RAG FastAPI example

This repository contains a minimal FastAPI app with a simple Retrieval-Augmented Generation (RAG) flow.

What I added

• main.py - main FastAPI app. generate_ai_response now implements a local RAG flow using OpenAI embeddings + ChatCompletion when OPENAI_API_KEY is present.
• knowledge/ - folder to store .txt documents used as the knowledge base (one sample file is provided).
• .rag_index.json (created at runtime) caches document embeddings.
• requirements.txt lists the python dependencies.

Quickstart (macOS / zsh)

1. Create a virtualenv and install deps:

python -m venv .venv
source .venv/bin/activate
python -m pip install --upgrade pip
pip install -r requirements.txt

2. Create a .env file (or copy .env.example) and set OPENAI_API_KEY plus ERPNext vars:

cp .env.example .env
# edit .env and set OPENAI_API_KEY and ERPNext values

3. Add knowledge files (plain .txt) into the knowledge/ folder. A sample is included.

4. Run the app:

uvicorn main:app --reload

Notes

• If OPENAI_API_KEY isn't set or openai isn't installed, generate_ai_response falls back to a simulated response.
• For production, use a proper vector DB (Qdrant/Pinecone) and secure secrets management instead of .env.