OpenFR: Lightweight Financial Research Agent | Powered by AKShare | Multi-LLM | Multi-Agent Deep Analysis

中文 | Quick Start • Features • Usage • Configuration • Architecture

---

📊 Overview

OpenFR (Open Financial Research) is a minimal, lightweight intelligent financial research Agent. Built on large language models and integrated with AKShare data APIs, it uses multi-agent collaboration to deliver in-depth investment research across stocks, funds, futures, indices, macroeconomics, and more.

✨ Features

• 🌱 Minimal & Lightweight — Pure Python package + Typer CLI, AKShare data only, one command to start researching
• 🧠 Multi-Agent Collaboration — Four analysts + bull/bear debate + three-way risk assessment, orchestrated via LangGraph StateGraph
• ⏱️ Per-Node Timing — Elapsed time displayed after each agent node, making it easy to identify performance bottlenecks
• 📋 Full Intermediate Reports — Market / fundamental / news / macro reports, debate transcripts, and risk assessments shown in full
• 📈 Rich Data — 35+ financial data tools: A-shares, HK stocks, funds, futures, indices, macro, and sectors
• 🔄 Multi-LLM — 15+ providers (domestic Chinese, overseas, local), compatible with OpenAI and Anthropic formats
• 🎨 Nice CLI — Rich terminal UI with live progress and complete analysis content per stage
• 🔌 Fallback Sources — East Money + Sina + Tonghuashun with automatic switch and retry
• 💾 Cache Friendly — Stock list cached 6h, some quote data cached 1min, reducing redundant requests
• 🛡️ Error Recovery — Retry, fallback, and "finish with available info" protection logic

---

🏗️ Architecture

OpenFR uses LangGraph StateGraph to orchestrate a three-phase multi-agent workflow:

START
  ↓
┌─────────────────────────────────────────────┐
│  Phase 1: Data Collection & Analysis         │
│                                             │
│  📈 Market Analyst      ← quote/index/sector │
│       ↓                                     │
│  📊 Fundamentals Analyst ← financials/flow   │
│       ↓                                     │
│  📰 News Analyst         ← news/announcements│
│       ↓                                     │
│  🏛️ Macro Analyst        ← CPI/PPI/PMI/GDP   │
└─────────────────────────────────────────────┘
  ↓
┌─────────────────────────────────────────────┐
│  Phase 2: Investment Debate (Bull vs Bear)   │
│                                             │
│  🐂 Bull Researcher ⇄ 🐻 Bear Researcher    │
│       ↓                 (1–3 rounds)        │
│  👔 Research Manager → Initial recommendation│
└─────────────────────────────────────────────┘
  ↓
┌─────────────────────────────────────────────┐
│  Phase 3: Risk Assessment (Three-Way Debate) │
│                                             │
│  🔥 Aggressive ⇄ 🛡️ Conservative ⇄ ⚖️ Neutral│
│       ↓                                     │
│  💼 Portfolio Manager → Final decision       │
└─────────────────────────────────────────────┘
  ↓
END

Final output:

• Rating: Buy / Overweight / Hold / Underweight / Sell
• Confidence: High / Medium / Low
• Detailed reasoning
• Action recommendations

Per-node timing: Each node name is followed by Xs elapsed time. Individual tool call timings are written to DEBUG logs.

---

🚀 Quick Start

Install

# Clone repo
git clone https://github.com/openmozi/openfr.git
cd openfr

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

# Install
pip install -e .

Configure

Create a .env file and set your API key:

# Recommended default: Zhipu AI
ZHIPU_API_KEY=your_zhipu_api_key_here
OPENFR_PROVIDER=zhipu
OPENFR_MODEL=glm-4.7

See Configuration for more providers and options.

Run

# Interactive chat (recommended)
openfr chat

# Single query
openfr query "Is Kweichow Moutai a good buy?" --target 贵州茅台
openfr query "Analyze BYD's investment value" --target 比亚迪 -p deepseek

# List tools and providers
openfr tools
openfr providers

---

📖 Usage

Interactive chat

openfr chat
openfr chat -p dashscope
openfr chat -p zhipu -m glm-4-plus

Then type your question:

You: What is Kweichow Moutai's price today?
You: Analyze today's hot sectors
You: How is the Shanghai Composite Index?
You: Is BYD worth buying?

