saiteja129/agentic-equity-research

Agentic Equity Research System

A state-of-the-art multi-agent workflow for equity research powered by LangGraph, LangChain, and GPT-4o.

🚀 Overview

This system automates the technical and regulatory analysis of stocks. It features a collaborative agent flow where an Analyst performs quantitative research, a Critic validates the findings, and a Human-in-the-Loop gate ensures final report quality.

Key Features:

• —Parallel Tool-Calling: The Analyst agent fetches RSI, VWAP, and filings concurrently.
• —Async Scrapers: Built with httpx and BeautifulSoup for real-time filing extraction.
• —Regulatory Gate: Validates findings against simulated compliance data.
• —3-Strike Refinement: Automatically simplifies analysis if the Critic rejects it multiple times.
• —Stateful Memory: Uses MemorySaver to persist conversations and HITL states.

🏗️ Architecture

graph TD
    START((Start)) --> analyst[Analyst Node]
    analyst --> |tool calls?| tools[Tool Executor]
    tools --> analyst
    analyst --> |no tool calls| critic[Critic Node]
    critic --> |errors found & revisions < 3| analyst
    critic --> |errors found & revisions >= 3| refinement[Refinement Prompt]
    refinement --> analyst
    critic --> |no errors| hitl_gate[HITL Safety Gate]
    hitl_gate --> |interrupt & wait| HUMAN{Human Decision}
    HUMAN --> |approve| final[Finalize Report]
    HUMAN --> |reject| analyst
    final --> END((End))

🛠️ Installation

1. 1.Clone the repository.
2. 2.Install dependencies:

   pip install -r requirements.txt

1. 1.Set up your .env file:

   OPENAI_API_KEY=your_openai_key

1. 1.Run the app:

   python app.py

🔒 Security & Compliance

The system includes a dedicated Critic node that checks for regulatory red flags and logical inconsistencies, ensuring that the final output is safe for consumption.

Built with ❤️ by Antigravity