted as an implicit inner join. Add an is not null filter on the joined column to drop unmatched rows. Metric (HAVING) filters share the same logic, so a negative metric filter such as SUM(amount) ≠ 0 now also keeps groups whose aggregate is NULL (e.g. all-NULL input).

  • 122ce35: More reliable connector runs during cancellation and long backfillsSection titled “More reliable connector runs during cancellation and long backfills”

    Previously, cancelling a connector run could fail to stick: the run reappeared as running minutes later, and its logs were lost. Long BigQuery backfills on OAuth also stopped after about an hour with an “Invalid Credentials” error. Now cancelled runs stay cancelled and keep their logs, and BigQuery refreshes its access token mid-run so long backfills continue.

    A run can still fail with “Invalid Credentials” if its saved token expired before the run started. Reconnect the BigQuery destination, or use a service account, until token saving arrives.

  • 8724c41: Fix dropped characters and lost focus in connector settingsSection titled “Fix dropped characters and lost focus in connector settings”

    Previously, typing into a field on the “Configure Settings” step dropped characters, so text seemed to appear only on the second try, and pasted values could revert. Credential fields also lost focus after the first character, forcing users to click back into the field. Both problems are fixed — fields now keep focus and accept every character, whether typed quickly or pasted.

    Nested credential fields now also show their readable labels everywhere. Some tabs previously showed the raw field name, such as “ServiceAccountKey” instead of “Service Account Key (JSON)”.

  • e02ff15: Preview SQL no longer requires edit accessSection titled “Preview SQL no longer requires edit access”

    Opening Preview SQL on a report used to fail with “edit access to the source data mart is required” for anyone without maintenance access — including Business Users on a Data Mart shared for reporting, and Technical Users who are not owners of it. Reading the query a report runs is now tied to seeing the report: if a Data Mart is visible to you, you can open the generated SQL of any report on it and copy it to the clipboard.

    The two actions in that dialog that genuinely need maintenance access — the SQL validator (dry run) and Copy as Data Mart — are now hidden for users without it instead of failing on click. Reports that join a Data Mart the viewer cannot access still refuse to render SQL, naming the inaccessible Data Mart.

  • e5d26a0: No more leftover columns when switching Data Marts in Google SheetsSection titled “No more leftover columns when switching Data Marts in Google Sheets”

    When creating a report in the Google Sheets extension, switching to another Data Mart now resets the column selection. Previously, columns picked for the first Data Mart stuck around and showed up as “Disconnected columns” under the new one — and could even reach the created report and make it fail.

  • a54bf5d: API surface maintenanceSection titled “API surface maintenance”

    • Project search API — GET /api/search now publishes the integer range and comma-separated serialization of its optional filters in OpenAPI. @owox/api-client adds search.query(query, options) and exports OWOXSearchResult, OWOXSearchEntityType, and OWOXSearchOptions for discovering visible Data Marts, data storages, and data destinations. Existing viewer access and search behavior are unchanged.
    • Project run history contract reconciled — GET /api/data-marts/runs now publishes the complete project-wide run-history contract, including viewer visibility, pagination normalization, enums, field presence and nullability, and the backend’s RFC3339 timestamp profile. createdByUser is the nullable run-author field; when present it includes userId and may include nullable fullName, email, and avatar. definitionRun remains present but can be null when a historical definition snapshot is unavailable. @owox/api-client validates this contract and exposes it as runs.list({ limit, offset }) — consumers using the previously released runs.getHistory(...) must rename those calls to runs.list(...); the response and option type exports remain available.
    • HTTP Data streaming contracts strengthened — GET /api/external/http-data/data-marts/{dataMartId}.ndjson now publishes its exact-column projection, bounded base64url controls, positive-integer limit, NDJSON response, run identifier, and failure contract in OpenAPI. @owox/api-client provides typed filter, sort, aggregation, and date-bucket controls for dataMarts.traverseData(...) and validates the NDJSON response media type before traversal. Consumers passing controls through unknown[] or widened variables must adopt the exported rule types or annotate their options with TraverseDataOptions; valid inline calls remain unchanged.
    • Data Mart list contract actualized — GET /api/data-marts now publishes viewer visibility, non-negative integer offset validation, owner-presence filtering, 1,000-item pages, and the complete nested list-item response contract, including nullable draft definition types and optional nullable user metadata strings. @owox/api-client validates every returned page and exposes the full OWOXDataMart shape; dataMarts.list({ offset, ownerFilter }) can start from an offset, filter by has_owners or no_owners, and follows subsequent pages automatically. The package exports OWOXDataMartListOptions, OWOXDataMartOwnerFilter, and the nested Data Mart enum and object types, and rejects invalid list options before sending a request. Existing dataMarts.list() calls remain compatible.
  • @owox/internal-helpers@0.31.0
  • @owox/idp-protocol@0.31.0
  • @owox/idp-better-auth@0.31.0
  • @owox/idp-owox-better-auth@0.31.0
  • @owox/backend@0.31.0
  • @owox/web@0.31.0
  • 86425d3: More reliable MCP reporting answersSection titled “More reliable MCP reporting answers”

    MCP reporting tools now expose only published Data Marts and use clearer, business-friendly field labels in query results. Each result identifies its source Data Mart and includes OWOX links for the related Data Marts, reports, destinations, and schedules.

    When a query result is incomplete because of the row limit or response-size cap, the assistant receives an explicit truncation reason and instructions to make that limitation clear. It also receives guidance to use OWOX aggregations for count questions and to distinguish OWOX-calculated values from calculations made in the assistant response.

OWOX Data Marts – v0.30.0

  • cb6bc43: Model canvasSection titled “Model canvas”

    New project-level, read-only visualization of the data marts visible to the user in a storage and their relationships, available from the new “Models” sidebar item under Data Marts. Includes filters by storage, status, and relationships; search with highlighting; join-field labels on edges; node-avoiding edge routing; selectable layout directions; and directional (including bidirectional) join edges. Both this canvas and the existing Joinable Data Marts graph are now rendered with React Flow (replacing Rete.js), preserving their established styling and interactions.

  • f3a83b3: Add project-based MCP server URLsSection titled “Add project-based MCP server URLs”

    OWOX MCP now supports project-specific server URLs in the form https://{projectId}.mcp.owox.com/mcp alongside the existing shared https://mcp.owox.com/mcp server. Shared connections keep the existing project-selection flow. Project-specific connections derive the project from the MCP host, skip project selection, and still verify that the authenticated user has access to that project before issuing MCP tokens. Project settings show the project-specific MCP URL with copy support.

  • 480e8bc: Add project-aware MCP instructionsSection titled “Add project-aware MCP instructions”

    OWOX MCP servers now provide clients with built-in instructions for discovering and querying Data Marts with the current tool workflow. Project admins can add project-specific business context in Project settings; the project description is stored per project and appended to the MCP instructions when a client connects.

  • 3a5abb3: Create and update reports for every destination type through MCPSection titled “Create and update reports for every destination type through MCP”

    The add_report tool now supports all destination types, not just Google Sheets. Data Studio reports are created with default settings (a 5-minute data cache lifetime), and the response includes a setup-guide link — now also part of the add_destination response for looker_studio. Email, Slack, Microsoft Teams, and Google Chat reports take a new message parameter with the subject (defaults to the report name) and body template (supports the {{table}} placeholder); recipients stay on the destination and the send condition defaults to “Send always”. update_report accepts the same message parameter to change the subject and/or body of an existing email-family report, preserving everything else including the send condition. Every add_report response now reports the destination_type it created.

  • ba53b3a: Create and update reports with query-parity output controls through MCPSection titled “Create and update reports with query-parity output controls through MCP”

    The add_report tool accepts new optional filters, slices, aggregations, date_buckets, sort, and limit parameters — the same shape and vocabulary as query_data_mart — applied to the report on every run, so a filtered or aggregated query can be exported to a report that matches exactly the numbers the user saw. update_report accepts the same parameters as replacements ([] removes a control, null removes the row limit, omitting a parameter keeps the current value; filters and slices each replace only their own kind of rule). All controls work for every destination type and are validated against the data mart schema before any side effect — for Google Sheets, before the sheet document is created.

  • 52b3245: Sort rows from the query_data_mart MCP toolSection titled “Sort rows from the query_data_mart MCP tool”

    query_data_mart now accepts a sort parameter — an ordered list of { field, direction } rules (asc/desc, first rule primary) — so an assistant can order results (e.g. “top 10 campaigns by spend, descending”) without post-processing. Each sorted field must also be listed in fields, matching the existing aggregations and date_buckets rules. This brings the MCP tool in line with sorting already available in the Reports, HTTP Data streaming, and CLI query paths.

  • 2740556: Improve Output Schema AI actionsSection titled “Improve Output Schema AI actions”

    The Refresh schema and global Generate field aliases & descriptions actions moved to the table toolbar, keeping all primary actions in one place. The global AI action dropdown was removed — a single click now generates both field aliases and descriptions. AI action buttons were added to the Alias and Description column headers, so users can generate values for an entire column where the action applies.

  • 091d502: Data Level selection for TikTok Ads performance reportsSection titled “Data Level selection for TikTok Ads performance reports”

    TikTok Ads ad_insights and ad_insights_by_country now honor the selected Data Level (advertiser, campaign, ad group, or ad), a dropdown in the main connector settings. The field selector pins only the unique-key fields that level requires — advertiser-level reports no longer force ad_id and can group by date alone. Choose Data Level before selecting fields; when changing it on a connector that already has data, use a new Data Mart so rows merge correctly.

  • 55fcba9: Reports table UX — discoverable actionsSection titled “Reports table UX — discoverable actions”

    • Moved Open document and Preview SQL next to the report title.
    • Kept the More actions menu in the actions column.
    • Added tooltips for truncated report titles.
  • b0f0247: Shared Login Customer ID field for Google AdsSection titled “Shared Login Customer ID field for Google Ads”

    The Login Customer ID is now a single top-level field shared across all authentication types, instead of a separate field nested inside each auth type. Previously the value could be lost when switching between authentication methods and was not reachable from the OAuth button flow. Existing connectors keep working — the old stored value is still read as a fallback. The field is now optional for Service Account authentication and appears directly under Customer ID in the configuration form.

  • 18c8d9d: Selective refresh and clearer failed-report errors in Google SheetsSection titled “Selective refresh and clearer failed-report errors in Google Sheets”

    • On Refresh all reports, each report shows its title, sheet, and last-run status, including a clear error message when the last run failed.
    • Choose which reports to refresh with checkboxes, select all, or target only failed, successful, or never-run reports.
    • On the single-report refresh screen, failed runs show the actual error details and an Open report action.
    • The same last-run error details are available from the All reports list.
  • 7d09e62: HTTP Data streaming supports aggregations and date bucketsSection titled “HTTP Data streaming supports aggregations and date buckets”

    The HTTP Data streaming API, the @owox/api-client traverseData() method, and the owox-ctl data-marts stream command now accept aggregation and dateTrunc parameters (--aggregation / --date-bucket on the CLI), matching the grouping already available for Reports. Any selected column without an aggregation rule becomes a grouping key, so you can stream pre-aggregated rows — for example monthly revenue totals — directly from a published Data Mart without building a Report. Grand totals and the applied aggregation are recorded in the HTTP_DATA run history.

  • 0781310: Clearer permission errors when managing Storages, Destinations, and Data MartsSection titled “Clearer permission errors when managing Storages, Destinations, and Data Marts”

    Permission errors now name the exact role required — such as owner with the Technical User role, or Project Admin — and the “access forbidden” toast stays on screen until dismissed. Previously these actions failed with a vague message that did not explain who could perform them, and the error toast disappeared after a few seconds.

  • 5595bbd: Google Sheets A1 note shows import finish timeSection titled “Google Sheets A1 note shows import finish time”

    The A1 cell note for a Google Sheets export now records when the import finished (written in the finalize step after all data is on the sheet), not when the first data batch started writing. While rows are still streaming, only the short ODM marker is present; the full provenance block (Imported at …, data mart title, and link) is written at the end.

  • 1fc1bd2: Clearer report details, errors, and recovery in Google SheetsSection titled “Clearer report details, errors, and recovery in Google Sheets”

    The Data Mart section on a report card no longer shows “Details are unavailable” while loading — you now see a loading indicator, a clear message if details fail to load, or “No details available” when there’s no description or owners. When a report cannot run because columns are missing from the Data Mart schema, the extension shows the same actionable error as OWOX Data Marts (including which columns to uncheck). If a report is created but its first run fails, the editor still opens so you can fix the setup and retry. Valid report slices stay in sync with OWOX Data Marts and no longer appear broken.

  • 33cb1ec: Fix BigQuery report runs when the table name matches a filter columnSection titled “Fix BigQuery report runs when the table name matches a filter column”

    BigQuery report runs failed when a Table Data Mart’s unaliased table short name matched a filtered column — BigQuery treats the short name as a row STRUCT alias, so the generated SQL compared a STRUCT to a string (for example on Google Sheets export for a table such as …country filtered on a country column). Output-controls and explicit-projection queries now alias the source as src and qualify filter/sort references (src.`country`) in WHERE, ORDER BY, and HAVING, while SELECT and GROUP BY stay unqualified so nested RECORD paths keep their shape. Existing reports need no reconfiguration.

  • b1fc8ff: Output schema edits no longer get silently revertedSection titled “Output schema edits no longer get silently reverted”

    Unchecking a column (Alias or Description) no longer becomes visible again after a few seconds or resets on tab switch, and toggling “Hidden from reports” on a field while an AI-generated alias or description was still loading is no longer undone once generation finishes. Column visibility now persists per data mart across tab switches and reloads, and edits made while AI generation is running are preserved. Generating metadata that fills in nothing new no longer marks the schema as unsaved.

  • 97f0c0f: Correct joined data mart aggregations and flag missing primary keysSection titled “Correct joined data mart aggregations and flag missing primary keys”

    When a joinable field is deduplicated with a count (COUNT / Count Unique), its value in the blended result is now a whole number, so reports can sum, average, or take min/max of it (arithmetic aggregations appear automatically, Sum active by default) instead of collapsing the joined events into a concatenated string and under-counting them — fixing funnel-style reports (sessions, add-to-carts, purchases across shared dimensions). Slices (pre-join filters) on such a field now offer the operators for its original type, since a slice runs on raw values before the join. Separately, a joined data mart with no primary key now shows a “No primary key” warning in both the relationship list and the graph, because without a key the join cannot deduplicate rows reliably and metrics can be double-counted (fan-out).

  • 345d78e: Report totals cover every selected metric, not just numeric fieldsSection titled “Report totals cover every selected metric, not just numeric fields”

    Report totals — returned by the query_data_mart MCP tool and retrievable for a Data Mart run via its x-owox-run-id — now cover every aggregated metric, including Count and Count Unique on text, date, or boolean columns across both native and joined Data Mart fields. Previously totals covered numeric fields only, so a scorecard aggregating a text column (for example the number of distinct countries) came back empty even though the run succeeded. Sample (ANY_VALUE) and Combined (STRING_AGG) are no longer included in totals, since neither reduces to a meaningful grand total.

  • 873c74f: Report column picker fixesSection titled “Report column picker fixes”

    Adding a column in the Aggregations panel now opens the aggregation editor immediately (instead of an icon that required a second click), and its Apply button stays disabled until you pick a function or bucket, so a new column can no longer be discarded by an empty Apply. The auto-generated “Unique count” row now uses the same font as the other fields. The relative-date “Last N days/months” filter input can be cleared with Backspace instead of snapping back to 0. When editing a single filter or slice, a trash button in the editor header now lets you delete it directly.

