AutoRAG Quick Start

Install and start the service through the Arango Control Plane. Configure your LLM and embedding providers at install time by passing them in the env object of the install request body. Token streaming is available with the OpenAI-compatible providers, openai and custom. With triton, the streaming endpoint still works but returns the full answer as a single chunk.

{
  "env": {
    "db_name": "your_database_name",
    "project_name": "your_project_name",
    "chat_api_provider": "openai",
    "chat_api_key": "your_openai_api_key",
    "embedding_api_provider": "openai",
    "embedding_api_key": "your_openai_api_key"
  }
}

Note the serviceIdPostfix from the response. For all provider options and parameters, see LLM Configuration.

Use Global Search (query_type: 1) for high-level themes, Local Search (query_type: 2) for specific entities, or Instant Search (query_type: 3) on the streaming endpoint for fast answers.

curl -X POST \
  https://<EXTERNAL_ENDPOINT>:8529/graphrag/retriever/{serviceIdPostfix}/v1/graphrag-query \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer <your-jwt-token>" \
  -d '{
    "query": "What are the main themes discussed in the document?",
    "query_type": 1,
    "level": 1,
    "include_metadata": true
  }'

The response contains the generated answer and, when include_metadata is true, the supporting references. Switch to /v1/graphrag-query-stream to receive tokens as they are generated.