Check out a quick demo video here

RAG-powered health insurance policy comparator

Upload your insurance plan PDFs, describe what matters to you in plain English, and get a cited, priority-weighted comparison table with an AI recommendation.

1. Upload policy PDFs (or try pre-loaded sample plans)
2. Describe your needs in plain English (For example, "I need low overall costs and prescription medication coverage. I also travel frequently so I would like my insurance coverage to be available across states.")
3. Review extracted criteria aligned to standard insurance terminology and assign priorities
4. Review the plan comparison table showing Covered / Partial / Not Covered / Unknown per plan per criterion, each with a cited source passage
5. Ask any follow-up questions in a chat interface grounded in your plan documents and comparison results

• Ragas Core Metrics — Faithfulness, Answer Relevancy, Context Precision, Context Recall.
• Synthetic Testset Generation — Ragas TestsetGenerator creates eval dataset from policy documents (no manual labeling).
• Quantitative Before/After Story — Metrics compared with reranking on vs. off to demonstrate measurable improvement. See eval/eval_report.ipynb.
• W&B Weave Tracing — All eval LLM calls logged with prompts, responses, and latency. View eval dashboard →

• Python 3.11+
• Node.js 20+
• Groq API key

cd backend
python -m venv .venv
source .venv/bin/activate
pip install -r requirements-core.txt
pip install -r requirements-eval.txt

cp .env.example .env
# Add your API keys to .env

uvicorn app.main:app --reload

cd frontend
npm install
npm run dev