Agentic AI ex-US equity evaluator (LangGraph+Gemini)

Multi-Agent International Equity Analysis System

This repository is a multi-agent international equity research system. It can analyze single tickers, run broader screening pipelines, and optionally reconcile saved results against an Interactive Brokers portfolio through either a CLI workflow or a local Flask dashboard.

You need Python 3.12+, Poetry, and at least one working LLM API key. A Gemini key is the minimum practical setup; optional data-provider keys improve coverage and reliability.

What This Repo Covers

• Multi-agent international equity analysis for individual tickers
• Structured markdown reports and charts
• Screening pipeline for broader exchange-wide discovery
• Optional IBKR portfolio reconciliation and watchlist handling
• Optional local Flask dashboard for portfolio and refresh monitoring

Architecture

This is not a single prompt wrapped in a CLI. The runtime fans out work across specialist agents, applies deterministic validation before debate, then routes the surviving analysis into valuation, risk, and portfolio decision stages.

graph TB
    Start(["User: Analyze TICKER"]) --> Dispatcher{"Parallel<br/>Dispatch"}

    Dispatcher --> MarketAnalyst["Market Analyst<br/>(Technical)"]
    Dispatcher --> SentimentAnalyst["Sentiment Analyst<br/>(Social)"]
    Dispatcher --> NewsAnalyst["News Analyst<br/>(Events)"]
    Dispatcher --> JuniorFund["Junior Fundamentals<br/>(API Data)"]
    Dispatcher --> ForeignLang["Foreign Language<br/>(Native Sources)"]
    Dispatcher --> LegalCounsel["Legal Counsel<br/>(Tax & Reg)"]
    Dispatcher --> ValueTrap["Value Trap Detector<br/>(Governance)"]
    Dispatcher -.-> Audi