BigQuery Full Resync with Warehouse Usage

BigQuery Full Resync now refreshes both schema and warehouse usage together in a single operation. When usage-history access requires additional permissions, you receive clear guidance on how to restore access without the connection being marked as broken.

Scheduled BigQuery Usage Intelligence

BigQuery usage intelligence can now refresh automatically on a schedule and report setup issues independently. This means your warehouse usage data stays current without requiring manual intervention, and permission issues are surfaced clearly so you can address them at your convenience.

Improvements

• Browser notifications now include a preview of Orion's response so you can see what was said while away from the tab.

Split View for Dashboard and Notebook with Inspect

You can now open a dashboard and the analysis notebook side by side, then use Inspect to click a tile and see which query it came from and how it was rolled up. This makes it easy to understand the mechanics behind any dashboard metric without switching between windows.

Auto-Updating Dashboard Commentary

Dashboard text cards now rewrite automatically whenever you rerun the underlying notebook, keeping your narrative commentary synchronized with the fresh data. Older dashboards with hardcoded numbers self-heal on the next refresh, upgrading their text to match the live numbers so your commentary always stays accurate.

Improvements

• Dashboards can now use Google Fonts for custom typography, enabling richer visual design.
• Assumptions in the Context panel now display in full without truncation, so you can see complete methodology notes at a glance.
• Forecast recommendations now require sufficient history and a measurable baseline, reducing misleading work suggestions.

Edit and Re-run Data Extraction Queries

You can now edit the SQL behind a data extraction directly in the Analysis Notebook panel. Orion reviews the change, pulls the data again, and re-runs your notebook so every chart and number below reflects the new query. Perfect for refining a query on the fly without leaving your analysis.

Bridged Slack Conversations Stay in Sync

When you ask Orion a question from the web and it has a linked Slack thread, Orion's answer now appears in that thread as well. Both sides of the conversation read the same messages in the same order, so no one misses context.

Notification Delivery Visibility

Workflow runs now show each place a notification was sent and whether it arrived. A message that failed to reach one Slack channel is visible instead of hidden behind a single status, making it easy to spot delivery issues at a glance.

Improvements

• Chat is significantly more responsive in long conversations and during notebook analysis, with virtual scrolling that renders only visible portions of the transcript.

• Notebook cell output no longer reprocesses repeatedly while a turn is streaming, and tables no longer rebuild themselves on every update.

• Workflow diagrams now display Slack and webhook delivery channels instead of showing "No recipients".

• Chat no longer stops unexpectedly when Orion deploys during a response; partial answers are preserved with a note inviting you to continue.

• Conversations that stopped responding after Orion looked up additional tools mid-answer now work again without any action from you.

• Live views reconnect immediately during deploys instead of waiting up to 30 seconds, and deploys finish faster overall.

• Questions about Delta Lake data no longer stall; you can now explore values, check date ranges, and run ad-hoc queries against Delta Lake sources reliably.

• BigQuery datasources whose tables live in a different project from the one they bill to now work correctly instead of failing with permissions errors.

• Orion now incorporates more of the surrounding channel conversation when answering in Slack threads.

• Fixed an issue where two people replying in the same Slack thread could be answered by different Orion workspaces; now resolves consistently to the correct workspace.

• Memory saving now works again after being broken in the last release.

Improvements

• Replies in Orion Slack threads now join the same room you see on the web, keeping conversations unified across platforms.
• Messages sent in an Orion room now appear in its linked Slack thread, enabling seamless two-way communication.

Improvements

• Typing in the chat box is now more responsive, particularly when composing long messages or using @-mentions.
• Tables in chat responses no longer flicker while streaming, providing a smoother viewing experience.
• Chat is significantly more responsive during notebook analysis sessions.

Conversations as a Deliverable

When you open a discussion room from a workflow's published report, you now land directly on the analysis with the chat beside it. You can see who else is in the room and address Orion directly within the conversation, making it easy to ask follow-up questions and explore the data together without switching between windows.

Improvements

• Slack workflow notifications now include full context about why they were sent, which workflow and run produced them, and which reports they include.
• On Snowflake, tables shown as enabled in a project's schema settings are no longer refused at query time due to case-sensitivity mismatches.

Shareable Portals

Share your dashboards and reports outside Orion in three ways: a public link that anyone with the URL can access, a protected link secured by a password or sign-in through your identity provider, and an embed that renders Orion content directly inside your own website.

Each link can be given an expiry date and revoked individually, and the Share button shows you every way your content is currently reachable.