  • be8a495: Cross-platform Windows support for dev scripts and git hooksSection titled “Cross-platform Windows support for dev scripts and git hooks”

    Two Unix-only assumptions broke first-time setup on Windows. Inline VAR=value command environment-variable assignments (e.g. npm run dev -w owox and the package test scripts) are now wrapped with cross-env so they run under cmd/PowerShell (no new dependency — cross-env is already a hoisted root devDependency — and no lockfile change). The husky setup no longer emits a Windows batch-style pre-commit hook (@echo off …) that failed with @echo: command not found under sh (which Git uses on every OS) and blocked commits; it now writes a POSIX shell hook in the husky v9 format, self-heals only the stale formats it previously generated (leaving hand-edited hooks untouched), and clears the husky - DEPRECATED warning that breaks under husky v10.

  • 72f3d47: API surface maintenanceSection titled “API surface maintenance”

    • Project settings API — the endpoints now publish explicit OpenAPI request/response contracts; @owox/api-client adds project.getSettings() and project.updateDescription().
    • Models canvas API — @owox/api-client exposes paginated Models canvas data marts via models.getDataMarts() and their visible relationship edges via models.getEdges().
    • Project setup progress API — GET /api/project-setup-progress publishes an explicit OpenAPI response contract; @owox/api-client adds project.getSetupProgress() and exports OWOXProjectSetupProgress, OWOXProjectSetupProgressSteps, and OWOXProjectSetupStepState.
    • Project run history API — @owox/api-client adds runs.getHistory({ limit, offset }) and exports OWOXProjectDataMartRunsResponse, OWOXProjectDataMartRun, OWOXProjectDataMartRunRef, OWOXProjectDataMartRunUser, OWOXProjectDataMartRunStatus, OWOXProjectDataMartRunType, OWOXProjectDataMartRunTriggerType, and OWOXProjectRunHistoryOptions.
    • Project insight-template discovery API — @owox/api-client adds insights.getTemplates({ limit, offset }) and exports OWOXProjectInsightTemplatesResponse, OWOXProjectInsightTemplate, OWOXProjectInsightTemplateDataMartRef, OWOXProjectInsightTemplateUser, and OWOXProjectInsightTemplateListOptions; each result includes its Data Mart reference, creator metadata when available, and the current member’s canDelete state.
    • Markdown rendering API — POST /api/markdown/parse-to-html publishes an explicit OpenAPI contract; @owox/api-client adds markdown.parseToHtml({ markdown }) and exports OWOXMarkdownParseRequest and OWOXMarkdownParseResponse.
    • Auth context introspection — GET /api/auth/context publishes an explicit OpenAPI response contract; @owox/api-client exposes auth.getContext() and OWOXAuthContext for validating a configured API key and reading its project/member context without exposing the secret.
    • Require interactive authentication for OAuth flows — OAuth routes under /api/connectors/{connectorName}/oauth, /api/data-destinations/oauth, /api/data-destinations/{id}/oauth, /api/data-storages/oauth, and /api/data-storages/{id}/oauth, plus POST /api/data-destinations/connect/google-sheets, now reject API-key authentication and require an interactive user session. API-key access to non-OAuth resource operations is unchanged.
    • Publish canonical API-key authentication headers — protected operations now declare X-OWOX-Authorization and, only when API-key-derived tokens are accepted, the optional X-OWOX-Api-Key-Id header (must match the token’s API key ID). Routes that reject API-key authentication omit that conditional header, while POST /api/auth/api-keys/exchange retains its required API key ID input. Runtime authentication behavior is unchanged.
  • 52149e4: Fix nested blended field name collisions in joinable data martsSection titled “Fix nested blended field name collisions in joinable data marts”

    Nested (struct) fields in joined sources no longer share report column names with flat siblings after dots are replaced with underscores. For example, flat campaign_id and nested campaign.id under the same join alias now produce distinct unified names (ads__campaign_id vs ads__campaign_id__a8702665).

    • Flat blended names stay byte-for-byte unchanged; existing reports that only use flat joined columns need no config changes.
    • Nested blended names always include a stable 8-character hash of (aliasPath, originalFieldName). New saves and reloads use the hashed form end-to-end.
    • Legacy nested names in already-saved columnConfig / filter / sort (pre-hash form such as customers__campaign_id for campaign.id) are intentionally not migrated in this release and will not resolve after deploy. That population was empty when this shipped; a follow-up migration can re-key them if needed. The pre-join filter migration (MigratePreJoinFilterToUnifiedColumn) still emits the pre-hash form for historical rules only.
    • Rolling deploy: during a mixed old/new backend fleet, nested-field reports saved or run mid-rollout can briefly orphan until every pod serves the hashed names. Self-heals once the fleet is fully on the new version; prefer rolling through before bulk-editing nested-column reports.
    • @owox/internal-helpers@0.30.0
    • @owox/idp-protocol@0.30.0
    • @owox/idp-better-auth@0.30.0
    • @owox/idp-owox-better-auth@0.30.0
    • @owox/backend@0.30.0
    • @owox/web@0.30.0

OWOX Data Marts – v0.29.0

  • 36fb4d3: Data Mart report aggregationsSection titled “Data Mart report aggregations”

    Reports can now aggregate Data Mart data server-side across all supported storages (BigQuery, Athena, Snowflake, Redshift, Databricks).

    • Group-by + metric functions — SUM, AVG, MIN, MAX, COUNT, COUNT_DISTINCT, STRING_AGG, and percentiles P25/P50/P75/P95. Every selected column without an aggregation rule becomes a grouping key.
    • Date bucketing (dateTruncConfig) — group a date/timestamp dimension by DAY/WEEK/MONTH/QUARTER/YEAR, with an optional per-rule IANA timeZone applied before truncation.
    • Aggregated column naming — outputs are named <column> | <TOKEN> (e.g. revenue | SUM, customer_id | COUNTUNIQUE) and carry the function’s effective type. A column can carry several functions, each as its own output column.
    • Auto Row Count — aggregated reports automatically include a Row Count (COUNT(*)) column.
    • Unique Count (uniqueCountConfig) — opt-in COUNT(DISTINCT <primary key>), with composite primary key support; rejected at save time if the Data Mart has no primary key.
    • Post-aggregation filtering (HAVING) — a filter on an aggregated output column is auto-routed to HAVING instead of row-level WHERE; the (column, function) pair must match a configured aggregation.
    • Joined Data Marts — post-join aggregation over the joined result, in addition to the existing pre-join join-rollup.
    • Totals — a per-column summary over the full filtered dataset (no grouping), computed as a separate query and surfaced through the HTTP Data API’s run subtree. Push destinations (Google Sheets, Email/Slack/Teams/Chat) don’t compute totals; Row Count and Unique Count are excluded from totals.
    • Governance — each schema field carries a dimension/metric role, a type-derived set of supported aggregations, and an on-by-default subset, with per-field override.
    • UI — a new AGG control next to output controls, plus a per-field AGG icon, configure grouping, multi-aggregation, date bucketing, and timezone.

    Backend adds three additive nullable report columns (aggregationConfig, dateTruncConfig, uniqueCountConfig) plus save-time validation. Deployment ordering: deploy the backend before shipping a Google Sheets Extension build that exposes new aggregation options, since the extension’s function list has no compile-time guard against drift from the API.

  • 627880f: Auto-create Google Sheets documents for reportsSection titled “Auto-create Google Sheets documents for reports”

    A “Create document” action in the Google Sheets report form auto-creates a new spreadsheet for the selected destination and fills in its link. Works for both auth methods — OAuth (document created in the connected account’s Drive and shared with the requesting user) and Service Account (created in a configured Shared Drive folder).

    • Destinations can target a Drive folder: paste a folder URL for Service Account destinations (validated on save — must be a Shared Drive folder the service account can write to), or pick one via the Google Drive Picker for OAuth destinations.
    • The Picker requires a new GOOGLE_PICKER_API_KEY environment variable.
    • The OAuth destination flow now also requests the drive.file scope — existing OAuth destinations must reconnect to grant it before folder placement and sharing work.
  • e575e99: Smart search for Data Marts, storages, and destinationsSection titled “Smart search for Data Marts, storages, and destinations”

    Added semantic search functionality for finding project Data Marts, Storages, and Destinations from the project search index.

  • e8b87e3: MCP toolsSection titled “MCP tools”

    • e8b87e3: Added summarize_data_catalog — high-level summary of the project’s published Data Mart catalog: total count, top Data Marts ranked by relationship connectivity, report/trigger usage, and recent updates. Returns only Data Marts visible to the current member; does not query row data.
    • e8b87e3: Added get_data_mart_details_by_id — a Data Mart’s id, name, description, and output schema fields, including joined_fields (blended fields with qualified names, source Data Mart, and allowed aggregations).
    • f680592: Added query_data_mart — query a Data Mart directly: pick fields, filter (pre-join slices and post-join filters), aggregate (SUM, COUNT, COUNT_DISTINCT, AVG, MIN, MAX, percentiles), bucket dates by day/week/month/quarter/year, and get server-side totals. Runs against the warehouse, recorded in Run History (query definition and SQL only, never row values), and costs credits per call.
    • 7e3ce65: Added get_data_mart_reports — list reports tied to a Data Mart, including destination, owner, all run schedules (cron, timezone, active flag), last run time, and last run status.
    • c1485c0: Added add_report — create a report exporting a Data Mart to a Google Sheets destination; a new sheet is created and linked automatically. Currently supports Google Sheets only.
    • 4fe2c6e: Added update_report — rename a report or change its exported fields; anything not mentioned (destination, filters, sorting, owners, schedules) stays unchanged.
    • 5ed5a23: Added run_report and get_report_run_status — start a report run by id (returns immediately with report id and run id; data delivers to the report’s push destination), then poll status (running, success, failed, cancelled, interrupted, restricted) with queued_at, started_at, and raw_status. Status responses include should_poll and stop_reason guidance, recommending ~15-second polling intervals. Pull-based destinations (Data Studio, HTTP Data API) can’t be run this way. Starting a run requires mcp:write; checking status only requires mcp:read.
    • b8732c5: Added delete_report — marked destructive so assistants confirm before calling. The underlying Data Mart, destination, and already-exported documents are unaffected.
    • e8499d4: Added report run schedule management — four tools to list, create, update, and delete recurring report run schedules by trigger id, respecting project and report permissions. Creating, updating, or deleting requires both mcp:read and mcp:write.
    • 5f7dbcf: Added list_destinations — list available destinations with name, type, owner, and shared-for-use status.
    • 9f2ca93: Added add_destination — create a new destination. Email-based destinations (email, Slack, Teams, Google Chat) and Looker Studio are created and ready immediately; Google Sheets destinations require the user to connect their Google account via a linked in-app page. Every destination created this way starts unshared until a human reviews and shares it.
  • 964d886: Protect unsaved Output Schema changes from being lostSection titled “Protect unsaved Output Schema changes from being lost”

    When a Data Mart’s Output Schema has unsaved changes, a prompt now appears before any action that would replace it, with three choices: Save & continue, Discard & continue, or Cancel. Triggered by generating field aliases or descriptions with AI, refreshing the schema, publishing the Data Mart, updating the input source, or leaving the page.

  • 662871c: API-key exchange and auth contextSection titled “API-key exchange and auth context”

    Project member API keys can now be exchanged for IDP access tokens in Better Auth and the development Null IDP. Exchanged tokens are bound to the API key ID and are rejected from API-key management, project-member administration, user provisioning, Intercom identity, and MCP OAuth authorization endpoints.

    • @owox/api-client now exposes client.auth.getContext(); owox-ctl status includes the project and member context resolved from the exchanged token, without exposing key secrets.
    • The auth-context endpoint uses IDP token introspection, and Better Auth refreshes API-key token context from current user and membership state during introspection.
    • Exchanged tokens include the project title used by auth-context responses, and now expire after 15 minutes, so clients re-exchange the underlying API key instead of holding indefinitely valid access tokens.
    • Magic links now carry the encrypted invite role through the callback path, so callback query normalization no longer drops the role before the user is added to the project.
  • 2a71167: Automatic retry and clearer error messages for Criteo Ads importsSection titled “Automatic retry and clearer error messages for Criteo Ads imports”

    Criteo Ads imports now automatically retry on server errors, rate limits, and network issues, so brief outages on Criteo’s side no longer stop an import. When an import still fails, logs now include the HTTP status, the provider’s own error message, and a note when the failure looks temporary. Previously a temporary server error failed the whole import immediately with only a raw stack trace in the logs.

  • d80b20d: Report Column Picker search and layout improvementsSection titled “Report Column Picker search and layout improvements”

    • Added a search bar that works across native and blended Data Mart columns.
    • Clear empty state when no matching columns are found.
    • Reorganized toolbar for easier access to selection controls and search.
    • Search works together with Show selected only and Select all.
  • 19a666c: Published OWOX Data Marts MCP in ChatGPT AppsSection titled “Published OWOX Data Marts MCP in ChatGPT Apps”

    OWOX Data Marts MCP in ChatGPT Apps catalog

  • @owox/internal-helpers@0.29.0
  • @owox/idp-protocol@0.29.0
  • @owox/idp-better-auth@0.29.0
  • @owox/idp-owox-better-auth@0.29.0
  • @owox/backend@0.29.0
  • @owox/web@0.29.0

OWOX Data Marts – v0.28.0

  • 0165f6b: Google Sheets Extension menu and sidebar updatesSection titled “Google Sheets Extension menu and sidebar updates”

    • All Reports — new menu item lists every report in the current document with its sheet name, last run status, and last run time, with direct access to open any report.
    • Create new report — new menu item creates a new sheet and opens the report creation form in one step.
    • Report run failure reason — hover the last run status icon in the sidebar to see why a run failed.
    • Reliable active sheet tracking — the sidebar correctly follows the active sheet when switching between sheets quickly.
  • dad37e4: Executed SQL in Run HistorySection titled “Executed SQL in Run History”

    Report runs now record the exact SQL that ran — with output controls applied and filter parameter values inlined as literals — in a dedicated Executed SQL block in Run History. Previously only the raw Data Mart query was stored, which didn’t reflect the parameters used at run time. Covers Google Sheets, Email, Slack, Google Chat, Microsoft Teams, and Looker Studio. The block appears only when output controls or blended SQL were applied.

  • 8b20d18, 2a2efa1: Output controls for Databricks and Snowflake

    Filters, slices, sort, and limit are now supported on Databricks and Snowflake Data Marts, matching support already available for BigQuery, Athena, Redshift, and Legacy BigQuery. Filter values are correctly escaped per storage type; substring matchers avoid LIKE so user % and _ stay literal. Available in the web app and the Google Sheets Extension.

  • 0165f6b: Quick run button for reportsSection titled “Quick run button for reports”

    Reports table now has a direct run button. A brief undo window appears before the run begins.

  • 0165f6b: Form validation and UI improvementsSection titled “Form validation and UI improvements”

    Submitting a form now opens any collapsed sections containing invalid fields and focuses the first error. Table filters now default to the is operator for fields that support both contains and is, and reset when switching filter fields. Project switcher menu received search, scroll-into-view, keyboard navigation, state reset on close, and accessibility improvements.

  • 113ef2e: Fail fast on disconnected report columns and output controlsSection titled “Fail fast on disconnected report columns and output controls”

    Reports now reject orphaned column references upfront — after a joined Data Mart alias is renamed, a relationship is removed, or a field disappears from the schema or is hidden — instead of letting them leak into generated SQL and fail with a cryptic “Unrecognized name” error. The column picker shows a “Disconnected columns” block for stale selections; stale filter/slice rules are highlighted in red in Output Controls; the Output Controls button badge turns red when any control references a disconnected field.

