query_data

Execute a read-only SQL query via DuckDB against data sources for a project. Supports two source types: (1) Uploaded files (upload_data) — table name is the source_name given at upload. (2) Google Sheet snapshots (connect_google_sheet) — each tab becomes its own table named {source_name}_{tab_name} where non-alphanumeric characters in the tab name are replaced with underscores (e.g. source ‘budget’, tab ‘Q1 Revenue’ → table ‘budget_Q1_Revenue’). FRESHNESS: Google Sheet results reflect the latest snapshot captured by connect_google_sheet. Call connect_google_sheet again to refresh before querying when current data is needed. The snapshot_taken_at field in the sources response shows when each sheet was last snapshotted. Only SELECT and WITH statements are accepted. Results are capped at row_limit rows (default 1 000, max 10 000). Returns: columns (list[str]) — ordered column names; rows (list[list]) — result rows as JSON-serialisable values; row_count (int) — number of rows returned; truncated (bool) — true when row_limit was reached; sources (list) — loaded sources with fields: source_name, source_type, tables (list of DuckDB table names), snapshot_taken_at (ISO 8601 UTC or null); warnings (list[str], optional) — present when Google Sheet tabs were skipped due to empty or malformed CSV snapshots.

Parameters

Name	Type	Required	Default	Description
project_name	str	Yes	—
sql	str	Yes	—
row_limit	int	No	1000

Returns

Returns a structured object. Top-level fields:

• columns (array<string>)
• rows (array<array<string>>)
• row_count (integer)
• truncated (boolean)
• sources (array<object>)

See the response example below for the full payload shape.

Example

Tool Call

{
  "name": "query_data",
  "arguments": {
    "project_name": "forecasts",
    "sql": "SELECT month, revenue FROM actuals ORDER BY month LIMIT 3",
    "row_limit": 1000
  }
}

Response

{
  "columns": [
    "month",
    "revenue"
  ],
  "rows": [
    [
      "2026-01-01",
      100.0
    ],
    [
      "2026-02-01",
      110.0
    ],
    [
      "2026-03-01",
      121.0
    ]
  ],
  "row_count": 3,
  "truncated": false,
  "sources": [
    {
      "source_name": "actuals",
      "source_type": "parquet",
      "tables": [
        "actuals"
      ],
      "snapshot_taken_at": null
    }
  ]
}

Notes

• Module: Data
• Requires authentication via API token or Auth0 JWT.