AI Agents

A multi-agent assistant framework built with LangGraph and LangChain.

Current Agents

Portfolio Assistant

An LLM-driven multi-agent system for portfolio analytics. The router agent decomposes user queries into execution plans across specialist agents:

• SQL Agent — Data retrieval from the finance database
• Compute Agent — Statistical computations and derived metrics
• Risk Agent — Risk measurement (volatility, beta, drawdowns, VaR)
• Explain Agent — Natural-language interpretations
• Viz Agent — Chart generation
• Report Agent — Structured report assembly

Project Structure

├── config.yaml              # LLM configurations (OpenRouter)
├── data/finance/            # SQLite database + schema docs
├── notebooks/               # Development notebooks
├── src/
│   └── portfolio_assistant/ # Router and agent implementations
└── requirements.txt

Setup

pip install -r requirements.txt

Create a .env file with your API key:

OPENROUTER_API_KEY=your_key_here

Usage

See notebooks/01_AC_multiagent_setup.ipynb for examples.