RAGCore

A production-grade Retrieval-Augmented Generation (RAG) API built with FastAPI, LangChain, ChromaDB, and PostgreSQL. Supports multi-tenant document ingestion, semantic search, and conversational Q&A over uploaded documents.

Stack

Layer	Technology
API	FastAPI 0.115 + Uvicorn
LLM orchestration	langchain-core / langchain-community / langchain-openai
Vector store	ChromaDB (collection-per-tenant)
Embeddings	OpenAI text-embedding-3-small
Cache	Redis (semantic embedding cache)
Database	PostgreSQL + SQLAlchemy async
Migrations	Alembic

Quick Start

cp .env.example .env
# add your OPENAI_API_KEY

docker-compose up -d postgres redis chromadb
pip install -r requirements.txt
alembic upgrade head
uvicorn app.main:app --reload

API docs: http://localhost:8000/docs

API Endpoints

Documents

Method	Path	Description
POST	/api/v1/documents/ingest	Upload and ingest a PDF/text file
GET	/api/v1/documents	List all documents for the tenant
DELETE	/api/v1/documents/{id}	Delete document and its vectors

Queries

Method	Path	Description
POST	/api/v1/queries/ask	Ask a question (returns full answer + sources)
POST	/api/v1/queries/stream	Stream answer tokens via SSE

Collections

Method	Path	Description
GET	/api/v1/collections	List ChromaDB collections for tenant
DELETE	/api/v1/collections/{name}	Drop a collection

Authentication

All requests require an X-API-Key header. In production this maps to a tenant. For local dev any string ≥ 8 characters works; the first 8 chars become the tenant_id.

curl -H "X-API-Key: devkey01" http://localhost:8000/api/v1/documents

Project Structure

app/
  api/           # Route handlers (documents, queries, collections)
  core/          # Config, database, cache
  models/        # SQLAlchemy ORM models
  schemas/       # Pydantic request/response schemas
  services/      # Business logic
    document_processor.py   # PDF parsing + chunking
    embedding_service.py    # OpenAI embeddings + Redis cache
    vector_store.py         # ChromaDB operations
    rag_chain.py            # LangChain ConversationalRetrievalChain
    llm_service.py          # OpenAI chat completions + token counting
alembic/         # Database migrations
tests/           # Pytest test suite

Running Tests

pytest tests/ -v

Environment Variables

See .env.example for all supported variables.