  • 19fbef7: Readable field names and search in Output ControlsSection titled “Readable field names and search in Output Controls”

    Filters, Slices, and Sort pickers now show the field’s alias (or leaf column name), with the joined Data Mart name as a muted second line for blended fields, instead of raw nested identifiers. Each picker is now searchable by alias, Data Mart name, or technical field name.

  • 25b43f6: Criteo placement, category, and transaction reportingSection titled “Criteo placement, category, and transaction reporting”

    Three new Criteo endpoints: placements, placement categories, and transactions. Added extra statistics dimensions (device, OS, channel) and reporting currency selection. Previously only campaign-level statistics were available.

  • ffd8ef8: Reliable incremental sync after partial connector runSection titled “Reliable incremental sync after partial connector run”

    The incremental date checkpoint is now written only after all accounts complete successfully. Previously, a mid-run failure could advance the checkpoint past data that was never fully fetched, silently skipping those dates on the next run. Affects MicrosoftAds, CriteoAds, GoogleAds, RedditAds, and XAds.

  • fddcd83: Criteo access token proactive refreshSection titled “Criteo access token proactive refresh”

    The Criteo connector now refreshes the access token 60 seconds before it expires and retries any request that still returns an expired-token error. Previously the cached token was reused for the entire run without expiry checks, causing mid-run failures on long-running syncs.

  • ec2fd8b: Reconnect notice when Google Storage authorization expiresSection titled “Reconnect notice when Google Storage authorization expires”

    Expired authorization is now detected before a connector run starts and surfaced as a “Reconnect Storage” prompt, instead of starting the run and failing with an opaque authentication error. The storage health indicator updates automatically when access is lost.

  • bbe8e34: Drag-and-drop for service account JSONSection titled “Drag-and-drop for service account JSON”

    Service account credential textareas now accept dropped JSON files, with JSON validation, file size limits, and error feedback.

  • b856d11: Shopify totalOutstanding fieldSection titled “Shopify totalOutstanding field”

    Added totalOutstanding to Shopify orders — the total amount not yet transacted for the order.

  • 39b4254: Session invalidation on password change and resetSection titled “Session invalidation on password change and reset”

    After a successful password change or reset, all other active sessions are revoked. The session that performed the change is preserved.

  • e6e9296: Unified column identifiers for output control slice filtersSection titled “Unified column identifiers for output control slice filters”

    Slice (pre-join) filters now reference columns by the same fully qualified identifier as regular output filters (e.g. category_details__item_event_count) instead of a raw column name plus a separate aliasPath. Existing saved reports are migrated automatically.

  • 1d2e04e: Bug fixes and improvementsSection titled “Bug fixes and improvements”

    • 1d2e04e: Fixed OAuth sign-in for ad platform connectors (Microsoft Ads, TikTok, LinkedIn, Google Ads) failing with “Not Found” after the provider authorization step — callback pages were routed to the backend instead of the web app. Affects both self-hosted and cloud deployments.
    • a222917: Fixed rotated Microsoft Ads refresh tokens not being persisted — OAuth responses returned a new token but it was discarded or overwrote the original user-provided token.
    • c110f35: Fixed AI Helper failing with Unrecognized name: <field> when the Output Schema contains disconnected fields. Disconnected fields are now excluded from the 30-row sample query; the fetch is skipped entirely when no connected fields remain.
    • f69f0b3: Fixed AI Helper failing with Syntax error: Expected end of input but got "-" for Data Marts whose fully qualified table name contains dashes (e.g. my-project.dataset.my-table-name). Table references and column names are now escaped per storage type via IdentifierEscaperFacade.
    • 4858830: Fixed missing validation feedback when a Google OAuth connection is required but not present.
  • @owox/internal-helpers@0.28.0
  • @owox/idp-protocol@0.28.0
  • @owox/idp-better-auth@0.28.0
  • @owox/idp-owox-better-auth@0.28.0
  • @owox/backend@0.28.0
  • @owox/web@0.28.0
  • c782374: Fix release build broken by duplicate google-auth-librarySection titled “Fix release build broken by duplicate google-auth-library”

    A transitive dependency drift left two copies of google-auth-library (v10) in the tree — one hoisted at the root and a second pinned exactly by googleapis-common — which produced incompatible OAuth2Client types and broke the TypeScript build of the Google Sheets and BigQuery integrations. A scoped npm override now makes googleapis-common reuse the root copy, so the packages build and publish reliably again.

OWOX Data Marts – v0.27.0

  • 3d80135: OWOX Data Marts API accessSection titled “OWOX Data Marts API access”

    Added owox-ctl, the OWOX Data Marts Control CLI, and @owox/api-client, a TypeScript/JavaScript API client for custom integrations. owox-ctl resolves credentials from environment variables, supports .env / --env-file, and defaults to OWOX Data Marts Cloud at https://app.owox.com.

    Available commands:

    • owox-ctl status
    • owox-ctl data-marts list
    • owox-ctl data-marts stream
    • owox-ctl storages list
    • owox-ctl destinations list
  • cb97897: HTTP Data API — stream Data Mart rows over HTTPSection titled “HTTP Data API — stream Data Mart rows over HTTP”

    A new GET /api/external/http-data/data-marts/{id}.ndjson endpoint streams a published Data Mart’s rows as newline-delimited JSON for project members authenticated with their ODM member token (x-owox-authorization).

    • columns=* returns Data Mart output columns; columns=** includes joined fields. Repeated column=<name> parameters select exact columns.
    • Optional base64url-encoded filter/sort and a limit are supported.
    • Only reporting-visible columns are selectable — hidden fields and excluded blend sources are rejected.
    • Every pull is recorded as an HTTP_DATA run in run history and counted for consumption.
  • c721847: User provisioning and project access requestsSection titled “User provisioning and project access requests”

    Added project-level user provisioning settings and context-scoped defaults for newly provisioned users. Users who receive a project token without assigned roles now stay inside a restricted ODM app shell, where they can request access to the current project or create a new one.

  • 02d3048: SQL preview icon on all reportsSection titled “SQL preview icon on all reports”

    The SQL preview icon is now visible on all reports (Google Sheets, Email, Looker Studio). For reports with output controls or joined fields, clicking it opens the generated SQL dialog. For reports without, it shows a tooltip explaining that the report queries the Data Mart source directly, with a link to Data Setup.

  • a47929c: Project-wide Data Mart activity pagesSection titled “Project-wide Data Mart activity pages”

    Added project-wide pages for Data Mart triggers, reports, insights, and run history. The lists support project-level search and filters and preserve row-level permissions for actions.

  • 8d83268, 3aefb24, 29c3af2: Output controls expanded to Athena, Legacy BigQuery, and Redshift

    Filters, slices, sort, and limit are now supported on AWS Athena, Legacy BigQuery, and AWS Redshift data marts, matching existing standard BigQuery support. Controls are available in both the web app and the Google Sheets Extension, including the generated-SQL preview and the Looker Studio cached path.

    Additionally, relative-date presets are tightened across all storages: last_n_days and last_n_months now stop at today (previously future-dated rows could leak in), and this_month / this_year are clamped to the current period.

  • 060e2f9: Automatic schema evolution for Snowflake and RedshiftSection titled “Automatic schema evolution for Snowflake and Redshift”

    Adding new fields to an existing Snowflake or Redshift Data Mart no longer causes MERGE or INSERT failures. Both storages now query the live table schema on every run and issue ALTER TABLE to add any missing columns before writing data.

  • 64fca08: Cancel Data Mart runs from Run HistorySection titled “Cancel Data Mart runs from Run History”

    In-progress connector and report runs can now be cancelled from Run History. Cancellation stops associated run triggers, marks the run as cancelled in history, and asks for confirmation that stopping a run can leave data incomplete.

  • 4f38486: Facebook Ads reach reporting at ad set and campaign levelSection titled “Facebook Ads reach reporting at ad set and campaign level”

    Two new Facebook Ads endpoints report daily performance aggregated at the ad set and campaign level, with reach deduplicated to match the numbers shown in Meta Ads Manager. Previously, reach was only available per ad, so totals summed across ads overcounted the same people.

  • fd7d9be: Google Sheets destination access email visible for both auth methodsSection titled “Google Sheets destination access email visible for both auth methods”

    The email address that needs document access was previously only shown for Service Account connections. It is now displayed and copyable in one click — in both the destination form and the report form — for Service Account and Google OAuth alike.

  • 62af9fb: Default row limit for Google Sheets reportsSection titled “Default row limit for Google Sheets reports”

    New Google Sheets reports now apply a default limit of 10,000 rows instead of returning all rows, to prevent hitting Google Sheets cell limits. The output controls icon reflects the applied limit, and a tooltip is shown before the first run.

  • 2f28f1d: Availability labels renamed to “Shared for…”Section titled “Availability labels renamed to “Shared for…””

    “Available for use”, “Available for reporting”, and “Available for maintenance” are now “Shared for use”, “Shared for reporting”, and “Shared for maintenance”. The “Not available” filter option is now “Not shared”.

  • 4409afe: Clearer API error messages for resource accessSection titled “Clearer API error messages for resource access”

    Error messages returned when access is denied or a resource is not found now use product names — “Data Mart”, “Data Storage”, “Data Destination”, “Technical User”, “Project Admin” — instead of the former internal identifiers DataMart, DataStorage, DataDestination, editor, admin.

  • 4656de0: Anonymous CLI usage telemetrySection titled “Anonymous CLI usage telemetry”

    owox serve now sends a single anonymous event on successful startup. Collected data: a random identifier (not derived from your machine or network), CLI/Node versions, OS platform/arch, and Docker/IDP/web flags. No hostnames, file paths, emails, or IP-derived identity are ever collected. Opt out by setting OWOX_TELEMETRY_DISABLED=1 or DO_NOT_TRACK=1. Automatically disabled in CI.

  • 0753de2: Faster Google Sheets import from BigQuery viewsSection titled “Faster Google Sheets import from BigQuery views”

    Data Marts based on a BigQuery VIEW now import into Google Sheets faster and more reliably for large datasets. Previously, large views could stop responding mid-import and deliver no data at all.

  • f573073: Improved SQLite runtime stabilitySection titled “Improved SQLite runtime stability”

    Improved runtime stability for SQLite-backed deployments, focusing on scheduled run processing and transaction coordination during concurrent background activity.

  • 62af9fb: Bug fixes and improvementsSection titled “Bug fixes and improvements”

    • 62af9fb: Fixed connector-based Data Mart definition type and connector settings appearing empty after switching tabs when no Storage was configured. Connector-based Data Marts can now be saved without a configured Storage.
    • c6d9af1: Fixed graph zoom controls for deep relationship graphs where fit-to-view scaled the graph below the previous fixed zoom floor.
  • @owox/internal-helpers@0.27.0
  • @owox/idp-protocol@0.27.0
  • @owox/idp-better-auth@0.27.0
  • @owox/idp-owox-better-auth@0.27.0
  • @owox/backend@0.27.0
  • @owox/web@0.27.0

OWOX Data Marts – v0.26.0

  • 10c540e: New OWOX Extension for Google Sheets — public release 🎉Section titled “New OWOX Extension for Google Sheets — public release 🎉”

    The next-generation OWOX Extension for Google Sheets is now published on the Google Workspace Marketplace, replacing the legacy add-on. It brings Data Marts directly into Google Sheets with a redesigned UX and reaches parity with the previous Extension on the workflows users relied on.

    • Column picker with field-level descriptions — pick which Data Mart fields to pull into the sheet without SQL, with the alias and description from the Data Mart’s Output Schema shown next to each field.
    • All Scheduled Reports page — a single place to see every report scheduled from the Extension, across spreadsheets and projects, subject to the user’s permissions.
    • Filters and slices on report output — apply pre-join slices and post-join filters to shape what a report returns without editing the Data Mart.
    • Refresh Current and Refresh All — re-run a single report or every report on the sheet in one click.
    • Granular permissions with Contexts support — the Extension honours the full permissions model: per-member, per-entity availability flags, and Project Contexts so a user sees only the Data Marts, Storages, and Destinations relevant to their team. When OAuth scopes are missing or revoked, the Extension explains what’s wrong and walks the user through re-granting access.
    • In-product support via Intercom — chat with the OWOX team without leaving the sidebar.
  • 85b150c: Default fields for connector endpointsSection titled “Default fields for connector endpoints”

    Each connector endpoint now ships with a curated set of pre-selected default fields, so users no longer have to choose fields from scratch on first use. Endpoint overviews and descriptions were also rewritten for clarity.

  • c083319: Data Mart run indicatorSection titled “Data Mart run indicator”

    An “Updating data” badge with an animated spinner now appears in the Data Mart header during active runs. The badge includes a “View runs” button for one-click access to run history. The manual run button is disabled while another run is in progress, with a tooltip explaining why.

  • 5b55bf2: Membership requests on the Members pageSection titled “Membership requests on the Members page”

    Project Admins can now review and act on pending project join requests from Project Settings → Members. Pending requests appear as a collapsible card above the members list. Clicking a request opens a side sheet to assign role, role scope, and contexts before approving, or to decline with a confirmation dialog. The section is admin-only and hidden when there are no pending requests.

  • 134884f: “Enable Google Sheets” checklist group for onboardingSection titled ““Enable Google Sheets” checklist group for onboarding”

    Users whose onboarding.use_case contains sync_dwh_sheets or import_external_sheets now see a dedicated “Enable Google Sheets” group in the Setup Checklist instead of the generic “Get data to your report” group. Three steps: create a Google Sheets Destination, install the Extension, and create and run a report from it. The group is non-blocking and auto-updates as users complete each step.

  • c520e17: Stricter OAuth credential checks when creating DestinationsSection titled “Stricter OAuth credential checks when creating Destinations”

    Creating a Google Sheets destination with a pre-existing credentialId now enforces two checks already present on destination updates:

    • Project ownership — the credential must belong to the same project. Cross-project references are rejected with 403 Forbidden — Credential does not belong to this project.
    • Copy permission — if the credential is already linked to another destination, the caller must have copy-credentials access on that destination. Without it, the request is rejected with 403 Forbidden — You do not have permission to copy credentials from this destination.

    Legitimate flows are unaffected — a fresh credentialId returned by the OAuth callback is not yet linked to any destination and passes both checks transparently.

  • 31ae838: Clarified Microsoft Ads account and customer ID fieldsSection titled “Clarified Microsoft Ads account and customer ID fields”

    Docs and in-product field tips now explicitly state that Account ID and Customer ID must be the numeric API identifiers from the aid and cid URL parameters, not the alphanumeric numbers shown elsewhere in the Microsoft Ads UI.

