bigquery.live Open demo
AI ANALYTICS WORKBENCH

Custom reports.
Built on your data.

Ask a business question. The agent finds the right Metabase query, reads your BigQuery data and assembles the answer in Python.

See the data flow

Analytics workbench / Product analytics

Sample data
INTERACTIVE DEMOThree reports · Sample data runs in your browser
01 / ARCHITECTURE

Follow the data.

BigQuery is the data source. Metabase gives agents a catalog of saved SQL. The BI layer turns returned rows into figures and explanations.

STEPCOMPONENTROLEEXAMPLE
  1. 01BigQuery

    Receive app events, website activity and transactions from server-side exports.

    product.events
  2. 02Metabase

    Keep approved SQL in parameterized saved questions the agent can reuse.

    question / 57
  3. 03OpenRouter

    Route the question to your chosen model to select a query and plan the output.

    question → report plan
  4. 04Python + BI

    Calculate derived metrics and render the right visualization for the result.

    rows → report
02 / SETUP

Connect your stack.

Bring the dataset, saved SQL collection and model API key. Configure credentials on the server.

  1. 01

    Export to a BigQuery dataset

    Point website or app exports at your Google Cloud project. Give reporting its own read identity.

  2. 02

    Select a Metabase collection

    Give the agent access to the saved questions it needs for reporting.

  3. 03

    Choose an OpenRouter model

    Set a model ID and API key in your backend to route report requests.

EXAMPLE CONFIGURATION
BIGQUERY_PROJECT=your-project
BIGQUERY_DATASET=product
METABASE_URL=https://bi.example.com
METABASE_COLLECTION_ID=42
METABASE_API_KEY=<server-secret>
OPENROUTER_API_KEY=<server-secret>
OPENROUTER_MODEL=<provider/model>