# LangGraph Comparison Agent

## Overview

This repository contains a **LangGraph** agent that, given three entities (e.g. vector databases, frameworks, or products), builds a comparative review. The agent:

1. **Plans criteria** – LLM generates 3‑5 comparison criteria based on the entities.
2. **Researches each entity‑criterion pair** – uses **Tavily** web search to fetch a short note for each pair.
3. **Builds a Markdown table** – rows are criteria, columns are entities.
4. **Produces a verdict** – LLM gives a concise recommendation.

The final output is a Markdown table followed by a verdict paragraph.

## Requirements

```bash
pip install -r requirements.txt
```

Environment variables:

- `OPENAI_API_KEY` – key for the OpenAI model used by LangChain.
- `TAVILY_API_KEY` – key for the Tavily API.

## Usage

```bash
# Default demo – compares Chroma, FAISS, Qdrant
python -m src.main

# Custom entities
python -m src.main "Chroma,FAISS,Qdrant"
```

The script prints the comparison table and verdict in Markdown format.