  • dcd4b55: Bug fixes and improvementsSection titled “Bug fixes and improvements”

    • dcd4b55: Fixed Business Owner assignment reducing a Technical User’s access. A Technical User tagged as Business Owner of a Data Mart that was Available for maintenance lost Edit, Delete, and Manage Triggers — actions they would have had as a non-owner Technical User. Permissions now combine two paths: an ownership floor (always guarantees See + Use for any owner) and the non-owner sharing path (role-based maintenance actions). Context and role scope still apply to maintenance actions: a member with Selected contexts only scope needs a context overlap with the Data Mart for maintenance access. Business Users assigned as Business Owner are unchanged — they still receive See + Use only, as the Business User role does not permit Data Mart maintenance.
    • ce7d14d: Fixed empty Data Mart titles when emoji or special characters are entered — users now see a clear message that these characters are not supported for Legacy BigQuery storage.
    • ce7d14d: Fixed unexpected delay when loading Destinations in the Extension; hidden reporting fields no longer appear in the column picker; improved empty state for new users.
    • da6e5f0: Fixed ambiguous column errors in blended report filters and sorting — WHERE and ORDER BY clauses now qualify every reference with the correct CTE alias (main.<col> for native columns, <subsidiary>.<alias> for blended fields). Fixed duplicate CTE names when two relationships shared the same targetAlias — CTE names are now derived from the full aliasPath (dots → underscores) to guarantee uniqueness.
    • d0f5455: Fixed React SVG attribute warnings in Microsoft Teams and Azure Synapse icon components.
    • c9f5bea: Fixed project notification settings endpoints (projects/:projectId/notification-settings*) accepting a caller-controlled project ID — an authenticated user could previously read or modify another project’s notification configuration.
  • @owox/internal-helpers@0.26.0
  • @owox/idp-protocol@0.26.0
  • @owox/idp-better-auth@0.26.0
  • @owox/idp-owox-better-auth@0.26.0
  • @owox/backend@0.26.0
  • @owox/web@0.26.0

OWOX Data Marts – v0.25.0

  • 9bb6393: AI helper for Data Mart metadataSection titled “AI helper for Data Mart metadata”

    A Sparkles ✨ button appears next to title, description, field alias, and field description inputs. Click it to get an AI-generated draft grounded in the Data Mart’s schema and up to 30 sample rows. Nothing is saved until you explicitly apply the suggestion.

    • Bulk fill available via Generate field descriptions and Generate field aliases next to Refresh schema — populates the whole Output Schema at once for review before saving.
    • Available for SQL, Table, View, and Table Pattern definitions. Connector-based Data Marts are not yet supported.
    • Self-hosted deployments: set AI_BASE_URL, AI_API_KEY, and AI_MODEL in the environment to enable the buttons.
  • 480a3be: Bulk import of Data Marts from BigQuerySection titled “Bulk import of Data Marts from BigQuery”

    A new Import… button on the Data Marts page opens a picker that lists every table and view in the selected BigQuery storage. Select up to 20 resources and click Create — each becomes its own Data Mart named after the resource. Sharded tables (e.g. events_20240101, events_20240102) are recognized automatically and shown as a single events_* entry that creates a Table Pattern Data Mart. If the storage has broken credentials, the dialog explains what’s wrong and offers a one-click shortcut to finish setup.

  • 9e5e3e0: Google Sheets export preserves column layout and side-by-side formulasSection titled “Google Sheets export preserves column layout and side-by-side formulas”

    Refreshing an OWOX → Google Sheets report no longer wipes the entire tab. The exporter only touches the cells it owns, so anything else in the same sheet — extra columns, formulas, pivot tables, charts, named ranges — stays intact across refreshes.

    • Your column order wins. Reordering columns in the Output Schema after the first refresh no longer rearranges the sheet.
    • New SQL columns are appended on the right; removed columns disappear cleanly and leave a #REF! signal on any formula that depended on them.
    • Formulas in the first data row are automatically filled down across all new rows.
    • Output Schema aliases update the sheet header without rewriting the whole tab.
    • A failed refresh no longer wipes existing data — the sheet stays as last seen and the Data Mart status flips to “failed”.
  • da9bd7e: Output controls per report — filters, sort, and row limitSection titled “Output controls per report — filters, sort, and row limit”

    A new Output controls panel in the Report Columns picker lets you shape what each report writes to its destination without changing the underlying Data Mart.

    • Filters apply WHERE-style conditions on the final SELECT. Multiple filters on the same column are combined with AND. Filtering by a column not selected for output is supported.
    • Sort orders output by one or more columns with asc/desc per column; drag to reorder priorities.
    • Limit caps the number of rows written. Leave empty for no limit.
    • Filter values are passed as named parameters (@p0, @p1, …), not inlined SQL.
    • Currently supported on BigQuery. Athena, Redshift, Snowflake, and Databricks will return a clear capability error until their adapters land.
  • 814790c: Report runs and trigger management for all Data Mart viewersSection titled “Report runs and trigger management for all Data Mart viewers”

    Manual report runs and Report Trigger CRUD are now available to any project member who can see the Data Mart and use the Destination — not just the Report owner. Report config editing (columns, filters, owners, destination) remains restricted to owners with an effective Destination.

  • 99acce6: New Data Marts, Storages, and Destinations are available by defaultSection titled “New Data Marts, Storages, and Destinations are available by default”

    Newly created entities now default to Available for reporting / use = ON, so other project members can see them immediately. The Available for maintenance toggle still defaults to OFF. Existing entities are not migrated.

  • 8cb62c9: Improved BigQuery location search and groupingSection titled “Improved BigQuery location search and grouping”

    Country names are now included in every location label (e.g. “Saudi Arabia” resolves to me-central2 Dammam). The catch-all “Other” group is replaced with dedicated Middle East, South America, and Africa groups. Three previously missing regions added: northamerica-south1 (Querétaro, Mexico), us-south1 (Dallas, United States), europe-west10 (Berlin, Germany).

  • 68ff850: Looker Studio destination renamed to Data StudioSection titled “Looker Studio destination renamed to Data Studio”

    Renamed consistently across the destination type dropdown, availability toggle, report cards, run history, sidebar help menu, setup checklist, and notification labels. Existing configurations and dashboards are unaffected.

  • a147210: Storage picker button integrated into input fieldSection titled “Storage picker button integrated into input field”

    The table/view picker button now appears inside the Fully Qualified Name input rather than next to it. Shows “Select…” with a database icon when empty; collapses to an icon-only button with “Change selection” tooltip when a value is set.

  • 3d3c8d3: Contextual promo and health UX for bulk Data Mart creationSection titled “Contextual promo and health UX for bulk Data Mart creation”

    When exactly one BigQuery Data Mart is configured, users now see a prompt to bulk-create Data Marts from existing tables and views. Resource browsers in storage pickers now show a health-state block instead of raw loading errors when the selected storage is not fully configured.

  • 118e029: Google Sheets promo and Email Reports owners columnSection titled “Google Sheets promo and Email Reports owners column”

    Published Data Marts without a Google Sheets destination now show a contextual promo with a direct CTA to create one. Email Reports table now displays report owners as an avatar group.

  • c2099e1: “New Report” and “New Trigger” labels and empty statesSection titled ““New Report” and “New Trigger” labels and empty states”

    Button labels standardized to “New Report” and “New Trigger” across the app. Empty states for destinations now show a prominent “New Report” button directly in place. Google Sheets destination empty state includes suggestions to invite teammates.

  • 0cbca4d: Report creation button enabled as soon as all fields are validSection titled “Report creation button enabled as soon as all fields are valid”

    The “Create & Run report” button no longer requires manually editing the default title to become active. It enables as soon as all required fields are filled.

  • 661b280: Data Mart table navigation and dialog improvementsSection titled “Data Mart table navigation and dialog improvements”

    • Cmd/Ctrl+Click on a Data Mart row opens it in a new tab.
    • Long Data Mart and report titles in delete dialogs now wrap correctly.
    • Sticky action cells no longer show a transparent background during scroll.
  • f717600: Accessibility and dark mode improvementsSection titled “Accessibility and dark mode improvements”

    Extracted reusable TableSelectionCheckbox component used across all tables. Added id/htmlFor label associations to Switch components. Enabled sorting by Contexts column in Data Mart, Storage, and Destination tables. Fixed dark mode styles for health status ring indicators.

  • 262a9c0: Bug fixes and improvementsSection titled “Bug fixes and improvements”

    • 262a9c0: Fixed wrong Input Source shown in the Data Marts list — the mismatch appeared when any Data Mart in the list had multiple owners or contexts assigned.
    • 4172cb7: Fixed incorrect column types and aggregations for blended fields from joined Data Marts in Looker Studio. COUNT/COUNT_DISTINCT now appear as numeric metrics; STRING_AGG/ANY_VALUE as dimensions; MIN/MAX/SUM use the correct matching aggregation. Fix applies across BigQuery, Snowflake, Redshift, Athena, and Databricks.
  • @owox/internal-helpers@0.25.0
  • @owox/idp-protocol@0.25.0
  • @owox/idp-better-auth@0.25.0
  • @owox/idp-owox-better-auth@0.25.0
  • @owox/backend@0.25.0
  • @owox/web@0.25.0

OWOX Data Marts – v0.24.0

  • b335552: Joinable Data Marts and joined reportsSection titled “Joinable Data Marts and joined reports”

    Data Marts can now be connected to each other, and reports can combine fields from several joined Data Marts in a single output.

    On a Data Mart’s Data Setup tab, the new Joinable Data Marts block lets you add joinable Data Marts, configure join conditions, choose which fields each joined Data Mart exposes, and override their aliases or aggregations. When editing a Google Sheets, Looker Studio, or Email report, the Report Columns picker now shows fields from the base Data Mart together with fields from all joined Data Marts. The generated SQL is available for inspection, and any joined report can be saved as a standalone Data Mart.

    Supported on BigQuery, Snowflake, Redshift, Athena, and Databricks.

  • d187cdf: Project contexts and role scopeSection titled “Project contexts and role scope”

    Added business-domain Contexts and per-member Role Scope to narrow what shared resources a non-owner can see.

    • Admins define Contexts in Project Settings, attach them to Data Marts, Storages, and Destinations, and pick between Entire project and Selected contexts for each member.
    • Owners and admins are never gated by Contexts.
    • Introduced a unified Project Settings page (Overview, Members, Contexts, Credit consumption, Subscription, Notification) that replaces the separate Members page.
  • e817c93: Setup checklist for guided onboardingSection titled “Setup checklist for guided onboarding”

    Added a Setup Checklist that guides new users through the key steps required to start working with the product. Each step includes a direct call to action, the checklist tracks progress in real time, and steps are grouped into logical stages.

  • c669d11: OAuth for LinkedIn Ads and LinkedIn PagesSection titled “OAuth for LinkedIn Ads and LinkedIn Pages”

    Added OAuth authorization flow for both LinkedIn connectors, including UI login, callback handling, credential exchange, and token-backed execution. Existing LinkedIn configurations are migrated to the OAuth2 structure with compatibility for previously saved settings.

  • ba44a8d: Table and view browser for BigQuery Data MartsSection titled “Table and view browser for BigQuery Data Marts”

    A new Fill from Storage button lets you browse available tables and views through a project → dataset → table hierarchy. Selecting a table or view automatically fills the Fully Qualified Name field. Manual entry remains available as an alternative.

  • d118b10: Improved Data Mart publishing experienceSection titled “Improved Data Mart publishing experience”

    • Data starts loading automatically after publishing a connector-based Data Mart.
    • Guidance now focuses on scheduling automatic updates instead of manual runs.
    • Promo messages appear only once.
    • Updated button labels and tooltips for better clarity.
  • 3f8b831: Data Mart page layout and table usability improvementsSection titled “Data Mart page layout and table usability improvements”

    • Sticky actions column keeps row actions always visible during horizontal scroll.
    • Row actions are emphasized on hover to reduce visual noise.
    • Long values (including strings without spaces) now wrap correctly in tables and titles.
    • Refined layout behavior across different screen sizes.
    • Scrollable tabs now show a gradient fade indicator at the edge to signal hidden overflow.
  • d8e7e84: Project Settings documentationSection titled “Project Settings documentation”

    New documentation section covering members, roles, and resource access. Includes three pages: Members Management, Roles and Permissions, and Ownership and Availability (with per-entity access tables for Storage, Data Mart, Destination, Report, and their Triggers).

  • ba44a8d: Enhanced security headersSection titled “Enhanced security headers”

    Improved HTTP security headers now apply automatically to all application pages and authentication screens: Content Security Policy against clickjacking, Strict Transport Security for HTTPS enforcement, XSS and content-type protection, and referrer policy controls.

  • ba44a8d: Improved BigQuery configuration validationSection titled “Improved BigQuery configuration validation”

    Invalid project ID values are now caught immediately when saving OAuth-authenticated BigQuery storages, with clear error messages to correct typos or formatting issues before they cause runtime errors.

  • 160574a: Bug fixes and improvementsSection titled “Bug fixes and improvements”

    • 160574a: Unconfigured storages now show a grey health indicator with the message “Complete setup to activate Storage” instead of a red one. Red is reserved for storages that have been configured but fail validation.
    • 50366c4: Fixed campaign_id and adgroup_id fields in tiktok_ads_ad_insights — both were always null due to missing parent-hierarchy IDs in the API request. Both fields are now correctly populated.
  • @owox/internal-helpers@0.24.0
  • @owox/idp-protocol@0.24.0
  • @owox/idp-better-auth@0.24.0
  • @owox/idp-owox-better-auth@0.24.0
  • @owox/backend@0.24.0
  • @owox/web@0.24.0

OWOX Data Marts – v0.23.0

  • d8b82e5: Ownership foundationsSection titled “Ownership foundations”

    Ownership is now explicit and visible across all major entities. This is the first stage of the Permissions Model Evolution — ownership is informational only and does not affect access control.

    • Data Mart has separate Technical Owner and Business Owner, editable on the Overview tab.
    • Storage, Destination, and Report each have an Owners list, editable in the configuration sheet and saved together with the entity.
    • Creator is auto-assigned as owner when a new entity is created.
    • Owners column is displayed in Storage, Destination, and Report list tables.
    • Filter by owner is available in Data Marts, Storages, and Destinations lists.
    • “Not assigned” state is shown when an entity has no owners.
  • 11c0c77: Availability model and owner-based access controlSection titled “Availability model and owner-based access control”

    Introduced explicit availability settings for DataMarts, Storages, and Destinations (Available for reporting / Available for maintenance and Available for use / Available for maintenance). Ownership now affects access: owners have direct access to their entities regardless of availability state, while non-owners see only available entities.

    • Implemented table-driven AccessDecisionService that evaluates entity type, role, ownership status, availability state, and action for every single-entity operation.
    • Enforced access checks on all endpoints including direct URL access, sub-operations (SQL dry run, validate definition, run history), and report creation.
    • Existing entities default to fully available for backward compatibility; new entities default to not available (secure by default).
  • 11c0c77: Extended role definitions and report ownershipSection titled “Extended role definitions and report ownership”

    Reinterpreted existing project roles without changing stored identifiers: Viewer becomes Business User with self-service reporting capabilities, Editor becomes Technical User with full technical maintenance access.

    • Business Users can now create, edit, delete, and run their own Reports and manage Report Triggers.
    • Business Users can also create, edit, and delete Destinations.
    • Technical Users retain project-wide Report management.
    • Implemented ineffective owner logic: if a Report owner loses access to the DataMart or Destination, they become read-only until access is restored.
  • 004b9a5: Invite teammates from setup screensSection titled “Invite teammates from setup screens”

    Added contextual “Invite teammates” helper blocks to setup and empty state screens. Includes a reusable component with responsive layout and optional documentation link.

  • 0f07594: Auto-select the only storage when creating a Data MartSection titled “Auto-select the only storage when creating a Data Mart”

    When creating a Data Mart, if the project has only one storage, that storage is now selected automatically.

  • 870de85: Geographic performance reports for Google Ads connectorSection titled “Geographic performance reports for Google Ads connector”

    Added Geo Stats and Geo Target Constants nodes to the Google Ads connector. The two tables are designed to be joined in BigQuery on country_criterion_id.

  • 4033235: X Ads: country breakdown for ad statsSection titled “X Ads: country breakdown for ad stats”

    Added stats_by_country node for daily ad stats broken down by country. A companion targeting_locations reference table maps X Ads location IDs to country names and ISO codes — join on the country field.

  • 15390c6: Progressive save for Facebook Ads dataSection titled “Progressive save for Facebook Ads data”

    Data is now saved to BigQuery page by page. A failure only affects the remaining pages, not the entire dataset.

