Multi tenant RAG.

Small RAG (Retrieval-Augmented Generation) reference app demonstrating:

• document ingestion (PDFs)
• embeddings (local / cloud providers)
• vector store (Qdrant)
• a simple chat/query API and a web demo

Repository layout

• app/ — main application code (API, chat service, ingestion pipeline, embedding factories)
• data/ and data_2/ — sample PDFs used for ingestion
• web/index.html — minimal demo UI
• docker-compose.yml / Dockerfile — containerized run options
• populate_database.py — run ingestion to populate vector store
• query_data.py / get_embedding_function.py — helpers and examples
• test_api_flow.py / test_rag.py — test examples

Quickstart (local, Python)

1. Create virtualenv and install deps

# Bash
python -m venv .venv
.venv/Scripts/activate    # Windows: .venv\Scripts\activate
pip install --upgrade pip
pip install -r requirements.txt

2. Ingest sample documents

python populate_database.py

3. Run the API

python app/main.py
# or use your preferred ASGI server:
# uvicorn app.main:app --reload

4. Query the running service (example)

• Check app/main.py for the exact route path used by this repo. A generic POST example:

# Bash
curl -X POST http://localhost:8000/chat \
  -H "Content-Type: application/json" \
  -d '{"query":"What is invoice INV-2023-001 about?"}'

• The web demo: open web/index.html (or point it at the running API as configured).

Docker / Docker Compose

• To run with Docker Compose:

docker compose up --build

• Configure provider and models via environment variables (docker-compose.yml or .env).

Important environment variables (examples)

• PROVIDER_EMBED — provider for embeddings (e.g., openai, ollama)
• PROVIDER_CHAT — provider for chat/completion
• EMBED_MODEL — embedding model name (e.g., text-embedding-3-small, nomic-embed-text)
• CHAT_MODEL — chat model name (e.g., gpt-4o, llama2)
• OLLAMA_BASE_URL — if using a local Ollama server (e.g., http://host.docker.internal:11434)
• OPENAI_API_KEY — OpenAI key (if using OpenAI)
• QDRANT_URL / QDRANT_API_KEY — Qdrant connection details

Example .env snippet

PROVIDER_EMBED=openai
PROVIDER_CHAT=openai
EMBED_MODEL=text-embedding-3-small
CHAT_MODEL=gpt-4o
OPENAI_API_KEY=sk-...
QDRANT_URL=http://localhost:6333

Testing

pytest -q

Troubleshooting

• ModuleNotFoundError: No module named 'langchain.document_loaders'
  • Cause: mismatched langchain package version vs code import paths.
  • Options:
    1. Pin to the known-working langchain version used by this repo (if the code expects 0.3.x):

pip install "langchain==0.3.14" "langchain-core==0.3.35"

2) Or update imports in ingestion code to match langchain 1.0+ layout (code changes required).

• Create and use a virtualenv to avoid system-wide package conflicts.
• If pip reports dependency resolver conflicts after upgrading, prefer creating a fresh venv and installing requirements.txt.

Notes about providers and compatibility

• Some third-party packages in the ecosystem require older langchain-core versions. If you need legacy compatibility, pin package versions in requirements.txt.
• If you want Ollama locally:
  • Start Ollama and set OLLAMA_BASE_URL.
  • Set PROVIDER_EMBED=ollama and EMBED_MODEL to the Ollama embedding model name.
• For hybrid setups (embed with local Ollama, chat with OpenAI), set PROVIDER_EMBED and PROVIDER_CHAT independently.

Contributing

• Open issues or PRs for improvements.
• Keep changes minimal and include tests where applicable.

License

• Default: MIT (update if you prefer another license)

Contact / References

• See app/ for implementation details and docker-compose.yml for environment variable examples.