bigRAG

Open-source, self-hostable RAG platform. Upload documents, auto-chunk, embed, and search — all behind a simple REST API.

• Document ingestion — PDF, DOCX, PPTX, HTML, Markdown, images, and more via Docling
• S3 bucket ingestion — ingest from S3 or any S3-compatible service (MinIO, R2, Spaces, etc.), including public buckets
• Embedding providers — OpenAI and Cohere, configurable per collection
• Vector search — semantic, keyword, and hybrid search modes via Milvus
• Reranking — Cohere reranking for improved result relevance
• Multi-collection queries — search across collections in a single request
• Batch operations — bulk upload, delete, status checks, and queries
• Real-time progress — SSE streaming for document processing status
• Analytics — per-collection query analytics and platform-wide stats
• Webhooks — get notified when documents are processed
• Self-hostable — single docker compose up to run everything
• SDKs — TypeScript and Python clients

docker compose up -d

This starts bigRAG API, Postgres, Redis, and Milvus. Open http://localhost:6100/docs for the interactive API docs.

# Create a collection
curl -X POST http://localhost:6100/v1/collections \
  -H "Content-Type: application/json" \
  -d '{"name": "docs", "embedding_api_key": "sk-..."}'

# Upload a document
curl -X POST http://localhost:6100/v1/collections/docs/documents \
  -F "file=@paper.pdf"

# Ingest from a public S3 bucket
curl -X POST http://localhost:6100/v1/collections/docs/documents/s3 \
  -H "Content-Type: application/json" \
  -d '{"bucket": "indian-supreme-court-judgments", "prefix": "judgments/2025/", "region": "ap-south-1", "no_sign_request": true}'

# Query
curl -X POST http://localhost:6100/v1/collections/docs/query \
  -H "Content-Type: application/json" \
  -d '{"query": "What are the main findings?"}'

./dev.sh  # starts Postgres, Redis, Milvus, and the API with hot reload

docker pull yoginth/bigrag:latest

graph TD
    Client([Client]) -->|REST API| API[bigRAG API<br/>Python / FastAPI]

    API --> Collections[Collections]
    API --> Documents[Documents]
    API --> Query[Query]
    API --> Webhooks[Webhooks]

    Documents -->|store files| Storage[(Storage<br/>Local / S3)]
    Documents -->|enqueue| Redis[(Redis<br/>Job Queue)]
    Redis -->|process| Worker[Ingestion Worker]

    Worker -->|parse| Docling[Docling<br/>PDF, DOCX, HTML, Images]
    Worker -->|embed| Embedding[Embedding Provider<br/>OpenAI / Cohere]
    Worker -->|store vectors| Milvus[(Milvus<br/>Vector DB)]

    Query -->|search| Milvus
    Query -->|embed query| Embedding
    Query -->|rerank| Reranker[Cohere Rerank]

    Collections --> Postgres[(Postgres<br/>Metadata)]
    Documents --> Postgres
    Webhooks --> Postgres

Full interactive docs at /docs (Swagger UI) when running.

npm install @bigrag/client

import { BigRAG } from "@bigrag/client";

const client = new BigRAG({ apiKey: "your-key", baseUrl: "http://localhost:6100" });

// Upload a document
const doc = await client.uploadDocument("docs", new File([pdf], "paper.pdf"));

// Stream processing progress
for await (const event of client.streamDocumentProgress("docs", doc.id)) {
  console.log(event.step, event.progress);
}

// Query
const { results } = await client.query("docs", { query: "What is RAG?" });

// Ingest from S3
await client.documents.ingestS3("docs", {
  bucket: "my-bucket",
  prefix: "reports/",
  no_sign_request: true,
});

pip install bigrag

from bigrag import BigRAG

client = BigRAG(api_key="your-key", base_url="http://localhost:6100")

# Upload a document
doc = await client.documents.upload("docs", "/path/to/paper.pdf")

# Query
result = await client.queries.query("docs", {"query": "What is RAG?"})

# Ingest from S3
result = await client.documents.ingest_s3(
    "docs",
    bucket="my-bucket",
    prefix="reports/",
    no_sign_request=True,
)

All settings use the BIGRAG_ prefix as environment variables, or configure via bigrag.toml:

PDF, DOCX, PPTX, XLSX, HTML, Markdown, CSV, TSV, XML, JSON, PNG, JPG, TIFF, BMP, GIF — powered by Docling with OCR support for scanned documents and images.

See CONTRIBUTING.md for development setup and guidelines.

If bigRAG is useful to you, consider sponsoring the project.

MIT