🧠 Equity Research Agent

A LangGraph-based multi-agent system for automated equity research. The system simulates the workflow of a professional equity research analyst, providing enriched, curated, and briefed insights through a structured multi-agent architecture.

🚀 Features

• 📊 Financial analysis
• 🗞️ News scanning
• 🧠 Sentiment analysis for news and report sections
• 🏭 Industry insights
• 🏢 Company profiling
• 🔍 Data collection and enrichment
• 📝 Auto-generated investment briefings
• ✍️ Final report editing
• 🌐 Real-time WebSocket progress updates
• 📈 Valuation analysis powered by Yahoo Finance
• 💬 Real-time subquery generation for deeper insights
• 🔎 Symbol lookup using Tavily and yfinance

🧱 Architecture

This project is built using LangGraph's StateGraph with the following node pipeline:

grounding
   ├── financial_analyst
   ├── news_scanner
   ├── industry_analyst
   ├── company_analyst
   ├── fundamental_analyst
   └── valuation_analyst
      ↓
   collector → curator → enricher → briefing → editor

Each node represents a specialized agent:

• GroundingNode: Sets research context.
• FinancialAnalyst, NewsScanner, IndustryAnalyzer, CompanyAnalyzer: Conduct domain-specific analysis.
• FundamentalAnalyst, ValuationAnalyst: Perform financial metrics evaluation and market-based valuation with Yahoo Finance API.
• Collector: Aggregates data.
• Curator: Filters and prioritizes findings.
• Enricher: Enhances findings with additional insights.
• Briefing: Creates structured reports.
• Editor: Finalizes output for human consumption.

🧪 How to Run

0. Create a new Python project using uv (optional but recommended)

uv venv
uv pip install -r requirements.txt

1. Install dependencies (if not using uv)

pip install -r requirements.txt

2. Run the Agent

from backend.graph import Graph
graph = Graph(company="Tesla")
# graph.run(thread={}) is an async generator
async for state in graph.run(thread={}):
    print(state)

Note: Make sure to configure WebSocket manager and other integrations as needed.

📂 Project Structure

backend/
├── graph.py         # Main workflow definition
├── classes/
│   └── state.py     # InputState dataclass
└── nodes/           # Modular node logic
    ├── collector.py
    ├── curator.py
    ├── enricher.py
    ├── briefing.py
    ├── editor.py
    └── researchers.py

📡 WebSocket Support

Progress updates are sent via WebSocket (websocket_manager) with:

• Job ID (job_id)
• Current node name
• Execution progress
• Result or error data

🙏 Acknowledgements

This project was inspired by and adapted from company-research-agent by Guy Hartstein. We thank the original author for the excellent foundation and contributions to the open-source research tooling ecosystem.

Licensed and modified under the Apache License 2.0.