🔍 OpenAI-Powered RAG (Retrieval-Augmented Generation)

A production-ready RAG system built with:

• 💡 OpenAI for embeddings and completions
• 📦 ChromaDB as the local vector store
• ⚡ FastAPI for serving as an API
• 🧱 Modular code structure with utilities, configs, and logging

---

📁 Project Structure

rag-gemini-app/
├── app/                  # Core logic: ingestion, retrieval, generation
│   ├── ingest.py
│   ├── retriever.py
│   ├── generator.py
│   ├── rag_pipeline.py
│   ├── utils.py
│   └── __init__.py
├── api/                  # FastAPI server
│   ├── main.py
│   ├── routes.py
│   └── schemas.py
├── data/                 # Your input text files
│   └── documents/
├── vector_store/         # Chroma persistence
├── .env                  # API keys and environment variables
├── .gitignore
├── requirements.txt
├── run.py                # CLI interface for RAG
└── README.md

---

🚀 Quick Start

1. 🔧 Setup Environment

git clone https://github.com/yourusername/rag-gemini-app.git
cd rag-gemini-app

python -m venv venv
source venv/bin/activate        # or venv\Scripts\activate on Windows

pip install -r requirements.txt

2. 📄 Configure .env

OPENAI_API_KEY=your-openai-key-here

3. 📥 Ingest Documents

Place your .txt files into data/documents/, then run:

python app/ingest.py

4. 🧠 Ask via CLI

python run.py

5. 🌐 Run the API Server

uvicorn api.main:app --reload

Then open: http://localhost:8000/docs

---

🧠 How It Works

1. Ingestion:

   • Text files → Chunked → Embedded using OpenAI
   • Stored in ChromaDB with metadata

2. Retrieval:

   • Query embedded → Similarity search via Chroma

3. Generation:

   • Top-k chunks + Query → Prompt sent to OpenAI ChatCompletion

---

📦 Requirements

• Python 3.9+
• OpenAI API Key
• Internet access for embedding & LLM

---

📌 TODOs

• Improve chunking with NLP (spaCy, LangChain)
• Add document upload via API
• Add /ingest and /health endpoints
• Optional: Swap ChromaDB for FAISS or Qdrant in production

---

🛡 License

MIT

---

🙋‍♂️ Maintainer

Made with ❤️ by Jenish Thapa