  • 126280a: Bug fixes and improvementsSection titled “Bug fixes and improvements”

    • 126280a: Fixed empty error text for invalid storage status — now shows “Access validation failed” instead of a blank space.
    • a848ac4: Fixed Storage column sorting on the Data Marts page — storages with custom names now sort correctly by their visible name.
    • 832bb73: Error boundaries for unexpected application crashes. Replaced the default React error screen with a user-friendly fallback UI. The sidebar stays visible for in-layout errors so users can navigate away without a full page reload.
  • @owox/internal-helpers@0.23.0
  • @owox/idp-protocol@0.23.0
  • @owox/idp-better-auth@0.23.0
  • @owox/idp-owox-better-auth@0.23.0
  • @owox/backend@0.23.0
  • @owox/web@0.23.0

OWOX Data Marts – v0.22.0

  • 4c9b96d: Business Owner & Technical Owner for Data MartsSection titled “Business Owner & Technical Owner for Data Marts”

    Assign Business and Technical Owners to each Data Mart to track accountability and maintainability.

    • Owners section on the Data Mart edit page — assign one or more project members per role via an inline selector.
    • Search and role display — the member selector shows each user’s project role with a search input for quick filtering.
    • Outbound member warnings — previously assigned owners who have been removed from the project are shown with a warning icon for easy reassignment.
    • Business Owner and Technical Owner columns in the Data Marts list table (hidden by default — enable via column settings).
    • Filter by owner in the Data Marts list to quickly find data marts by a specific team member.
    • Auto-assigned Technical Owner — the creator is automatically set as Technical Owner when a new Data Mart is created.
  • b178723: Created By visibility across major entitiesSection titled “Created By visibility across major entities”

    See who created each entity directly in the list views.

    • Created By column is now available in Data Storages, Data Destinations, Reports, Scheduled Triggers, and Insights tables.
    • Filter by creator is available in Data Destinations and Data Storages lists.
  • 85abd92: Trigger-based execution for connector and report runsSection titled “Trigger-based execution for connector and report runs”

    Connector and report runs are now processed through a task queue instead of running immediately in the background. This improves reliability by ensuring runs are not lost on server restart, adds per-project concurrency limits, and automatically retries runs that cannot start due to concurrency limits. Includes a safety mechanism to detect and fail runs stuck in the queue for too long.

  • 2406874: Onboarding video for Email-based ReportsSection titled “Onboarding video for Email-based Reports”

    A new onboarding video to improve adoption of Email-based Reports in Data Marts.

    • Shown once to new users on the Destinations tab in Data Mart.
    • Available in Help menu → Video tutorials.
    • Embedded in Email Reports documentation.
  • 9395757: Sign In page redesign with product-focused brand panelSection titled “Sign In page redesign with product-focused brand panel”

    Redesigned the auth screen to better communicate product value and reduce sign-in friction.

    • Brand panel — replaced placeholder with a carousel of product use cases (Google Sheets, Looker Studio, Email delivery).
    • Headline — “Build once. Share anywhere.” with outcome-driven copy and short supporting text per slide.
    • Product preview — visual workflows introduced directly on the auth screen.
    • Carousel — auto-rotation with lazy-loaded images.
  • 737f292: Externalized connector secrets from Data Mart definitionsSection titled “Externalized connector secrets from Data Mart definitions”

    Non-OAuth secrets are moved from inline storage in Data Mart definitions to a separate connector_source_credentials table. Centralizes credential storage and reduces secret exposure in definition JSONs.

    • Added _secrets_id reference pattern (aligned with existing _source_credential_id for OAuth).
    • Secrets are extracted on save and injected during connector execution.
    • Includes data migration for existing Data Marts.
  • a258b72: Improved UX for setting triggersSection titled “Improved UX for setting triggers”

    Reduced friction in the trigger setup flow: smart default type based on Data Mart configuration, improved empty state with a CTA button, and one-click schedule presets (Daily 9:00, Every hour, Every 6h, Weekdays 9:00).

  • 2399254: Searchable Storage Selector in Data Mart formSection titled “Searchable Storage Selector in Data Mart form”

    The storage picker in the Create Data Mart form is upgraded to a searchable combobox.

    • Storages are listed in alphabetical order by title.
    • Typeahead search filters the list by typing — useful when many storages exist.
    • Create new storage option remains accessible at the bottom of the list.
  • 62f435a: Auto-subscribe new project members to notification settingsSection titled “Auto-subscribe new project members to notification settings”

    New team members with Admin or Editor roles are automatically subscribed to existing notification settings when joining a project. Members who leave and rejoin are re-subscribed automatically. Manual unsubscribes are respected and not overridden. Members downgraded to Viewer are automatically removed from the receivers list.

  • 6d53e58: Bug fixes and improvementsSection titled “Bug fixes and improvements”

    • Connector definition validation — fixed validators to allow early validation success; connector definitions no longer require credential validation during publish, resolving failures for Athena, BigQuery, Databricks, Redshift, and Snowflake.
    • Delete confirmation dialog — fixed undefined Data Mart title; dialog now correctly displays the actual data mart name.
    • Edit button in oneOf config — secret editing state is now tracked per-field, preventing unintended resets of sibling fields and auth type switches.
  • @owox/internal-helpers@0.22.0
  • @owox/idp-protocol@0.22.0
  • @owox/idp-better-auth@0.22.0
  • @owox/idp-owox-better-auth@0.22.0
  • @owox/backend@0.22.0
  • @owox/web@0.22.0

OWOX Data Marts – v0.21.0

  • 7494cd6: Introducing reusable Insights with AI assistanceSection titled “Introducing reusable Insights with AI assistance”

    Insights is now available in OWOX! Use it to analyze your data mart results directly — create reports, explore your data, and share findings with your team. To get started, check out the Insights setup guide.

  • a7bbb8c: Add OAuth for Google Ads connectorSection titled “Add OAuth for Google Ads connector”

    Added OAuth2 authentication flow for Google Ads connector. Users can now authorize access using a “Sign in with Google” button instead of manually entering RefreshToken, ClientId, and ClientSecret. The DeveloperToken is managed via environment variable and stored securely with other OAuth credentials. Also fixed a COOP SecurityError in the OAuth popup polling that affected all OAuth connectors.

  • d906258: Add “Copy Credentials” button to Storage and DestinationSection titled “Add “Copy Credentials” button to Storage and Destination”

    Added a “Copy Credentials” button to Storage and Destination edit forms, allowing users to copy credential configuration from other storages or destinations of the same type. Includes new backend endpoints for listing storages and destinations by type with credential identity information.

  • e68ca61: Improve Data Mart publish status messageSection titled “Improve Data Mart publish status message”

    The publish status message for data marts has been improved to provide more clarity and actionable information.

    • Publish after: The publish status message now shows what you need to do to publish the data mart. For example, if you need to complete storage configuration or configure an input source.
    • Ready to publish: The publish status message now shows that the data mart is ready to publish.
  • cb57600: Add video tutorials about completing storage setup and insightsSection titled “Add video tutorials about completing storage setup and insights”

    In the Help menu, we added two new video tutorials:

    • how to complete the Google BigQuery (used in OWOX extension) storage setup to publish Data Marts from OWOX Reports (Google Sheets extension)
    • how to get started with Insights
  • 27e5af8: AWS Athena storage descriptionsSection titled “AWS Athena storage descriptions”

    Added support for column descriptions (comments) in AWS Athena storage and Output Schema.

  • 90b7ee7: Implement destination name validation for all data storage typesSection titled “Implement destination name validation for all data storage types”

    Implements comprehensive destination name validation for CONNECTOR and TABLE definition types across BigQuery, Snowflake, Athena, Redshift, and Databricks. This update strengthens security by preventing malformed table names and ensuring all storage configurations are properly validated before processing.

  • 1a7b5f8: Improve form UI with collapsible sectionsSection titled “Improve form UI with collapsible sections”

    This update introduces collapsible sections in forms to make configuration flows easier to complete. Less important or advanced settings can now be hidden by default, allowing users to focus on the most important fields first. For example, this approach can help when configuring storages that are automatically created during integration with the OWOX Reports extension for Google Sheets.

    The interaction follows the same collapsible pattern used in wizard screens, keeping the experience consistent across the product. Existing forms remain unchanged and continue to appear expanded by default.

  • 6f1f61a: Show storage health status in the configuration formSection titled “Show storage health status in the configuration form”

    This update improves visibility of the storage connection status when configuring a Data Storage. Users can now see whether the storage access is valid directly in the General section of the configuration form. This provides immediate feedback that the storage is correctly configured and ready to be used by Data Marts. This makes the setup experience clearer and helps users configure their storage with more confidence.

    Benefits:

    • Instantly understand if the storage connection is working
    • Quickly detect configuration issues
    • Maintain a consistent status indicator between the Data Storage list and the configuration form
  • 7494cd6: Bug fixes and improvementsSection titled “Bug fixes and improvements”

    • Fix data storage validation for Google Legacy BigQuery connector
    • Fix missing dependencies for data destination and storage forms to avoid errors
    • Fix duplicate trigger runs caused by MySQL deadlocks not being handled as transient errors
    • Fix Google BigQuery storage not saving OAuth credentials and blocking OAuth-authenticated users from saving storage settings
    • Fix field descriptions and primary key flags being lost after schema sync in Redshift and Snowflake data marts
  • @owox/internal-helpers@0.21.0
  • @owox/idp-protocol@0.21.0
  • @owox/idp-better-auth@0.21.0
  • @owox/idp-owox-better-auth@0.21.0
  • @owox/backend@0.21.0
  • @owox/web@0.21.0

OWOX Data Marts – v0.20.0

  • @owox/internal-helpers@0.20.0
  • @owox/idp-protocol@0.20.0
  • @owox/idp-better-auth@0.20.0
  • @owox/idp-owox-better-auth@0.20.0
  • @owox/backend@0.20.0
  • @owox/web@0.20.0

OWOX Data Marts – v0.19.0

  • @owox/internal-helpers@0.19.0
  • @owox/idp-protocol@0.19.0
  • @owox/idp-better-auth@0.19.0
  • @owox/idp-owox-better-auth@0.19.0
  • @owox/idp-owox@0.19.0
  • @owox/backend@0.19.0
  • @owox/web@0.19.0

OWOX Data Marts – v0.18.0

  • 68d72af: Add Databricks storage type with Personal Access Token authenticationSection titled “Add Databricks storage type with Personal Access Token authentication”

    This change adds support for Databricks as a new storage type in OWOX Data Marts platform. Users can now connect to Databricks SQL warehouses using Personal Access Token authentication.

    • Authentication: Personal Access Token
    • Storage Configuration: Host, HTTP Path
    • Driver: @databricks/sql
  • ee0459e: Add OAuth flow for TikTok Ads connectorSection titled “Add OAuth flow for TikTok Ads connector”

    • Added support for OAuth2 authentication in the TikTok Ads connector
    • Implemented OAuth credential exchange
    • Added TikTok login button UI component for OAuth flow
    • Added OAuth callback page for handling TikTok authorization redirect
    • Manual credential entry option remains available as fallback
  • 38d3593: Improved Looker Studio destination stabilitySection titled “Improved Looker Studio destination stability”

    Enhanced Looker Studio destination stability and performance by implementing data streaming. This update ensures a smoother user experience and more reliable delivery of large datasets from Data Marts.

  • 38d3593: Descriptions for fields in AWS Redshift tablesSection titled “Descriptions for fields in AWS Redshift tables”

    After launching the connector, field descriptions will be synchronized with tables in Redshift

  • f52376f: Enrich Facebook Marketing ad-account/ads endpointSection titled “Enrich Facebook Marketing ad-account/ads endpoint”

    The Facebook Marketing connector’s ad-account/ads endpoint now returns real ad data instead of null values.

    What’s New:

    • Added 19 useful fields including ad name, status, campaign/adset relationships, creative details, and timestamps
    • 8 essential fields are now pre-selected by default (id, name, status, effective_status, adset_id, campaign_id, created_time, updated_time)

    Before: Only 3 placeholder fields that returned null After: Comprehensive ad data ready for analytics

    Note: For performance metrics (impressions, clicks, spend), continue using the ad-account/insights endpoint.

  • bcf4b10: Fix Snowflake data mart schema derivation to properly handle queries with LIMIT clauses by wrapping them in subqueries instead of naive concatenationSection titled “Fix Snowflake data mart schema derivation to properly handle queries with LIMIT clauses by wrapping them in subqueries instead of naive concatenation”

    Enhanced Snowflake data mart schema derivation to properly handle queries with LIMIT clauses by wrapping them in subqueries instead of naive concatenation. This ensures that the schema is derived correctly even when the query contains a LIMIT clause.

  • 997fcba: Update Snowflake storage UI to use PAT terminologySection titled “Update Snowflake storage UI to use PAT terminology”

    Updated the Snowflake storage settings interface to refer to “PAT (Programmatic Access Token)” instead of “Password” to align with Snowflake’s current terminology.

    This is a visual change only. Your existing configurations remain unchanged, and no action is required on your part. Everything will continue to work as before.

  • @owox/internal-helpers@0.18.0
  • @owox/idp-protocol@0.18.0
  • @owox/idp-better-auth@0.18.0
  • @owox/idp-owox@0.18.0
  • @owox/backend@0.18.0
  • @owox/web@0.18.0

OWOX Data Marts – v0.15.0

  • bfdbc6b: Insights feature is now available for all projects in the cloud versionSection titled “Insights feature is now available for all projects in the cloud version”

    Users can now try out the new Insights functionality and share their feedback.

  • 3283bcd: Insights: Improved permission handling and read-only modeSection titled “Insights: Improved permission handling and read-only mode”

    We’ve enhanced how permissions work within Insights to provide a more seamless and informative experience.

    • Read-only state: You can now view Insights in a read-only mode when you don’t have edit permissions.
    • Action enforcement: Buttons and actions (like editing or deleting) are now properly disabled based on your access level.
    • Better feedback: New tooltips and messages explain why certain actions are restricted, helping you understand your permission levels at a glance.
  • 51a9bdd: Added instant health status indicators to the Data Marts listSection titled “Added instant health status indicators to the Data Marts list”

    Improved Data Marts list with instant health status indicators and automatic prefetching of recent run data, making it easier to monitor Data Mart health at a glance.

    • Color-coded health status indicators for each Data Mart (success, failure, mixed results, in progress, or no recent runs)
    • Draft Data Marts clearly communicate that publishing is required before runs are available
    • Automatic prefetching of run statuses for visible rows — no waiting, no hover-to-load surprises
    • Helpful hover details with recent run info and a quick path to full Run History
  • d15a3c4: Fix bugs in Data Mart “Run History” tabSection titled “Fix bugs in Data Mart “Run History” tab”

    Fixed bugs that users may have encountered in the Data Mart “Run History” tab, namely:

    • Unexpected visual effects when clicking the “Load More” button.
    • Automatic deletion of run items that were loaded via the “Load More” button, and displaying only the last 20 runs.
  • c10c3ff: Refactor: Standardize Data Types Across ConnectorsSection titled “Refactor: Standardize Data Types Across Connectors”

    Introduced centralized data type definitions and standardized type handling across all storage connectors and API references.