Improvements

• Knowledge base pages you lock or mark human-only are no longer overwritten by Orion's nightly knowledge base updates.
• After you connect or re-sync a LookML repository, table descriptions and join paths reach Orion immediately instead of waiting for the next daily sync.

Improvements

• 3x faster response times for chat conversations in the published Dashboard chat panel

dbt Descriptions in Schema Explorer

Table and column descriptions from your dbt project now appear directly in the schema explorer, labeled dbt and ranked above warehouse-native descriptions because these are the definitions Orion's agents read when answering your questions.

Interactive Count Tiles on LookML Panel

Count tiles on a data source's LookML panel are now interactive. Select any count to see exactly which views, dimensions, measures, or descriptions are behind it, with anything that did not come across listed first and the findings that explain each gap in the same place.

LookML Joins in Schema Tab

The Schema tab now shows join paths taken from a data source's own LookML for warehouses with a LookML repository connected. Orion uses these declared joins when writing SQL, so it no longer has to guess how tables relate on a warehouse that declares no foreign keys. Each path is marked "declared" because the warehouse has not verified them.

Full Augmentation Inventory

The Augmentation tab now shows exactly what Orion extracted from a connected LookML repository: how many views, dimensions, measures, and descriptions it could match to real warehouse tables, and which ones it could not. Anything Orion cannot match is listed with the file it came from, so you can spot gaps instead of discovering them in an answer. Orion never modifies your LookML.

Column Descriptions from LookML

Column descriptions written in a data source's LookML now appear in Orion for warehouses with a LookML repository connected. Orion uses them when writing SQL, and warehouse-provided descriptions are never replaced.

Improvements

• Context panel now cites dbt and LookML description sources and quotes the descriptions Orion used when answering
• Context panel now shows which file LookML or dbt join definitions came from, displaying the join both ways: the SQL Orion ran and the definition in your file, with direct links to the file
• dbt metadata tools now report what they read to the context panel when Orion explores your dbt project's models, columns, lineage, or declared foreign keys
• Orion now recognizes which columns your LookML declares as primary keys and tracks whether a column description came from LookML or the warehouse
• Long-running conversations and rooms compact context more reliably

Improvements

• Orion now recognizes foreign key relationships declared in your dbt project through relationships tests, using them to determine the correct join paths between tables.

Behavioral Analytics Reports in the Knowledge Base

Comprehensive behavioral analytics reports for your Looker models now automatically appear in your knowledge base under Internal and stay up to date after every nightly analysis.

These reports include executive summaries, dashboard analysis, user activity, field catalogs, data quality assessments, usage patterns, proactive intelligence, and user personas—giving you instant access to rich context about your data without manual compilation.

Connect LookML Repositories Directly to Warehouse Data Sources

You can now connect a LookML repository to warehouse data sources such as Snowflake, BigQuery, or Postgres, not only to Looker connections. If your business logic lives in LookML but Orion queries your warehouse directly, that logic can now reach Orion without a Looker connection in between.

You can also copy an existing LookML repository configuration from a Looker connection to a warehouse data source instead of entering the repository and token a second time.

Improvements

• Notebook cell edits requested in chat are now reliably applied instead of being replaced with a review of the change.
• Orion now provides the standard OAuth resource-metadata endpoint that MCP clients look for, eliminating extra failed requests during connector discovery.

Metrics Catalog in Project Settings

You can now add catalog metrics to a project directly from the project's own settings. Open Project Settings, go to the Knowledge Base tab, and the Metrics Catalog appears at the top of the list alongside your wiki pages. Select metrics and folders with checkboxes instead of managing them one at a time from the Knowledge Base interface.

LookML File Browser and Actionable Sync Errors

When working with LookML integrations, you can now browse the exact files Orion pulled from your GitHub repository in the Augmentation tab, so you know precisely what Orion is reading. When a sync fails, Orion now tells you what went wrong and which specific setting to change, instead of reporting a generic error message.

End-to-End Metrics Catalog Building

Orion can now complete your entire metrics catalog workflow without handoff. It creates the folder structure, files metrics into the right folders as it promotes them, reorganizes the catalog for clarity, and writes knowledge base pages that explain how your metrics fit together. What previously required you to finish the organization and documentation by hand is now fully automated.

Improvements

• Dashboard edits now detect conflicting updates, preventing newer changes from being accidentally overwritten when multiple people edit simultaneously.

Metric Ownership and Deliverable Lineage

You can now record who is responsible for each metric in your Metrics Catalog as people or teams. When you're about to edit a shared metric, Orion shows you every published report that depends on it, so definition changes become conversations with stakeholders instead of surprises.

