HackRF Conversational RAG

Lightweight FastAPI service that exposes a conversational retrieval‑augmented generation workflow for HackRF documents.

---

Project layout

• Service entrypoint: main.py – wires FastAPI routes to the RAG pipeline.
• RAG utilities: rag.py – loads documents, builds the Chroma vector store, and defines the conversational chain.
• SQLite session/message store: database.py – persists chat history for each session.
• API models: serializers.py – Pydantic schemas for every endpoint.
• Response formatting helper: utils.py.
• Dependencies: requirements.txt.
• Data sources live in documents and the persisted embeddings in chroma_db.

---

Installation

python -m venv .venv
. .venv/Scripts/activate  # or source .venv/bin/activate on Unix
pip install -r requirements.txt

Initialize the vector store once (rerun when documents change):

python -c "from rag import create_vector_store; create_vector_store('./documents')"

---

Running the API

uvicorn main:app --reload

Conversational Voice Agent + RAG

I have created a working example of a Livekit + RAG

python agent.py dev

• Now go to Livekit Playground to test it out.

Open http://localhost:8000/docs for interactive OpenAPI docs.