Document Chat (RAG)

Upload PDFs, ingest them into a local vector store, and chat with your documents using retrieval-augmented generation.

How RAG Works

Retrieval-Augmented Generation (RAG) combines the power of semantic search with LLM reasoning. Documents are chunked, embedded, and stored in a vector database. When you ask a question, the system retrieves the most relevant chunks and uses them to ground the AI's response.

Document Ingestion API

Use the REST API to upload documents and trigger the ingestion pipeline.

curl -X POST http://localhost:5000/api/v1/documents/upload \
  -F "file=@research_paper.pdf" \
  -F "metadata={\"title\":\"AI Research\",\"tags\":[\"ai\",\"ml\"]}"

# Response:
{
  "document_id": "doc_abc123",
  "chunks_created": 45,
  "status": "indexed"
}

Querying Documents

Ask natural language questions and get answers grounded in your uploaded documents with source citations.

{
  "query": "What are the key findings about neural networks?",
  "document_ids": ["doc_abc123"],
  "top_k": 5
}

# Response:
{
  "answer": "The key findings indicate that...",
  "sources": [
    { "chunk_id": "chunk_12", "page": 3, "relevance": 0.92 },
    { "chunk_id": "chunk_45", "page": 8, "relevance": 0.88 }
  ]
}