FundAI - Multi-Agent Fund Analysis & Decision System

This project is currently in the early development stage. Features, APIs. Production use is not recommended at this time. Contributions and feedback are welcome!

中文 | English

---

Overview

FundAI is a multi-agent collaborative fund analysis and decision-making system. Through coordinated work of multiple specialized agents (Fundamental Analyst, Technical Analyst, Risk Analyst, Cost Analyst, and Sentiment Analyst), it provides users with investment recommendations that have undergone cost verification.

Key Features

• Multi-Agent Collaboration: Five specialized analytical agents working in parallel/series
• Intelligent Decision Making: Decision agent synthesizes all analysis results to provide short-term/long-term investment recommendations
• Real-time Streaming: SSE-based streaming analysis process visualization
• Debate Mechanism: Detects scoring divergence and triggers agent debates for adjustment
• RAG Knowledge Enhancement: LLM-driven agents with tool invocation and RAG knowledge enhancement
• Data Quality Validation: Built-in data quality verification and provenance tracking
• Flexible Data Sources: Support for AKShare and Tushare Pro data sources with auto-switching and caching

Architecture

Backend

Component	Technology
Runtime	Python 3.12+
Framework	FastAPI 0.109+
ORM	SQLAlchemy 2.0+ (Async)
Database	PostgreSQL 15+
Cache	Redis 7.0+
Agent Framework	LangChain / LangGraph
Vector Store	ChromaDB
Data Sources	AKShare, Tushare Pro

Frontend

Component	Technology
Framework	Nuxt 4.x
UI	Vue 3.5+
UI Components	Element Plus
State Management	Pinia
Charts	ECharts
Language	TypeScript

Project Structure

FundAI/
├── app/                        # Backend application
│   ├── api/v1/endpoints/       # API endpoints
│   ├── core/                   # Core configuration
│   │   ├── config.py           # Settings management
│   │   ├── database.py         # Async database engine
│   │   ├── cache.py            # Redis cache
│   │   └── calculations/       # Technical indicators
│   ├── models/                 # SQLAlchemy ORM models
│   ├── schemas/                # Pydantic models
│   ├── agents/                 # Multi-agent system
│   │   ├── base.py             # BaseAgent implementation
│   │   ├── orchestrator.py     # Agent orchestration
│   │   ├── fundamental.py      # Fundamental analyst
│   │   ├── technical.py        # Technical analyst
│   │   ├── risk.py             # Risk analyst
│   │   ├── cost.py             # Cost analyst
│   │   ├── sentiment.py        # Sentiment analyst
│   │   ├── decision.py         # Decision analyst
│   │   ├── prompts/            # System prompts
│   │   └── tools/              # Agent tools
│   ├── data_sources/           # Data source adapters
│   ├── services/               # Business services
│   └── main.py                 # Application entry
├── frontend/                   # Frontend application
│   ├── app/                    # Nuxt app directory
│   │   ├── components/         # Vue components
│   │   ├── composables/        # Composable functions
│   │   ├── pages/              # Page routes
│   │   ├── services/           # API services
│   │   ├── stores/             # Pinia stores
│   │   └── utils/              # Utility functions
│   ├── server/                 # Nuxt server
│   └── nuxt.config.ts          # Nuxt configuration
├── database/                   # Database scripts
├── tests/                      # Test modules
├── requirements.txt            # Backend dependencies
└── README.md                   # Project documentation

Quick Start

Prerequisites

Ensure the following are installed:

• Python 3.12+
• Node.js 22+
• PostgreSQL 15+
• Redis 7.0+

Backend Setup

# Create virtual environment
python -m venv venv
.\venv\Scripts\activate

# Install dependencies
pip install -r requirements.txt

# Configure environment variables
copy .env.example .env
# Edit .env file to configure database connection and other settings

# Initialize database
# Execute database/init_db.sql

# Start backend server
uvicorn app.main:app --reload --host 0.0.0.0 --port 8000

Frontend Setup

# Navigate to frontend directory
cd frontend

# Install dependencies
npm install

# Start development server
npm run dev

Access the Application

• Frontend Application: http://localhost:3000
• Backend API Docs: http://localhost:8000/docs
• Health Check: http://localhost:8000/health

API Endpoints

Authentication

Method	Endpoint	Description
POST	/api/v1/auth/send-code	Send verification code
POST	/api/v1/auth/register	User registration
POST	/api/v1/auth/login	User login
POST	/api/v1/auth/logout	User logout
POST	/api/v1/auth/refresh	Refresh token

Funds

Method	Endpoint	Description
GET	/api/v1/funds/search	Search funds
GET	/api/v1/funds/{fund_code}	Get fund details
GET	/api/v1/funds/{fund_code}/nav	Get NAV history
GET	/api/v1/funds/{fund_code}/holdings	Get holdings information
GET	/api/v1/funds/{fund_code}/fees	Get fee information

Analysis

Method	Endpoint	Description
POST	/api/v1/analysis/sessions	Create analysis session
GET	/api/v1/analysis/sessions/{session_id}/stream	SSE streaming analysis
GET	/api/v1/analysis/sessions/{session_id}/report	Get analysis report

Sessions

Method	Endpoint	Description
GET	/api/v1/sessions	Get session list
GET	/api/v1/sessions/{session_id}	Get session details

Development

Backend Development

# Run tests
pytest

# Code formatting
black app/
isort app/

Frontend Development

cd frontend

# Build for production
npm run build

# Preview production build
npm run preview

# Type checking
npm run typecheck

Docker Deployment

# Start PostgreSQL + Redis + Application
docker-compose up -d

Core Architecture

Multi-Agent System

• 5 Analytical Agents execute in parallel/series: Fundamental, Technical, Risk, Cost, Sentiment
• 1 Decision Agent synthesizes all analysis results and provides short-term/long-term investment recommendations
• Each agent is LLM-driven with tool invocation and RAG knowledge enhancement capabilities
• Agents push SSE events to frontend through EventCallback

Analysis Workflow

1. Build context (fetch fund data → data quality validation → data provenance tagging)
2. Execute 5 analytical agents in parallel
3. Debate Mechanism: Detect scoring divergence (threshold 1.5), diverging parties review each other and adjust scores
4. Decision agent synthesizes scores and generates investment recommendations
5. Historical results stored in AgentMemory for future reference

SSE Streaming

• Real-time event pushing via EventCallback + asyncio.Queue
• Event types: agent status changes, thinking process, tool calls, deep reasoning, debate rounds, progressive updates
• Frontend subscribes via EventSource with pause/resume support

API Proxy

• Frontend Nuxt proxies all /api/v1/* requests to backend via server/routes/api/v1/[...].ts
• Frontend uses process.env.NUXT_API_URL or defaults to http://localhost:8000

LLM Configuration

• LLM settings managed through frontend settings page, stored in data/config.json
• Supports any OpenAI-compatible API (API URL, API Key, model name, temperature, etc.)
• Configuration takes effect immediately without service restart

Data Sources

• Supports AKShare and Tushare Pro data sources
• DataSourceManager provides unified management with auto-switching and caching

Testing Strategy

• Framework: pytest + pytest-asyncio (asyncio_mode = auto)
• Database: In-memory SQLite with automatic table creation/cleanup per test
• HTTP Client: httpx.AsyncClient + ASGITransport for direct ASGI application calls
• Key Mock Fixtures: mock_llm_service, mock_datasource_manager, mock_rag_service
• Test Organization: tests/ organized by module (test_agents/, test_api/, test_core/, etc.)

License

MIT License - see LICENSE file for details.