Run a read-only SQL query

Executes a single read-only Postgres `SELECT` statement against a curated set of views scoped to this workspace. Call `GET /v1/sql/schema` first — it lists every queryable view (table) and column; unqualified names in your query (e.g. `FROM leads`) resolve against those views automatically. Only one top-level `SELECT` is allowed: no `INSERT`/`UPDATE`/`DELETE`/DDL, no `EXPLAIN`, no table-valued sources, no unbounded result constructors/collection aggregates, and no multiple statements — anything else is rejected with `400`. Results are capped at 10,000 rows or 8 MB of response body, whichever comes first (`truncated: true` when either cap is hit); the database query is stopped when that byte cap is reached, PostgreSQL temporary-file spill is capped at 64 MB per process, and a query is aborted after 30 seconds (`408`). A single row that alone exceeds the size budget is rejected with `400` rather than truncated — select fewer columns, or wrap large text in `left(col, 1000)`. Requires the `sql:query` permission and workspace enrollment in the SQL query private alpha (`403` otherwise, contact us to request access). Rate limited to 2 queries/second and one in-flight query per workspace (`429` when either limit is exceeded).

curl -X POST "https://example.com/sql/query" \  -H "Content-Type: application/json" \  -d '{    "query": "string"  }'

{  "columns": [    {      "name": "string",      "type": "string"    }  ],  "rows": [    {}  ],  "rowCount": 0,  "truncated": true,  "executionMs": 0}