title	Election Intel Agent
emoji	🗳️
colorFrom	blue
colorTo	indigo
sdk	docker
pinned	false

🗳️ India Election Intelligence Agent

A multi-agent AI system for querying and analyzing Indian election data from 1962 to 2019, powered by LangGraph.

🎥 Video Walkthrough

🚀 Live Demo

What it does

Ask any question about Indian elections in plain English and the system automatically routes it to the right agents, queries the database, searches the web, retrieves policy documents, and generates visualizations — all in one response.

Example queries:

• Who won Varanasi in 2019?
• Show voter turnout trend in Varanasi across all elections
• What did the 2019 budget say about farmers?
• Tell me about Narendra Modi
• BJP vs INC seats in 2019 Lok Sabha
• Latest news about Rahul Gandhi

Architecture

The system uses a LangGraph state machine to orchestrate 5 specialized agents:

Agent	Purpose
SQL Agent	Queries SQLite election database (575K rows, 1962–2019)
Wikipedia Agent	Fetches background on politicians, parties, constituencies
News Agent	Searches recent news via DuckDuckGo
RAG Agent	Searches budget speeches (1947–2026) and President addresses (1952–2026) via BM25 + ChromaDB
Visualization Agent	LLM writes Plotly code to auto-generate charts

Data Sources

• Election results — Lok Sabha and Vidhan Sabha data from Lok Dhaba, 1962–2019
• Budget speeches — All Union Budget speeches 1947–2026 from indiabudget.gov.in (91 PDFs)
• President addresses — All Presidential addresses to Parliament 1952–2026 from PRS India (70 PDFs)
• RAG corpus — 3,229 chunks, ~1.4M words

Tech Stack

• LLM — Llama 3.1 8B via HuggingFace Router API
• Agent framework — LangGraph
• Vector DB — ChromaDB (local persistent)
• Hybrid search — BM25 (rank-bm25) + ChromaDB cosine similarity
• Embeddings — sentence-transformers/all-MiniLM-L6-v2
• Backend — FastAPI + Uvicorn
• Frontend — Vanilla HTML/JS + Plotly
• Deployment — HuggingFace Spaces (Docker)

Setup

git clone https://github.com/shan1322/Election-Intelligence-agent
cd Election-Intelligence-agent
pip install -r requirements.txt
export HF_TOKEN=your_token_here
python app.py

Environment Variables

Variable	Description
HF_TOKEN	HuggingFace API token for LLM inference