Register a data source — profile, index, detect joins to all existing sources

A data source to map. Every format normalizes to list[dict] internally.

Inline (no external dependency): records — JSON array of objects (fastest) csv — raw CSV text json_str — raw JSON array/object text (nested JSON is auto-flattened) url — HTTP(S) URL; format auto-detected from Content-Type / extension

Binary files (base64-encoded for API transport): excel_b64 — Excel .xlsx file, base64-encoded parquet_b64— Parquet file, base64-encoded

Connector config (routes through ConnectorRegistry): connection — dict with type + connector-specific keys: {"type": "sql", "connection_string": "postgresql+asyncpg://...", "query": "SELECT ..."} {"type": "sql", "connection_string": "sqlite+aiosqlite:///db.sqlite", "query": "..."} {"type": "rest_api", "url": "https://...", "data_path": "data.items"} {"type": "s3", "bucket": "my-bucket", "key": "data/sales.parquet"} {"type": "gcs", "bucket": "my-bucket", "blob": "data/orders.csv"} {"type": "azure_blob", "connection_string": "DefaultEndpointsProtocol=...", "container": "data", "blob": "sales.parquet"}