    • Created Constants/DataTypes.js with standardized type definitions (STRING, BOOLEAN, INTEGER, NUMBER, DATE, DATETIME, TIME, TIMESTAMP, ARRAY, OBJECT)
    • Updated all storage connectors (AWS Athena, AWS Redshift, Google BigQuery, Snowflake) to use standardized DATA_TYPES constants for type mapping
    • Refactored field definitions across all source connectors (Facebook Marketing, Google Ads, LinkedIn Ads, Microsoft Ads, Reddit Ads, TikTok Ads, X Ads, Shopify, GitHub, etc.) to use consistent data type references
    • Changed AbstractStorage.getColumnType() to throw an error if not implemented, enforcing proper implementation in child classes
    • Eliminated storage-specific type constants (e.g., GoogleBigQueryType) from API reference files where they were inappropriately used
  • 7d7ddd1: Add new fields to Facebook Ads source connectorSection titled “Add new fields to Facebook Ads source connector”

    Added cost_per_result, results and result_rate fields to the Facebook Ad Account Insights report schema, allowing for better cost efficiency analysis.

  • dad1fd2: Feat: Add default fields for Microsoft AdsSection titled “Feat: Add default fields for Microsoft Ads”

    We’ve improved the Microsoft Ads connector to help you set up reports faster.

    Now, when you select the User Location Performance Report or Campaigns, we automatically pre-select common useful fields (like Impressions, Clicks, Spend) so you don’t have to search for them.

    • Convenient: Standardization of commonly used metrics.
    • Flexible: You can still uncheck these fields if you don’t need them.
  • 4d4b7e9: Add new fields to Shopify ordersSection titled “Add new fields to Shopify orders”

    • discountCodes: Array of discount code strings applied to the order
    • discountApplications: Detailed discount applications with code, amount/percentage, target type, and allocation method
  • 3f381f7: Fix: TikTok Connector Type HandlingSection titled “Fix: TikTok Connector Type Handling”

    Fixed “Unknown type STRING” errors in the TikTok Ads connector by updating type handling to use standardized DATA_TYPES constants.

  • @owox/internal-helpers@0.17.0
  • @owox/idp-protocol@0.17.0
  • @owox/idp-better-auth@0.17.0
  • @owox/idp-owox@0.17.0
  • @owox/backend@0.17.0
  • @owox/web@0.17.0

OWOX Data Marts – v0.15.0

  • 81112e2: Floating Popovers and Help Menu for Tutorials and SupportSection titled “Floating Popovers and Help Menu for Tutorials and Support”

    • Added FloatingPopover component with context to manage popover state
    • Added video tutorial popovers for Google Sheets and Looker integration
    • Implemented HelpMenu with dropdown functionality for quick access to help resources
    • Integrated FloatingPopoverProvider and HelpMenu into AppSidebar for a better user experience
    • Enhanced Intercom integration with launcher visibility control and payload adjustments
    • Added popover tutorial in EmptyDataMartsState to guide new users
  • 18c2a9f: AI-Powered Insight CreationSection titled “AI-Powered Insight Creation”

    Added the ability to create the first insight using AI, significantly simplifying the user’s journey to start using insights functionality.

    Users can now generate intelligent, data-driven insight templates with pre-configured AI prompts tailored to their data mart structure, making it easier to discover valuable insights without manual setup.

  • 8d8b75b: AWS Redshift Storage SupportSection titled “AWS Redshift Storage Support”

    Introduced AWS Redshift as a new data storage type with comprehensive support for both Serverless and Provisioned clusters.

    • Authentication Methods: Username/Password authentication for secure connections
    • Flexible Configuration: Support for both Redshift Serverless (workgroup-based) and Provisioned (cluster-based) deployments
    • Complete Schema Management: Added schemas, guards, and services for Redshift data mart configuration and credentials
    • Enhanced Frontend: Updated UI components to handle Redshift configuration with relevant descriptions and validation
    • Type Safety: Implemented comprehensive DTOs, mappers, and type guards for Redshift data structures
    • Full Documentation: Included detailed documentation for setting up and using the Redshift connector

    This update enables users to seamlessly connect OWOX to AWS Redshift data warehouses, expanding the platform’s data storage capabilities to one of the most popular cloud data warehouse solutions.

  • 7d9eb45: Streamlined Data Mart OnboardingSection titled “Streamlined Data Mart Onboarding”

    Introduced interactive “next-step” suggestions to help you set up and use your Data Marts faster. Now, after publishing or loading data, you’ll get clear guidance on the best next action—like running a connector, scheduling triggers, or creating reports—ensuring a smoother transition from configuration to results.

  • 09c6270: Add direct linking for Data Storages and DestinationsSection titled “Add direct linking for Data Storages and Destinations”

    • Added support for direct links to specific Data Storages and Destinations via URL parameters, making it easier to share specific entities with others.
    • Automatically synchronizes the URL with the currently opened entity.
  • 2f99ff4: Improve paginationSection titled “Improve pagination”

    • Improved pagination on lists of Data Marts, Storages and Destinations: added customizeble page-size selector and info about selected items and pages
    • Added persistent page size support for storing and retrieving the user’s preferred page size from localStorage
    • Fixed “Select all” checkbox to operate on the current page only
  • 2d311f6: Improve UX in ReportsSection titled “Improve UX in Reports”

    • added buttons for running the report and opening the Google Sheets document directly in the table row
    • changed the default action for the new report from “Create” to “Create and Run” action. It’s affected Google Sheets and Email destinations
  • 6940365: Add Shopify ConnectorSection titled “Add Shopify Connector”

    Added new Shopify connector with support for multiple data nodes including orders, products, customers, metafields, and more

  • 469e64f: Improved Facebook Marketing Connector Stability and Data DepthSection titled “Improved Facebook Marketing Connector Stability and Data Depth”

    • More Reliable Imports: Fixed an issue where missing or invalid dates could interrupt data transfers. Your pipelines will now be more resilient to data inconsistencies.
  • c9e7dc5: Microsoft Ads: AccountID changes to AccountIDsSection titled “Microsoft Ads: AccountID changes to AccountIDs”

    The Microsoft Ads connector configuration field has been renamed from AccountID to AccountIDs to better reflect its capability. You can now specify multiple Account IDs (comma-separated) in a single field, allowing you to load data for several accounts using one connector instead of creating separate connectors for each account.

    Existing Microsoft Ads connectors will be automatically migrated with no action required on your part.

  • @owox/internal-helpers@0.16.0
  • @owox/idp-protocol@0.16.0
  • @owox/idp-better-auth@0.16.0
  • @owox/idp-owox@0.16.0
  • @owox/backend@0.16.0
  • @owox/web@0.16.0

OWOX Data Marts – v0.15.0

  • 8298a39: Improve Storage Creation UX and Interaction SafetySection titled “Improve Storage Creation UX and Interaction Safety”

    The Storage creation flow is now smoother and more predictable.

    • Added click-lock protection to prevent accidental double-clicks when selecting a Storage type. This prevents duplicate creation requests and ensures users trigger the action only once.
    • Buttons now automatically switch to a disabled state while a new Storage is being created. This provides clear visual feedback and blocks unwanted interactions during processing.

    These updates improve safety, reduce friction, and help users stay in control while the system handles their request.

  • Enhance Report Creation Workflow and Scheduling

    The Report creation experience has been upgraded to support immediate execution and seamless scheduling.

    • Integrated schedule triggers directly into the creation screen. This allows users to configure data delivery rules upfront, maintaining consistency with the Data Mart Triggers tab.
    • Implemented a split-action button for the creation step. Users can now choose to not only save but also immediately run the report, eliminating the need to navigate back to the list to trigger a manual run.

    These updates reduce navigation friction and unify the definition and automation steps.

  • 81aedad: Fix Report Reader for View-Defined Data Marts in BigQuery StorageSection titled “Fix Report Reader for View-Defined Data Marts in BigQuery Storage”

    Fixed report reader functionality for data marts defined by views in BigQuery Storage.

    • Enhanced definition type checking in BigQuery Storage report reader to properly distinguish between definition types.
    • Added explicit definitionType parameter validation to ensure correct handling of view-based data mart definitions in BigQuery Storage.
  • @owox/internal-helpers@0.15.0
  • @owox/idp-protocol@0.15.0
  • @owox/idp-better-auth@0.15.0
  • @owox/idp-owox@0.15.0
  • @owox/backend@0.15.0
  • @owox/web@0.15.0

OWOX Data Marts – v0.14.0

  • 30d95a8: Fix migration error on application start with SQLite databaseSection titled “Fix migration error on application start with SQLite database”

    Fixed an issue where users running OWOX with SQLite database could encounter errors like “SQLITE_ERROR: no such column: runType” during database migrations.

  • 1c03024: Added additional information about users in UISection titled “Added additional information about users in UI”

    • Added display of the Data Mart creator in the Data Marts list.
    • Added display of the run initiator on the Run History tab.
  • 2c636b8: Add Snowflake support for Data Marts and ConnectorsSection titled “Add Snowflake support for Data Marts and Connectors”

    You can now use Snowflake as a data storage destination for both Data Marts and Connectors, giving you more flexibility in how you store and manage your data.

    Snowflake Data Storage

    • Create and manage Data Marts with Snowflake as the destination
    • Configure Snowflake connections with username/password or key-pair authentication
    • Automatic schema detection and validation for Snowflake tables
    • Run SQL queries directly on your Snowflake data warehouse

    Connector Support

    • Load data from any connector directly into Snowflake tables
    • Automatic table creation and schema management
    • Support for merging and upserting data based on unique keys
    • Proper handling of Snowflake data types including VARIANT for JSON data

    Data Mart Features

    • Read data from Snowflake tables for reports and exports
    • Support for custom SQL queries as Data Mart sources
    • Schema customization with Snowflake-specific field types
    • Case-sensitive table and schema name support
    • Support Data Studio features
    • Support for AI Insights features
  • @owox/internal-helpers@0.14.0
  • @owox/idp-protocol@0.14.0
  • @owox/idp-better-auth@0.14.0
  • @owox/idp-owox@0.14.0
  • @owox/backend@0.14.0
  • @owox/web@0.14.0

OWOX Data Marts – v0.13.0

  • 5adbb9f: New: Email-based destinationsSection titled “New: Email-based destinations”

    New: Email as a data destination

    • You can now add Email as a destination and deliver reports directly to inboxes.

    New: Email report editor

    • Create and configure email-based report deliveries with a dedicated editor.
    • Compose messages using Markdown for clear, readable emails.
    • Inline descriptions help you set up destination type, message template, and sending conditions.
    • Preview and save your email delivery configuration.

    New: Slack, Microsoft Teams and Google Chat integrations added via the email channel.

    Edition-aware visibility

    • Features are shown or hidden based on your app edition to keep the UI clear and relevant.

    No breaking changes

    • This release adds new capabilities; existing flows remain unchanged.
  • fc1dca7: The creative field in the Facebook Marketing connector is no longer supported; use creative_id insteadSection titled “The creative field in the Facebook Marketing connector is no longer supported; use creative_id instead”

  • dc9b5ab: Implement OAuth2 flow for Facebook Marketing connectorSection titled “Implement OAuth2 flow for Facebook Marketing connector”

    • Add OAuth2 base oauth flow implementation
    • Add OAuth2 authentication option for Facebook Marketing connector
    • Implement token exchange and refresh logic
    • Store OAuth credentials securely in database
    • Add “Set up manually” option for manual token configuration
  • @owox/internal-helpers@0.13.0
  • @owox/idp-protocol@0.13.0
  • @owox/idp-better-auth@0.13.0
  • @owox/idp-owox@0.13.0
  • @owox/backend@0.13.0
  • @owox/web@0.13.0

OWOX Data Marts – v0.12.0

  • bd09d56: Enhanced Run History: Google Sheets Reports and Looker Studio Data FetchingSection titled “Enhanced Run History: Google Sheets Reports and Looker Studio Data Fetching”

    The Run History tab now displays all Data Mart runs in one place:

    • Google Sheets Export report runs and Looker Studio report runs are now tracked in Run History alongside Connector runs
    • Each run shows its title, connector logo or destination icon, start datetime, and trigger type (manual/scheduled)
    • Hover over start time to see detailed start/finish datetimes and execution duration
    • Added “Pending” status for queued operations
    • ☝️ All historical Runs before the update are considered manual
  • ba8ca14: Improve ConnectorEditForm with Auto-Save and Smarter DataMartDefinition HandlingSection titled “Improve ConnectorEditForm with Auto-Save and Smarter DataMartDefinition Handling”

    This update enhances the connector setup experience with automatic saving, improved validation, and better handling of connector configurations.

    • Added auto-saving for connector settings in the ConnectorEditForm
    • Introduced safeguards for unsaved changes to prevent data loss
    • Enabled auto-updates for DataMartDefinition upon form submission
    • Refactored related components for clearer validation and smoother configuration flow
  • 5c98ca4: Safer and Smoother Connector Editing ExperienceSection titled “Safer and Smoother Connector Editing Experience”

    We’ve made it easier — and safer — to edit your connector settings. Now, if you make changes and try to close the form before saving, you’ll see a confirmation dialog to prevent losing your work.

    We’ve also simplified how configuration details are managed and improved tooltips for better clarity — so you can focus on setting up your data connections with confidence and less friction.

  • e2da6ef: Facebook Connector: Added support for nested creative fields in ad-group endpointSection titled “Facebook Connector: Added support for nested creative fields in ad-group endpoint”

    New flat fields available:

    • creative_id - Unique ID for the ad creative
    • creative_name - Name of the ad creative
    • creative_url_tags - UTM parameters and URL tags
    • creative_object_story_spec - Object story spec with page_id and other details
    • creative_effective_object_story_id - Page post ID used in the ad
  • 66e494d: Fixed false error notification about actualizing schemaSection titled “Fixed false error notification about actualizing schema”

    When configuring the Connector-based Data Mart, attempts to update the table schema would cause users to receive an error message in the UI that was not actually an error. For the Connector-based Data Mart, the table and schema are created on first run, so attempting to update the schema before the first run would result in an error in the UI. Now updating schema trigger checks the Data Mart type and doesn’t try for these cases

  • 061b00c: Refactor connector execution architecture by removing the standalone @owox/connector-runner package and integrating its functionality directly into @owox/connectors packageSection titled “Refactor connector execution architecture by removing the standalone @owox/connector-runner package and integrating its functionality directly into @owox/connectors package”

    ⚠️ Breaking changes:

    • Removed @owox/connector-runner package entirely
    • Moved connector execution logic to @owox/connectors/src/connector-runner.js
    • Migrated DTOs to @owox/connectors/src/Core/Dto/

    Improvements:

    • Simplified dependency management by consolidating connector-related packages
    • Updated connector execution service to use new DTOs and exports from connectors package
    • Removed redundant GitHub workflows for connector-runner
    • Cleaned up repository structure
  • 961140b: Remove @kaciras/deasync and sync-request dependencies and migrate to async/awaitSection titled “Remove @kaciras/deasync and sync-request dependencies and migrate to async/await”

    This is a minor breaking change that removes the @kaciras/deasync and sync-request dependencies from connectors package and migrates all synchronous blocking code to modern async/await patterns.

    Changes:

    • Removed @kaciras/deasync dependency
    • Removed sync-request dependency
    • Removed Google Apps Script support - Only Node.js environment is now supported
    • Refactored EnvironmentAdapter into specialized utility classes:
      • HttpUtils - HTTP requests
      • DateUtils - Date formatting
      • AsyncUtils - Async delays
      • CryptoUtils - Cryptographic operations
      • FileUtils - File parsing and decompression
    • Removed ENVIRONMENT enum and environment detection logic
    • Updated connector documentation
  • 87aed3f: Show Connector State in Manual Run menuSection titled “Show Connector State in Manual Run menu”

    • In Manual Run → State Info for incremental runs, you can now view the Connector State.
    • For connectors with multiple configurations, the state is shown for each configuration with a “Created at” tooltip.
    • If no state is available, we show “No state available”.
    • The state is displayed as read-only JSON with a copy option.

    This helps you quickly understand where incremental loading will continue from and simplifies troubleshooting.

