bigRAG Python SDK

Async Python client for the bigRAG API.

pip install bigrag

import asyncio

from bigrag import BigRAG


async def main() -> None:
    async with BigRAG(api_key="bigrag_sk_...", base_url="http://localhost:4000") as client:
        doc = await client.documents.upload("docs", "/path/to/paper.pdf")
        result = await client.queries.query("docs", {"query": "What is RAG?"})
        print(doc["id"], result["total"])


asyncio.run(main())

The SDK is fully typed, ships py.typed, and uses CalVer releases like 2026.5.22.

Namespaces

• client.collections for collection CRUD, stats, re-embedding, event tokens, and event streams.
• client.documents for uploads, batch operations, file URLs, and status polling.
• client.queries for single, multi-collection, and batch retrieval queries.
• client.vectors for raw vector upsert and delete.
• client.webhooks for webhook management and delivery replay.
• client.auth for session login, setup, preferences, and identity.
• client.admin for users, API keys, access logs, audit logs, runtime settings, backups, vector storage overview, admin realtime streams, connector config, embedding presets, and MCP server keys.
• client.connectors.google for Google Drive account, file browsing, sources, and sync jobs.
• client.evaluations for golden-set retrieval evaluations.

Authentication

API-key endpoints accept api_key or the BIGRAG_API_KEY environment variable. Session-only admin endpoints can be used after calling client.auth.login(...) or client.auth.setup(...); the client keeps the session cookie in its underlying httpx.AsyncClient.

Versioning

Published artifacts use CalVer in the form YYYY.M.D.