PrepGraph is an AI-powered RAG chatbot built with LangGraph, LangChain, FAISS, and Groq’s LLaMA-3.
It’s designed to understand your uploaded PDFs/PPTX, retrieve the most relevant context, and chat naturally — while remembering your past messages.

• Each user has their own persistent chat memory (SQLite-based)
• Remembers your last messages and maintains context automatically
• Memory stored locally — no external database required

• Combines BM25 (keyword search) + FAISS (semantic embeddings)
• Embeddings generated using SentenceTransformers
• Supports both PDFs and PPTX — ideal for lecture slides, project docs, or reports
• Intelligent subject detection (CN, DOS, SE, etc.) for improved retrieval accuracy

• Powered by Groq’s ultra-fast inference
• Context-aware and concise responses with safe fallback extraction
• Strictly follows provided document context to avoid hallucination

• Clean, responsive Gradio UI (minimal design)
• Automatically loads your previous conversations
• Includes “Send” + “Clear Chat” buttons for smooth interaction
• Works even if backend restarts — chat memory persists!

• Modular architecture: Retriever, Memory, and Graph layers separated
• Cached FAISS + BM25 indexes for lightning-fast reloads
• .env-driven configuration for flexible environment setup
• Structured logging and fault-tolerant persistence
• Fully compatible with Hugging Face Spaces deployment

User Query → Memory (SQLite)
          ↓
      Retriever Node
     (BM25 + FAISS Hybrid)
          ↓
     Context Chunk
          ↓
      LLaMA-3 (Groq)
          ↓
    Answer + Memory Save
          ↓
     Gradio Chat UI

├── chatbot_graph.py # Main Gradio app (UI + LangGraph) ├── chatbot_retriever.py # Hybrid retriever (FAISS + BM25) ├── memory_store.py # SQLite-based persistent memory ├── data/ # Upload PDFs / PPTX here ├── .env # API keys and environment configs └── requirements.txt

git clone https://github.com/07Codex07/PrepGraph.git cd PrepGraph

python -m venv myenv source myenv/bin/activate # or myenv\Scripts\activate (Windows)

pip install -r requirements.txt

echo "GROQ_API_KEY=your_groq_api_key_here" > .env

python chatbot_graph.py

🖼️ Screenshots

Chat UI Document Context Retrieval Response Generation

🚀 Project Status

• Backend: Fully functional — LangGraph, Groq, FAISS, BM25, and SQLite memory integrated.
• Frontend: In progress — UI work ongoing (will connect to backend via Gradio/Bolt).
• Deployment: Will be deployed once frontend workflow is ready.

Built by Vinayak Sahu