GitHub RAG AI Assistant

A full-stack Retrieval-Augmented Generation (RAG) application that lets you chat with an AI about any GitHub developer's public repositories. Load a user's profile, and the system fetches, chunks, embeds, and indexes their repo data — then answers your questions using only that context.

---

✨ Features

• Profile Ingestion — Fetches all public repos + READMEs for any GitHub user via the GitHub API.
• RAG Pipeline — Chunks text, generates embeddings with all-MiniLM-L6-v2, and stores them in an in-memory FAISS index.
• Context-Aware Chat — Questions are answered strictly from retrieved repository context — no hallucination.
• Ollama-Powered LLM — Uses a locally-running Ollama model for answer generation (no cloud API keys needed).
• Modern Frontend — Clean, responsive chat UI built with vanilla HTML/CSS/JS.

---

🏗️ Architecture

frontend/          → Vanilla JS chat interface (served via static file server)
backend/
├── api/           → FastAPI route handlers (profile loading, Q&A)
├── config/        → App settings & environment config
├── github/        → GitHub API client & data parser
├── models/        → Pydantic request/response schemas
├── rag/           → RAG pipeline (chunker, embedder, retriever, LLM client, prompt builder)
├── utils/         → Logging utilities
└── vector/        → FAISS vector store (in-memory, per-user indexes)

---

🚀 Getting Started

Prerequisites

• Python 3.10+
• Ollama — Install from ollama.com and pull a model:

  ollama pull llama3

• GitHub Token (optional) — Increases API rate limits. Generate one at github.com/settings/tokens.

1. Clone & Install

git clone <your-repo-url>
cd github-chat

python3 -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt

2. Configure Environment

Create a .env file in the project root:

# GitHub (optional — increases rate limits)
GITHUB_TOKEN=ghp_your_token_here

# Ollama LLM Configuration
OPENAI_API_KEY=ollama
OPENAI_API_BASE=http://localhost:11434/v1
LLM_MODEL=llama3

Note: Ollama exposes an OpenAI-compatible API at http://localhost:11434/v1, so no code changes are required — just set the environment variables above.

3. Start the Backend

uvicorn backend.main:app --reload --port 8000

4. Start the Frontend

cd frontend
python3 -m http.server 3000

Open http://localhost:3000 in your browser.

---

💬 Usage

1. Enter a GitHub username in the sidebar and click Load Context.
2. Wait for the system to fetch repos, chunk text, and build the vector index.
3. Ask questions in the chat — the AI answers based only on the loaded repository data.

---

⚙️ Configuration

All settings are managed via environment variables or a .env file:

Variable	Default	Description
GITHUB_TOKEN	""	GitHub personal access token (optional)
OPENAI_API_KEY	"sk-placeholder"	API key (set to ollama for Ollama)
OPENAI_API_BASE	""	LLM API base URL (http://localhost:11434/v1 for Ollama)
LLM_MODEL	"gpt-3.5-turbo"	Model name (llama3, mistral, etc.)
EMBEDDING_MODEL	"all-MiniLM-L6-v2"	Sentence-transformer model for embeddings
CHUNK_SIZE	500	Characters per text chunk
CHUNK_OVERLAP	50	Overlap between consecutive chunks
TOP_K_RETRIEVAL	5	Number of context chunks to retrieve

---

🛠️ Tech Stack

Layer	Technology
Backend	FastAPI, Uvicorn
LLM	Ollama (OpenAI-compatible API)
Embeddings	Sentence-Transformers (all-MiniLM-L6-v2)
Vector Store	FAISS (in-memory)
Frontend	Vanilla HTML / CSS / JavaScript
Data Source	GitHub REST API

---

📄 License

This project is for educational and portfolio purposes.