Data Talks
    Integration

    Talk to BigQuery in plain English

    Data Talks turns Google BigQuery into a conversational data warehouse. Connect a dataset, ask a question, get a chart — no Standard SQL required, no semantic layer to maintain.

    Get Started

    BigQuery is fast, cheap, and serverless — but writing the SQL still gets in the way for non-engineers. Data Talks puts a natural language layer in front of any BigQuery dataset. The AI agent introspects your tables, infers JOINs from column naming, and generates Standard SQL that runs against your project.

    You stay in control: every answer ships with the raw query the agent ran, so analysts can audit, save, or copy it into their own pipelines. Costs stay predictable because Data Talks lets you cap dataset scan size per query and cache results.

    Setup

    Connect your BigQuery project in 2 minutes

    1. 1

      Create a service account in Google Cloud with BigQuery Data Viewer + Job User roles.

    2. 2

      Download the JSON key and add it as a new source in Data Talks.

    3. 3

      Pick the project and dataset. Schema is detected automatically.

    4. 4

      Start asking questions in any conversation.

    Examples

    Questions you can ask out of the box

    • › Show monthly active users by region for the last 6 months.
    • › Which 10 customers had the largest order spike vs their 90-day average?
    • › Compare this week’s pipeline to the same week last quarter.
    • › List all events with non-null user_id where session length > 30 min.
    Highlights

    Why teams pick Data Talks for BigQuery

    Standard SQL generation with dialect-aware syntax (no GoogleSQL surprises).

    Auto-detected schema and JOIN inference, configurable for star/snowflake schemas.

    Per-query scan cap to keep BigQuery slot costs predictable.

    Inspect raw SQL behind every answer — audit, edit, or save it.

    Recurring alerts on any natural-language question, delivered to webhooks or Telegram.

    Self-hosted: your service account key never leaves your infra.

    FAQ

    BigQuery FAQ

    Does it support partitioned and clustered tables?

    Yes. Data Talks reads partition and clustering metadata when generating SQL so queries respect partition pruning and avoid full-table scans.

    Can I query multiple BigQuery projects?

    Yes. Add each project as a separate source. The AI agent can be scoped to one or many sources per conversation.

    What about cost control?

    Set a max bytes scanned per query at the source level. Queries that exceed it are blocked before they run.

    Does the data leave BigQuery?

    Only the result rows the agent fetches to render charts. The dataset itself stays in BigQuery.

    Connect your BigQuery dataset today

    Open source, self-hosted, free. Run Data Talks against BigQuery in your own infrastructure.

    Get Started