  • f97c4e7: Add new Facebook Marketing insights endpoints and improve Facebook field schema filteringSection titled “Add new Facebook Marketing insights endpoints and improve Facebook field schema filtering”

    Introduced several new Facebook Marketing API insights endpoints with specific breakdowns:

    • ad-account/insights-by-age-and-gender — provides age and gender breakdowns
    • ad-account/insights-by-device-platform — provides device platform breakdown
    • ad-account/insights-by-product-id — provides product ID breakdown
    • ad-account/insights-by-publisher-platform and ad-account/insights-by-publisher-platform-and-position — provide publisher platform and platform position breakdowns
    • ad-account/insights-by-region — provides region-level breakdown

    ⚠️ Breaking Changes The legacy ad-account/insights endpoint no longer supports breakdown fields.

    If your Data Mart previously used ad-account/insights with breakdowns (such as age, gender, country, device_platform, link_url_asset, product_id, publisher_platform, platform_position, or region), please migrate to the appropriate new endpoint:

    Breakdown TypeNew Endpoint
    Age / Genderad-account/insights-by-age-and-gender
    Countryad-account/insights-by-country
    Device Platformad-account/insights-by-device-platform
    Link URL Assetad-account/insights-by-link-url-asset
    Product IDad-account/insights-by-product-id
    Publisher Platform / Positionad-account/insights-by-publisher-platform-and-position
    Regionad-account/insights-by-region

    Recommendation: ☝️ Recreate your Data Mart using the correct endpoint to ensure compatibility with the latest Facebook Marketing API structure.

  • cd3bcd9: Hidden optional connector config knobsSection titled “Hidden optional connector config knobs”

    Marked shared connector config fields as either hidden manual backfill dates or “Advanced” tuning options so the UI only surfaces essential settings by default.

  • @owox/internal-helpers@0.12.0
  • @owox/idp-protocol@0.12.0
  • @owox/idp-better-auth@0.12.0
  • @owox/idp-owox@0.12.0
  • @owox/backend@0.12.0
  • @owox/web@0.12.0

OWOX Data Marts - v0.11.0

  • 7617b79: Enhanced Data Mart run history monitoring with automatic updatesSection titled “Enhanced Data Mart run history monitoring with automatic updates”

    Improved the overall experience when working with Data Mart by introducing automatic data refresh and better run handling:

    Automatic data updates:

    • Run history now automatically refreshes, keeping you informed about the latest execution status
    • Google Sheets reports automatically update with fresh data without manual page refresh
    • Auto-refresh can be toggled on/off in run history, with your preference saved for future sessions
    • Updates happen silently in the background without disrupting your work

    Better connector run experience:

    • Clear loading indicators when manually running connectors, with the ability to cancel
    • Improved error messages when attempting to run a connector that’s already in progress
    • No more technical error messages - you’ll see friendly notifications like “Connector is already running. Please wait until it finishes”

    These improvements ensure you always have up-to-date information about your Data Mart executions without needing to manually refresh the page.

  • 0a99a0b: Add ability to copy connector configuration from existing Data MartsSection titled “Add ability to copy connector configuration from existing Data Marts”

    Added a new feature that allows users to copy connector configuration settings from existing Data Marts when creating or editing connector-based Data Marts.

    • Copy configuration button: New dropdown menu in the connector configuration step that shows all Data Marts with the same connector type
    • Multi-configuration support: For Data Marts with multiple configurations, a nested menu allows selecting specific configuration
    • Configuration preview: Tooltip on each item shows required fields with masked secrets
    • Secure secret copying: Secrets are properly masked and merged from source on backend
  • 5cd552c: Improve Data Mart Creation Flow and Connector Editor ExperienceSection titled “Improve Data Mart Creation Flow and Connector Editor Experience”

    This update brings several enhancements to the Data Mart creation flow and connector-related components, improving UI consistency, usability, and workflow efficiency.

    Changes

    • Data Mart creation flow:
      • Added new icons for Facebook Ads, X Ads, and LinkedIn Ads
      • Updated Empty Data Marts state with options to create Data Marts in different modes
      • Improved DataMartDefinitionSettings to handle mode-based initialization
      • Enhanced CreateDataMartPage to set default titles based on selected mode
      • Added animations on the Empty Data Marts page for a smoother user experience
    • UX improvements: implemented auto-open logic for Connector Setup Sheet when selecting a definition type
    • UI improvements: updated theme handling in DataMartCodeEditor for consistent styling
  • c929eb0: Fix BigQuery data duplication with NULL in unique keysSection titled “Fix BigQuery data duplication with NULL in unique keys”

    Fixed MERGE query in BigQueryStorage to correctly handle NULL values in unique key columns using IS NOT DISTINCT FROM instead of =. This prevents duplicate records when fields like AssetGroupId are NULL.

  • ccb4fef: Fix Boolean type in Connector Data Mart configurationSection titled “Fix Boolean type in Connector Data Mart configuration”

    Fixed an issue with boolean field types in connector configurations when setting up data marts. This fix ensures that boolean fields in connector configurations are properly handled, making them interactive and displaying appropriate UI indicators regardless of their default values.

  • 6059657: Fixed connector configuration fields editingSection titled “Fixed connector configuration fields editing”

    Fixed an issue where connector configuration fields with default values were difficult to edit when setting up a data mart. Fields now properly handle user input and allow modification of default values.

  • 7617b79: Improved SQL validation flow in Data Marts to prevent timeout issuesSection titled “Improved SQL validation flow in Data Marts to prevent timeout issues”

    Previously, users were unable to save SQL queries in Data Marts when validation took longer than 30 seconds, causing timeout errors.

    This update resolves the issue by:

    • Made SQL validation asynchronous and non-blocking for saving SQL in Input Sources
    • SQL validation is no longer required for Publishing Data Marts

    Users can now save SQL queries regardless of validation time, improving the overall experience when working with complex queries or large datasets.

  • 7617b79: Fix messages for Data Mart publish buttonSection titled “Fix messages for Data Mart publish button”

    Improved the clarity of status messages displayed on the Data Mart publish button. You’ll now see more accurate and informative feedback when publishing your Data Marts.

  • f96f9aa: Add Google Ads connectorSection titled “Add Google Ads connector”

    Added new Google Ads connector with Service Account authentication

    Available data nodes:

    • campaigns, campaigns_stats - Campaign data
    • ad_groups, ad_groups_stats - Ad group data
    • ad_group_ads_stats - Ad performance data
    • keywords_stats - Keyword performance data
    • criterion - Criteria data
  • b11b726: Added support for oneOf fields with recursive secret maskingSection titled “Added support for oneOf fields with recursive secret masking”

    This release adds comprehensive support for oneOf configuration fields with nested secret handling. The connector secret service now recursively masks and merges secret fields within oneOf structures, ensuring sensitive data like API keys and tokens in nested authentication configurations are properly protected.

    New UI components include ButtonGroup for value-based selection and AppWizardCollapsible for expandable sections. Fixed an issue where the wrong oneOf variant was pre-selected when editing existing configurations.

    Added Advanced Fields section to the connector configuration form, allowing users to configure advanced settings for the connector.

  • 1b97886: Remove MaxFetchingDays parameter and improve incremental fetching logicSection titled “Remove MaxFetchingDays parameter and improve incremental fetching logic”

    Removed the MaxFetchingDays parameter from all data source connectors. The incremental data fetching now works as follows:

    • First run (no state): Data fetching starts from the 1st of the previous month
    • Subsequent runs: Data is fetched from the LastRequestedDate (with ReimportLookbackWindow applied) up to today
    • Manual backfill: Continues to work as before, fetching data for the specified date range
  • @owox/internal-helpers@0.11.0
  • @owox/idp-protocol@0.11.0
  • @owox/idp-better-auth@0.11.0
  • @owox/idp-owox@0.11.0
  • @owox/backend@0.11.0
  • @owox/web@0.11.0

OWOX Data Marts - v0.10.0

  • 7b8747c: Fix incremental state management for multiple connector configurationsSection titled “Fix incremental state management for multiple connector configurations”

    Fixed an issue where incremental updates only saved state for the last configuration when a Data Mart had 2+ connector configurations. Now each configuration’s state is tracked separately using its _id. Also enhanced logging with structured metadata (dataMartId, projectId, runId, configId).

    Changes:

    • Updated state structure to support array of states per configuration: {at, states: [{_id, state, at}]}
    • Modified ConnectorStateService to handle configId parameter for getting and updating state
    • Updated ConnectorExecutionService to extract and pass configId from configuration
    • Added database migration to transform existing state data from old to new format
    • Enhanced logging with structured metadata (dataMartId, projectId, runId, configId)
  • 526abdc: Improved Connector Setup and Usability EnhancementsSection titled “Improved Connector Setup and Usability Enhancements”

    • Simplified the connector setup flow with a cleaner layout and improved step structure
    • Added keyboard shortcuts for faster field selection in the Connector Editor (Command + Shift + A on macOS, Control + Shift + A on Windows)
    • Refined default titles and interface texts for better clarity
    • Adjusted side sheet layouts for more consistent visuals
  • 2898354: Improved environment variable loggingSection titled “Improved environment variable logging”

    • Reduced verbose logging from EnvManager that was confusing users with unnecessary technical details about environment variable processing.
    • Environment setup now shows only essential information instead of detailed variable counts and processing steps.
  • 3370b36: Added migration to rename Bing Ads connector to Microsoft AdsSection titled “Added migration to rename Bing Ads connector to Microsoft Ads”

    • Fixed an issue where Run History tab was not displaying history if the user previously used the Bing Ads connector.
  • Fixed Looker Studio Connector error with deleted Data Marts.

  • @owox/internal-helpers@0.10.0
  • @owox/idp-protocol@0.10.0
  • @owox/idp-better-auth@0.10.0
  • @owox/idp-owox@0.10.0
  • @owox/backend@0.10.0
  • @owox/web@0.10.0

OWOX Data Marts - v0.9.0

  • 701a05f: Add System Theme Option to User MenuSection titled “Add System Theme Option to User Menu”

    • Added System option to the theme switcher for automatic theme selection.
    • Enhanced UserMenu with theme selection and submenu support for better usability.
  • 54df91e: Convert boolean parameters to proper boolean typeSection titled “Convert boolean parameters to proper boolean type”

    Updated boolean configuration parameters to use proper boolean type instead of string or bool types:

    • ProcessShortLinks (FacebookMarketing): string → boolean
    • SandboxMode (TikTokAds): bool → boolean
    • IncludeDeleted (TikTokAds): bool → boolean
  • 8402b05: Add new CLI commands for database migrationsSection titled “Add new CLI commands for database migrations”

    • migrations up - run all pending migrations
    • migrations down - revert last migration
    • migrations status - migration’s status check
  • 8fffa5e: Mask connector secrets in UISection titled “Mask connector secrets in UI”

    • Secret fields in connector configuration are masked on the configuration page and in the Run History tab.
  • 0b0a8fb: Enhanced Connector Setup FlowSection titled “Enhanced Connector Setup Flow”

    • Improved structure with AppWizard components for a more consistent and flexible setup layout
    • Better usability across all setup steps
    • Refined accessibility and visual design throughout the connector editing interface
  • 32b0314: Enhanced connectors to support CreateEmptyTables configuration optionSection titled “Enhanced connectors to support CreateEmptyTables configuration option”

    • Now tables will be created even when no data is fetched, if the CreateEmptyTables parameter is set to “true”.
  • 8e673e9: Enhanced Google BigQuery Location OptionsSection titled “Enhanced Google BigQuery Location Options”

    • Updated location labels to include region codes alongside city names for better clarity (e.g., us-central1 (Iowa) instead of just Iowa).
    • Improved Combobox component with better search functionality using keywords and increased minimum width for better display of longer location names.
  • 43adfcb: Split Facebook Marketing insights endpoint into three separate endpointsSection titled “Split Facebook Marketing insights endpoint into three separate endpoints”

    • Split ad-account/insights into three endpoints: base insights, insights by country, and insights by link URL asset
    • Added ad-account/insights-by-country endpoint with country breakdown
    • Added ad-account/insights-by-link-url-asset endpoint with link_url_asset breakdown
    • Refactored insights data fetching to use object parameters and separate fields from breakdowns
    • ⚠️ Breaking Changes: ad-account/insights endpoint no longer supports breakdown fields
    • ⚠️ Breaking Changes: if your data mart was using ad-account/insights with breakdown fields (e.g., country, link_url_asset), you need to recreate it using the appropriate new endpoint:
      • Use ad-account/insights-by-country for country breakdown
      • Use ad-account/insights-by-link-url-asset for link URL asset breakdown
  • 646511d: Fix data mart run history time logsSection titled “Fix data mart run history time logs”

    • Fixed bad time in data mart run history logs. Now the time is displayed in the correct timezone.
  • 9773ba4: Improvements & Bug FixesSection titled “Improvements & Bug Fixes”

    This update includes general interface improvements, performance enhancements, and minor fixes to ensure a smoother and more reliable user experience.

  • 95dcaec: Intercom chat integrationSection titled “Intercom chat integration”

    💬 Intercom chat integration is now available in the Web app for faster support and onboarding.

  • @owox/internal-helpers@0.9.0
  • @owox/idp-protocol@0.9.0
  • @owox/idp-better-auth@0.9.0
  • @owox/idp-owox@0.9.0
  • @owox/backend@0.9.0
  • @owox/web@0.9.0

OWOX Data Marts - v0.8.0

  • 2932470: Better Auth: Primary Admin Setup & Password ResetSection titled “Better Auth: Primary Admin Setup & Password Reset”

    • Primary admin auto-creation: Configure IDP_BETTER_AUTH_PRIMARY_ADMIN_EMAIL to automatically create or manage primary admin on server startup
    • Password reset UI: Admins can reset user passwords through Admin Dashboard (/auth/dashboard) with automatic magic link generation
    • Enhanced documentation: Added comprehensive user management guide at /docs/getting-started/setup-guide/members-management/better-auth.md

    Features:

    • Auto-creates admin if doesn’t exist (generates magic link in logs)
    • Generates new magic link if admin exists without password
    • Password reset button for existing users with passwords
    • Magic link generation for users without passwords

    New Environment Variables:

    • IDP_BETTER_AUTH_PRIMARY_ADMIN_EMAIL – Email for automatic primary admin creation
  • 518cfe1: refactor: rename Bing Ads to Microsoft Ads and update documentation, images, and referencesSection titled “refactor: rename Bing Ads to Microsoft Ads and update documentation, images, and references”

  • 29f72ea: Enhance DataMartCreateForm with New Storage CreationSection titled “Enhance DataMartCreateForm with New Storage Creation”

    • Updated storage selection to allow creating new storage directly from the form.
    • Refined CreateDataMartPage styling for better visual consistency.
  • 099befb: fix: allow deleting a datamart within a project if it was created by another userSection titled “fix: allow deleting a datamart within a project if it was created by another user”

  • 25ab28e: fix: a user with the viewer role is not allowed to modify objects in the application with idp = better-authSection titled “fix: a user with the viewer role is not allowed to modify objects in the application with idp = better-auth”

  • edb4478: ✨ Google Tag Manager integration

    • 🚀 Added Google Tag Manager support across the web app. Enable by setting GOOGLE_TAG_MANAGER_CONTAINER_ID in your environment. This allows non‑technical teams to ship marketing/analytics tags without deployments.

    Why this matters

    • 📊 Faster iteration on analytics and marketing experiments (no code release required for common changes).
  • 19c21a1: ⚠️ Breaking Change: LinkedIn Authentication UpdateSection titled “⚠️ Breaking Change: LinkedIn Authentication Update”

    What changed: LinkedIn connectors now require 3 credentials instead of 1 Access Token: Client ID, Client Secret, and Refresh Token.

