Multi-Agent Investment Research Platform

AI-powered equity research platform that generates comprehensive investment reports using multi-agent orchestration, RAG, and real-time financial data.

---

Screenshot

📎 More Screenshots(Interactive Charts)

🚀 What It Does

Ask any investment question in natural language:

• English: "What is the investment outlook for Microsoft?"
• 中文: "微软的投资前景如何？"

Get a comprehensive report with:

• Real-time market data & 52-week trends
• Sentiment analysis from news
• Analyst consensus & price targets
• Peer valuation comparison (11 sectors)
• EDGAR SEC filings context

---

⚡ Quick Start

1. Prerequisites

• Python 3.11+
• Node.js 18+
• OpenAI API key (Get one here)
• MongoDB

2. Setup

# Clone repository
git clone <your-repo-url>
cd multi-agent-investment

# Create and activate virtual environment
python -m venv .venv
source .venv/bin/activate  # On Windows: .venv\Scripts\activate

# Install Python dependencies
pip install -r requirements.txt

# Install frontend dependencies
cd frontend
npm install
cd ..

3. Configure Environment Variables

# Copy template
cp .env.template .env

# Edit .env and add your credentials:
# - OPENAI_API_KEY: Your OpenAI API key
# - MONGODB_URI: Your MongoDB connection string

Required Environment Variables:

• OPENAI_API_KEY - OpenAI API key for GPT-4o and embeddings
• MONGODB_URI - MongoDB connection string
• MONGODB_DB_NAME - Database name (default: investment_research)
• CHROMA_PERSIST_DIR - Local directory for vector store (default: ./data/chroma)

4. Initialize Databases

python -m backend.scripts.init_db
python -m backend.scripts.init_ticker_cache

5. Run the Application

# Backend (Terminal 1)
uvicorn backend.main:app --reload --port 8000

# Frontend (Terminal 2)
cd frontend && npm install && npm run dev

Access:

• Frontend: http://localhost:3000
• API: http://localhost:8000/docs

---

💡 Core Features

✅ Multi-Agent Workflow - LangGraph orchestration (router, market data, sentiment, report)
✅ Real-time Data - Yahoo Finance, SEC EDGAR, financial news
✅ Smart Analysis - 52-week trends, peer valuation, analyst consensus
✅ RAG Pipeline - ChromaDB vector search for EDGAR filings
✅ Bilingual - Auto-detects language (EN/CN)
✅ Conversation Memory - MongoDB session history (24h TTL)
✅ Quality Assurance - Automatic report reflection & refinement (up to 3 iterations)
✅ Explainable AI - Reasoning chain tracking for all agents
✅ Smart Retry - Automatic recovery from transient API errors

---

🛠️ Technology Stack

Component	Technology
Backend	FastAPI, Python 3.11+
Multi-Agent	LangGraph, LangChain
LLM	OpenAI GPT
Databases	MongoDB, ChromaDB
Frontend	React, TypeScript, Tailwind CSS
Data Sources	Yahoo Finance, SEC EDGAR

---

📝 Example Usage

Via Frontend

1. Open http://localhost:3000
2. Type: "Analyze Apple's recent performance"
3. Get comprehensive report in seconds

Via API

curl -X POST http://localhost:8000/api/research/query \
  -H "Content-Type: application/json" \
  -d '{"query": "What is the investment outlook for NVDA?"}'

---

📁 Project Structure

.
├── backend/
│   ├── agents/
│   ├── api/             # FastAPI REST endpoints
│   ├── memory/          # MongoDB conversation memory
│   ├── rag/             # RAG pipeline (EDGAR, news)
│   ├── services/        # Yahoo Finance, ChromaDB, Ticker Resolver
│   └── config/          # Settings & environment config
├── frontend/
│   └── src/
│       ├── components/  # React UI components
│       ├── api/         # API client
│       └── types/       # TypeScript definitions
├── data/
│   ├── chroma/          # Vector store (local)
│   ├── edgar_filings/   # Downloaded SEC filings (not in git)
│   └── ticker_cache.json # Ticker resolution cache
├── tests/               # Test suite
└── PLAN.md              # Development roadmap & progress tracking

---

🎨 Features Showcase

🔍 Deep Analysis Mode

• On-demand SEC 10-K filing analysis
• Automatic download and vector embedding
• Comprehensive business insights and risk analysis

