Constitutioner is a Retrieval-Augmented Generation (RAG) tool that allows users to query the Indian Constitution interactively. It fetches the most relevant sections using semantic search with Pinecone and provides context-aware answers using an LLM.
- Chunked storage of the Indian Constitution using Pinecone DB
- Cosine similarity search to fetch relevant context
- RAG-based inference for accurate and context-rich answers
- Both GUI and CLI versions – run locally or access via the web
- Lightweight backend (no heavy embedding models hosted manually)
- Backend: FastAPI
- Vector DB: Pinecone (with server-side embeddings)
- LLM API: OpenRouterAI API
- Deployment: Render (backend) + Vercel (frontend)
1️⃣ The Indian Constitution is preprocessed and chunked into meaningful sections. 2️⃣ Each chunk is embedded and stored in Pinecone DB. 3️⃣ When a user queries, Pinecone fetches the top-k relevant chunks via cosine similarity. 4️⃣ These chunks are passed as context to the LLM, which generates a final answer.
git clone https://github.com/varshithpilli/constitutioner.git
cd constitutioner
pip install -r requirements.txtCreate a .env file:
API_KEY="sk-or-v1-...."
BASE_URL="https://openrouter.ai/api/v1/chat/completions"
MODEL="mistralai/mistral-small-3.2-24b-instruct:free"
PINECONE_API="pcsk_...."
Run the CLI version:
py cli_model.pyConstitutioner shows how RAG + Pinecone can be used to build lightweight, scalable, and cost-efficient AI apps without hosting heavy embedding models locally.
CLI Version
Desktop Version
Mobile Version