    What you need to do:

    1. Go to LinkedIn Developer Portal → your app → Auth tab
    2. Copy these 3 values:
      • Client ID (top of Auth page)
      • Client Secret (top of Auth page)
      • Refresh Token (generate via OAuth 2.0 tools)

    How to update:

    1. Go to your OWOX Data Marts
    2. Find your LinkedIn connector configuration
    3. Enter the 3 new credentials instead of the old Access Token:
      • Client ID
      • Client Secret
      • Refresh Token
  • b41b62d: Logging System Architecture RefactorSection titled “Logging System Architecture Refactor”

    • Refactored logging architecture: Extracted Pino logger creation from LoggerFactory into a provider-agnostic architecture while maintaining backward compatibility
    • Simplified configuration: Removed environment presets from LoggerConfig, now only LogLevel controls logging behavior
    • Environment variables update:
      • Changed from LOG_LEVELS (comma-separated) to LOG_LEVEL (threshold-based)
      • Updated .env.example with clear documentation
    • Enhanced TypeORM integration: Improved CustomDataSourceLogger with proper parameter usage and context formatting

    Breaking Changes:

    • LOG_LEVELS environment variable renamed to LOG_LEVEL
    • Removed environment field from LoggerConfig interface

    Migration:

    • Replace LOG_LEVELS=log,warn,error with LOG_LEVEL=info (threshold-based) or app will use default info level
    • Remove environment parameter from LoggerFactory calls
  • 8a1ef12: Secure MySQL connections (TLS/SSL)Section titled “Secure MySQL connections (TLS/SSL)”

    • New, simple way to enable encrypted MySQL connections via environment variables:
      • Backend (NestJS/TypeORM): DB_SSL
      • Identity provider (Better Auth): IDP_BETTER_AUTH_MYSQL_SSL

    Learn more

  • 32cd6c9: Revamp NotFound Page and Improve Mobile LayoutSection titled “Revamp NotFound Page and Improve Mobile Layout”

    • Redesigned 404 page with a new foreground card and animated background tunnel effect.
    • Updated styles for improved responsiveness and visual appeal.
    • Added icons and navigation button to guide users.
    • Improved mobile layout and updated SidebarTrigger icon for consistency.
  • e19073a: Refactor OpenHolidays connector according to common architecture and fix bugsSection titled “Refactor OpenHolidays connector according to common architecture and fix bugs”

  • 90a8711: Simplified MySQL configuration in the idp-better-authSection titled “Simplified MySQL configuration in the idp-better-auth”

    • idp-better-auth uses the DB_* environment variables unless IDP_BETTER_AUTH_MYSQL_* is specified.
  • fc17562: Enhance Error Handling and NotificationsSection titled “Enhance Error Handling and Notifications”

    • Enhanced API error handling and notifications across components.
    • Updated Toaster component styles for improved clarity and consistency.
    • Other UI improvements
  • af2e412: Updated Google BigQuery & Google Sheets authenticationSection titled “Updated Google BigQuery & Google Sheets authentication”

    • Switched to JWT-based auth client (google-auth-library).
    • Removed deprecated credential paths and warnings.
    • Improved reliability of loads/queries.
    • No action required — existing service account JSON keys continue to work.
  • 58e2ead: Updated Looker Studio data destinationSection titled “Updated Looker Studio data destination”

    • Clarify PUBLIC_ORIGIN: base public URL of the application (scheme + host [+ optional port]).
      • Examples: http://localhost:3000, https://data-marts.example.com
      • Default: http://localhost:${PORT}
      • In production, set this to your actual deployment URL.
    • Introduce LOOKER_STUDIO_DESTINATION_ORIGIN: public origin used to generate the deployment URL for the Looker Studio destination.
      • If empty, it falls back to PUBLIC_ORIGIN.
      • Example: https://looker.example.com

    Heads up

    • When retrieving the current JSON config for a Looker Studio Data Destination, the deploymentUrl field is now generated from LOOKER_STUDIO_DESTINATION_ORIGIN (fallback: PUBLIC_ORIGIN). If you previously set deploymentUrl manually during creation, it is now populated from the environment variable values.

    Learn more

  • @owox/internal-helpers@0.8.0
  • @owox/idp-protocol@0.8.0
  • @owox/idp-better-auth@0.8.0
  • @owox/idp-owox@0.8.0
  • @owox/backend@0.8.0
  • @owox/web@0.8.0

OWOX Data Marts - v0.7.0

  • @owox/backend@0.7.0
  • @owox/idp-protocol@0.7.0
  • @owox/idp-better-auth@0.7.0
  • @owox/idp-owox@0.7.0
  • @owox/internal-helpers@0.7.0
  • @owox/web@0.7.0

OWOX Data Marts – v0.6.0

  • Updated dependencies [4749749]
    • @owox/idp-owox@0.6.0
    • @owox/backend@0.6.0
    • @owox/idp-protocol@0.6.0
    • @owox/idp-better-auth@0.6.0
    • @owox/web@0.6.0
  • d129eb0: Triggers and reports columns available in the Data Marts listSection titled “Triggers and reports columns available in the Data Marts list”

    • Added columns for the number of triggers and reports to the Data Marts list
  • 6335c25: Fixed BingAds report data export and added proper field mappingSection titled “Fixed BingAds report data export and added proper field mapping”

    • Fixed data export issues in BingAds reports by separating into two report types with proper field schemas
    • Fixed issue where values were being saved with quotes in database
  • 2f2d4bf: Add manual backfill functionality for data mart connectorsSection titled “Add manual backfill functionality for data mart connectors”

    • Added support for manual connector runs with custom payload parameters
  • 0f590bb: Connector Target step: editable dataset/database and tableSection titled “Connector Target step: editable dataset/database and table”

    • Added editable dataset/database and table fields with sensible defaults
    • Defaults come from sanitized destination name: dataset/database ${sanitizedDestinationName}_owox, table ${sanitizedDestinationName}
    • Inline validation: required, only allowed characters, accessible error state
    • Helper text shows full path: {dataset}.{table}
  • db3a03a: Show Individual Destination Cards in Destination TabSection titled “Show Individual Destination Cards in Destination Tab”

    The Destination tab now displays a separate card for each specific destination in the project. Each card shows only the reports belonging to that destination, making it easier to find and manage reports at a glance.

  • 863ad3e: Enhanced Output Schema FormattingSection titled “Enhanced Output Schema Formatting”

    The Output Schema has received a major upgrade to improve control over data readability in Destinations.

    • Added support for column header descriptions as cell notes in the Google Sheets Destination, so you can define metrics everyone is aligned on
    • Implemented automatic formatting for BigQuery and Athena timestamp fields
    • Introduced the ability to control the order of fields delivered from Data Mart to Destination via simple drag & drop in the Output Schema
  • aac5411: Update API version and refactor insights data fetching logicSection titled “Update API version and refactor insights data fetching logic”

    • Updated the Facebook Graph API base URL to use version 23.0 directly in the code, removing the configurable ApiBaseUrl parameter.
    • Refactored the insights data fetching logic to pass the API base URL explicitly to helper methods.
    • Modified _fetchInsightsData and_buildInsightsUrl to accept and use the API base URL as a parameter.
    • Removed the unsupported activity_recency field from adAccountInsightsFields.
    • Improved code clarity and maintainability by simplifying how the API URL is constructed and used throughout the Facebook Marketing source integration.
  • b6cdb5a: TypeORM Entity Migration MechanismSection titled “TypeORM Entity Migration Mechanism”

    • Introduced an automatic migration system for TypeORM entities.
    • Ensures database schema stays up-to-date with entity definitions.
    • Runs migrations automatically on application startup—no manual steps required.
    • Prevents data loss and supports seamless schema evolution.
  • 66a6c38: Improving credentials management security for Data Storage and Data DestinationSection titled “Improving credentials management security for Data Storage and Data Destination”

    • API no longer returns credential secrets to the UI.
    • Credential secrets are no longer displayed in the UI.
    • Credentials are only updated if explicitly changed.
    • Added a link to manage Google Cloud Platform service accounts.
  • 6f772ee: Added Looker Studio Connector supportSection titled “Added Looker Studio Connector support”

  • e4e59f0: Remove unsupported fieldsSection titled “Remove unsupported fields”

    • Removed the following unsupported or deprecated fields from adAccountInsightsFields in the Facebook Marketing API reference:
      • age_targeting
      • estimated_ad_recall_rate_lower_bound
      • estimated_ad_recall_rate_upper_bound
      • estimated_ad_recallers_lower_bound
      • estimated_ad_recallers_upper_bound
      • gender_targeting
      • labels
      • location
    • Cleaned up the field definitions to avoid including unsupported fields for Facebook API v19.0 and above.
    • Improved maintainability and reduced the risk of API errors related to invalid fields.
  • f351f63: Hover Cards in Triggers List — Now Smarter and More VisualSection titled “Hover Cards in Triggers List — Now Smarter and More Visual”

    The Triggers list just got a big usability boost! Hover over any Report Run or Connector Run to instantly see key details — no extra clicks needed.

    • For Reports: name, last edit, run history, and 1-click access to Google Sheets.
    • For Connectors: source name, field count, run history, and direct Google BigQuery or AWS Athena link.

    Check status, spot issues, and jump to your data faster than ever — all right from the Triggers list.

  • 6e76c87: Implement column visibility and sorting persistenceSection titled “Implement column visibility and sorting persistence”

    Previously, user interface configurations such as selected columns in tables and accordion states were reset upon every page refresh. This change ensures that the system now remembers these chosen states at the browser level for:

    • Data Marts list
    • Storages list
    • Data Marts details (Destinations, Triggers, and Reports lists).
  • db0732e: Connector-Based Data Mart UX improvementsSection titled “Connector-Based Data Mart UX improvements”

    • Used connector-based data mart for data mart setup right destination name in Target Setup step.
    • Added in connector-based data mart inline validation for target dataset/database name in Target Setup step with accessible error state.
    • Enabled double-click on a connector card to select and advance to the next step.
    • Added field sorting controls in Fields Selection step:
      • A–Z, Z–A, and Original order
      • Unique key fields always appear at the top across all sorting modes
    • Minor UI polish: sort icon with dropdown next to search input; helpful link to open an issue from fields step.
    • Added helpful link to open an issue from nodes step.
  • 229c7a1: Updated connector configuration stepSection titled “Updated connector configuration step”

    • Added type to date fields.
    • Moved field descriptions to tooltips.
    • Used field labels as titles instead of field names.
  • @owox/backend@0.5.0
  • @owox/idp-protocol@0.5.0
  • ac64efd: # Add Data Mart Connector IconsSection titled “# Add Data Mart Connector Icons”

    Enhance data-mart with connetors:

    • add connector icons
    • can cancel connector run
    • add connector documentation link
  • ae26689: # Fixed unexpected behaviourSection titled “# Fixed unexpected behaviour”

    • 404 error after reloading page
    • error with crashing the react app
    • error with publishing connector-based data mart
  • 09aaade: # Add data mart run history feature that allows users to view and track execution history of their data martsSection titled “# Add data mart run history feature that allows users to view and track execution history of their data marts”

    This feature provides

    • New “Run History” tab in the data mart details view
    • Comprehensive run history display with pagination support
    • Real-time tracking of data mart execution status and results
    • Load more functionality for viewing extensive run history
    • Integration with existing data mart context and state management

    Additional improvements include:

    • Ability to edit source fields in already published connector-based data marts
    • Enhanced connector runner with better config handling for non-string values
    • Improved AWS Athena storage with optimized query execution and DDL handling
    • UI refinements including conditional chevron display in list item cards
    • Cleanup of unused connector-related code from data storage features

    This enhancement improves monitoring capabilities and gives users better visibility into their data mart execution patterns and performance.

  • ca4062c: # Add data mart schema management feature that allows users to view, edit, and manage the structure of their data martsSection titled “# Add data mart schema management feature that allows users to view, edit, and manage the structure of their data marts”

    This feature provides:

    • Visual schema editor for both BigQuery and Athena data marts
    • Ability to add, remove, and reorder fields in the schema
    • Support for defining field types, modes, and other properties
    • Schema validation to ensure compatibility with the underlying data storage
    • Ability to actualize schema from the data source to keep it in sync

    This enhancement gives users more control over their data mart structure and improves the data modeling experience.

  • 2b6e73d: # ✨ Add SQL validation for Data MartsSection titled “# ✨ Add SQL validation for Data Marts”

    Enhance your data mart experience with real-time SQL validation:

    • 🚀 Instant feedback on SQL query validity
    • ❌ Clear error messages when something goes wrong
    • 📊 Estimated data volume for successful queries
    • ⏱️ Automatic validation as you type

    This feature helps you write correct SQL queries with confidence, reducing errors and saving time when working with your data marts.

  • 6d97d91: # UX/UI ImprovementsSection titled “# UX/UI Improvements”

    Add Planned Data Storages with “Coming Soon” Status

    • Snowflake
    • Databricks
    • AWS Redshift
    • Azure Synapse

    UI Updates: Triggers Table and Reports Table

    • Minor UI updates to the Triggers Table
    • UI improvements to the Reports Table for consistency

    More Friendly and Consistent Forms

    We’ve improved the interface to make working with forms in OWOX Data Marts more intuitive and user-friendly.

    • Unified form layout. All forms — for Triggers, Reports, Storage, and Destinations — now share a consistent design. This makes it easier to navigate and work with confidence.
    • Helpful hints where you need them. Tooltips and inline descriptions have been added next to form fields, so you can better understand what’s expected without second-guessing.
    • Improved tooltip styling. Tooltips now feature a more noticeable background, making important information easier to spot.
    • Faster editing. You can now enter edit mode in the Storage and Destinations tables with a single click on a row — no need to hunt for buttons.
    • Warnings before leaving with unsaved changes. If you make changes to a Storage or Destination and try to leave without saving, you’ll see a confirmation dialog — helping prevent accidental data loss.

    Refined Data Mart Page: Layout, Menu, and Texts

    • Updated the layout of the Connector block
    • Polished the dropdown menu on the Data Mart page

    Redesigned “Create Data Mart” Page

    • The form on the Create Data Mart page has been updated for visual consistency and a better user experience.

    Extra Visual and Text Tweaks

    • We’ve also made a few small improvements to the UI and copy to make everything feel more polished and cohesive.
  • @owox/backend@0.4.0
  • 543f30d: ⏰ Time Triggers: Schedule Your Reports and ConnectorsSection titled “⏰ Time Triggers: Schedule Your Reports and Connectors”

    We’re excited to introduce Time Triggers - a powerful new feature that allows you to schedule your reports and connectors to run automatically at specified times!

    Benefits

    • ✅ Save Time: Automate routine data refreshes without manual intervention
    • 🔄 Stay Updated: Keep your data fresh with regular scheduled updates
    • 📊 Consistent Reporting: Ensure your reports are generated on a reliable schedule
    • 🌐 Timezone Support: Schedule based on your local timezone or any timezone you need
    • 🔧 Flexible Scheduling Options: Choose from daily, weekly, monthly, or interval-based schedules

    Scheduling Options

    • Daily: Run your reports or connectors at the same time every day
    • Weekly: Select specific days of the week for execution
    • Monthly: Schedule runs on specific days of the month
    • Interval: Set up recurring runs at regular intervals

    Now you can set up your data workflows to run exactly when you need them, ensuring your dashboards and reports always contain the most up-to-date information without manual intervention.

  • @owox/backend@0.3.0
  • 71294b2: 2 July 2025 demo
  • @owox/backend@0.2.0