🌐 Multi-Language Support

• Automatic language detection (English/Chinese)
• Bilingual UI and reports
• Natural query understanding

📊 Real-time Market Data

• Yahoo Finance integration
• 52-week price trends
• Peer sector comparison (11 sectors)
• Analyst consensus ratings

🤖 Advanced Agent Features

Reflection Loop (Generate → Reflect → Refine):

• Automatic quality evaluation (completeness, consistency, actionability, clarity)
• Iterative refinement up to 3 times for optimal report quality
• Quality threshold: 8.5/10 (88% score)

Reasoning Chain Tracking:

• Complete visibility into agent decision-making process
• Step-by-step reasoning logs for debugging and explainability
• Detailed execution metrics (time, attempts, success rate)

Smart Error Handling:

• Automatic retry with exponential backoff (1s → 2s → 4s)
• Intelligent error classification (transient vs permanent)
• Graceful degradation for partial failures

Performance Monitoring:

• Per-agent execution time tracking
• Token usage monitoring
• Success/failure statistics

---

🏗️ System Architecture

Multi-Agent Workflow

User Query
    ↓
┌─────────────────────────────────────────────┐
│ Router Agent (Reasoning Chain Enabled)     │
│ • Extract tickers (dynamic resolver)       │
│ • Analyze intent                            │
│ • Set routing flags                         │
└─────────────────────────────────────────────┘
    ↓
┌─────────────────────────────────────────────┐
│ Parallel Execution (4 Agents)              │
│                                             │
│ ┌─────────────┐  ┌──────────────┐         │
│ │ Market Data │  │  Sentiment   │         │
│ │   Agent     │  │    Agent     │         │
│ └─────────────┘  └──────────────┘         │
│                                             │
│ ┌─────────────┐  ┌──────────────┐         │
│ │  Forward    │  │     RAG      │         │
│ │  Looking    │  │  Retrieval   │         │
│ └─────────────┘  └──────────────┘         │
│                                             │
│ All agents track: reasoning, metrics, retry│
└─────────────────────────────────────────────┘
    ↓
┌─────────────────────────────────────────────┐
│ Aggregator (Sync Point)                    │
└─────────────────────────────────────────────┘
    ↓
┌─────────────────────────────────────────────┐
│ Visualization Agent                         │
│ • Price history charts                      │
│ • Peer comparison data                      │
└─────────────────────────────────────────────┘
    ↓
┌─────────────────────────────────────────────┐
│ Report Agent (Reflection Loop)             │
│                                             │
│  Iteration 1:                               │
│  ┌──────────┐  ┌──────────┐  ┌──────────┐ │
│  │ Generate │→ │ Reflect  │→ │  Refine  │ │
│  └──────────┘  └──────────┘  └──────────┘ │
│                     ↓                       │
│              Quality >= 0.85?               │
│                ✓ Yes → Done                 │
│                ✗ No → Iteration 2 (max 3)   │
└─────────────────────────────────────────────┘
    ↓
Final Report + Metrics + Reasoning Chains

Quality Assurance Pipeline

Report Reflection Process:

1. Generate: Create initial report from all agent data
2. Reflect: LLM evaluates 4 dimensions (0-10 scale each):
   • Completeness: Uses all available data sources
   • Consistency: No internal contradictions
   • Actionability: Clear investment insights
   • Clarity: Well-structured and understandable
3. Refine: If overall score < 8.5, regenerate with improvement feedback
4. Iterate: Repeat up to 3 times or until quality threshold met

Example Quality Feedback:

{
  "completeness": 9,
  "consistency": 9,
  "actionability": 8,
  "clarity": 9,
  "overall_score": 8.8,
  "strengths": [
    "Comprehensive data integration",
    "Clear investment recommendation"
  ],
  "gaps": [
    "Missing forward-looking catalysts"
  ]
}

---

🤝 Contributing

This is a personal portfolio project demonstrating:

• Multi-agent AI systems with LangGraph
• RAG pipeline implementation
• Full-stack development (FastAPI + React)
• Financial data integration

For questions or suggestions, please open an issue or contact me at [email protected].

---

📄 License

This project is licensed under the MIT License - see the LICENSE file for details.

---

🙏 Acknowledgments

• LangGraph & LangChain - Multi-agent orchestration framework
• OpenAI - GPT for analysis and embeddings
• Yahoo Finance - Real-time market data
• SEC EDGAR - Official company filings

---