PDFMate.Ai

Your Personalised Ai powered PDF buddy !!

Django-based RAG app to upload PDFs, index them into Pinecone using embeddings from Hugging Face and use ollama distilled model for RAG inference, and answer questions with Ollama zephyr:7b-alpha.

Features

• Upload PDF and auto-split into chunks
• Embeddings via all-mpnet-base-v2
• Pinecone v3 serverless vector store
• RAG inference via local Ollama zephyr-7b-alpha (distilled model)
• Evidence contexts

Setup

1. Python 3.10+ and PyTorch.
2. Install Ollama and pull model:
   • ollama pull zephyr:7b-alpha
3. Create and fill .env (see .env.example).
4. Install dependencies:
   • pip install -r requirements.txt
5. Django init:
   • python manage.py migrate
   • python manage.py runserver
6. Open http://127.0.0.1:8000

Environment Variables

See .env.example for all settings:

• PINECONE_API_KEY required.
• If the Hugging Face model is gated, set HUGGINGFACE_TOKEN.
• Ensure Ollama is running locally at OLLAMA_BASE_URL.