Improvements

• Commenting on interactive dashboard elements is now more responsive, especially on complex dashboards with many charts and controls.
• Orion more persistently validates chart type changes and refuses unsupported formats, then confirms its own edits by reading back current metric settings.
• Orion displays metric owner information and shows which published reports will be affected by a metric definition change before the change is made.
• Workflows built from Looker queries, weather lookups, or news searches now carry full definition information, preventing scheduled run failures.
• Shared Orion links through the Claude integration now resolve correctly, and slide decks and podcasts produced by workflows are now accessible through shared links.

MotherDuck Warehouse Connection

You can now connect Orion to MotherDuck and analyze your data the same way you would with any other warehouse.

Add your MotherDuck database using a token, and Orion automatically reads your schema, answers questions about your data in chat, and pulls tables into your notebook for analysis. Read Scaling tokens are supported and recommended, and your table and column comments are automatically used by Orion to understand which tables are relevant to your questions.

Schema Diagram View

The Schema tab now displays your data source's table relationships as an interactive diagram, letting you see the shape of your schema at a glance. Relationship lines are color-coded to show their origin, and dashed lines indicate relationships your warehouse declares but does not enforce.

Intelligent Table Relationships

Orion now understands how your tables relate to each other by reading the foreign keys your warehouse declares and any dbt relationship tests you have written.

When Orion writes queries, it uses these verified paths. The Schema tab shows you which joins are verified and which are not, and you can correct relationships Orion misidentified or add ones your warehouse never declared.

Improvements

• The conversation Context panel now displays the data sources, knowledge base pages, and metrics Orion started with, not just content pulled in mid-conversation

Improvements

• In a shared room, Orion's answer now stays where it was asked—messages people send while Orion is still writing appear below the answer instead of pushing it down, so a conversation reads in the order it happened.
• Slide Decks built from saved templates now fill in placeholder company and metric names with the real ones, so finished presentations no longer contain template text like "[Company Name]".

Unique Value Counts in Dashboard Headers

Dashboard headers can now display a count of unique values—such as unique sessions, unique customers, or unique users—instead of only showing a row count. This gives you more flexibility in summarizing what your data represents at a glance.

Chat Annotations for Direct Input

Orion can now raise questions or suggest analyses directly in chat when it needs your input on something. These annotation cards appear alongside the reply and persist in the conversation, so you can see them again when you return to the chat.

Improvements

• Dashboard metrics that divide a value by itself now display a "verify" badge to flag results that may need review, instead of silently showing 1.
• Dashboard metrics no longer show placeholder values when cost or credit columns contain small quantities.
• Charts covering several years of daily data now render successfully instead of reporting too many groups.
• Chat responses now stream properly in shared rooms when the response involves tool use and other participants send messages concurrently.

Improvements

• PDF uploads now clearly explain why a file was rejected and what to do about it, with immediate feedback for scanned documents with no readable text, password-protected files, and damaged files
• Slide decks built from saved templates now preserve the template's logo and background images

Analysis Recommendations with Intake Conversations

Orion now suggests analyses worth running directly in your conversations, surfacing opportunities at the moment you're already discussing the relevant topic.

When you click through on a recommendation, a short intake conversation opens to clarify exactly what you want—whether that's a slide deck, dashboard, metrics report, or something else—before any analysis work begins. This ensures the final result matches what you asked for rather than a generic guess.

Improvements

• Gemini Enterprise users can now ask Orion about their workflows and receive accurate answers about which workflows exist, when they last ran, and what each run produced.

Editable dbt Configuration

You can now edit a data source's dbt configuration after it is created. Update dbt Cloud, GitHub, or uploaded-manifest settings and rotate dbt credentials directly from the data source's Connection tab, without deleting and recreating the connection.

Per-Datasource Schema Auto-Enable Control

Data source admins can now control, per data source, whether newly discovered tables and explores are enabled or disabled by default. This gives you fine-grained control over which new schema elements appear in your workspace without requiring manual review of every discovery.

Improvements

• Resolved bug that affected dashboard dropdown rendering
• Fixed dashboard data pipeline to improve reliability

Workspace Panel: See the Sources Behind Every Answer

Open the Workspace panel (or click "Sources for this reply" under any answer) to see exactly which knowledge base pages, memory, metrics, and project settings Orion used to reach that result.

Each source is linked directly so you can open a Knowledge Base page to verify or edit it. This makes it much easier to understand and trust how Orion reached a conclusion.

Improvements

• Dashboards no longer display partial query results as complete data; when a result exceeds the calculation limit, you are prompted to narrow the data with a filter instead.