After each agent node completes, you'll see timing info, for example:

📈 Market Analyst · ✓ Market report generated (951 chars) 28.3s
📊 Fundamentals Analyst · ✓ Fundamentals report generated (778 chars) 31.7s
...
⏱ 11 nodes executed, total time 363.5s

Single query

openfr query "Is Kweichow Moutai a good buy?" --target 贵州茅台
openfr query "How did the Shanghai Composite Index perform today?"
openfr query "Analyze BYD's investment value" --target 比亚迪 -p zhipu

Supported query types

📈 A-shares

Real-time quote for 000001
Kweichow Moutai last week trend
Search new energy related stocks
Today's hot stocks

🇭🇰 HK stocks

Real-time quote for HK 00700
Tencent Holdings price today
Search HK Li Auto

💼 Funds

ETF data for 510300
Top equity fund ranking

📊 Indices and sectors

Shanghai Composite today
Today's top sectors by gain

🌍 Macro

Latest CPI data
Recent GDP growth
PMI trend

---

⚙️ Configuration

Model setup

Set your provider and model in .env (see .env.example):

OPENFR_PROVIDER=zhipu          # provider name
OPENFR_MODEL=glm-4.7           # model name (leave empty to use provider default)
ZHIPU_API_KEY=your_api_key     # API key for the chosen provider

provider	API Key env var	Default model
deepseek	DEEPSEEK_API_KEY	deepseek-chat
doubao	DOUBAO_API_KEY	doubao-1-5-pro-256k
dashscope	DASHSCOPE_API_KEY	qwen-max
zhipu	ZHIPU_API_KEY	glm-4.7 (default provider)
modelscope	MODELSCOPE_API_KEY	qwen2.5-72b-instruct
kimi	KIMI_API_KEY	moonshot-v1-128k
stepfun	STEPFUN_API_KEY	step-2-16k
minimax	MINIMAX_API_KEY	MiniMax-Text-01
openai	OPENAI_API_KEY	gpt-4o
anthropic	ANTHROPIC_API_KEY	claude-sonnet-4-20250514
openrouter	OPENROUTER_API_KEY	anthropic/claude-sonnet-4
together	TOGETHER_API_KEY	meta-llama/Llama-3.3-70B-Instruct-Turbo
groq	GROQ_API_KEY	llama-3.3-70b-versatile
ollama	OLLAMA_BASE_URL	qwen2.5:14b
custom	CUSTOM_API_KEY + CUSTOM_BASE_URL + CUSTOM_API_STYLE	(specify)

You can also switch provider/model at runtime with -p / -m:

openfr chat -p deepseek
openfr chat -p openai -m gpt-4o
openfr query "Analyze Moutai" -p groq

Other options

# Debate rounds (more rounds = deeper analysis, longer runtime)
OPENFR_MAX_DEBATE_ROUNDS=1          # bull/bear debate rounds (default 1)
OPENFR_MAX_RISK_DISCUSS_ROUNDS=1    # risk debate rounds (default 1)

# Custom OpenAI-compatible endpoint
OPENFR_PROVIDER=custom
CUSTOM_BASE_URL=https://your-api.example.com
CUSTOM_API_KEY=your-api-key
CUSTOM_API_STYLE=openai             # openai or anthropic

---

🐛 Troubleshooting

Common issues

1. API key not set

# In .env
ZHIPU_API_KEY=your-api-key-here

# Or temporarily
export ZHIPU_API_KEY=your-api-key-here

2. Network errors

Auto retry (up to 3 times) with fallback to backup data source. If it keeps failing, check your network.

3. Data API unavailable

Some real-time data is only available during market hours (weekdays 9:30–15:00 CST). Use history APIs instead.

4. Slow execution

Multi-agent mode requires at minimum ~15 serial LLM calls. Total time depends heavily on model latency. To speed up:

• Use a fast model such as groq or deepseek
• Lower OPENFR_MAX_DEBATE_ROUNDS and OPENFR_MAX_RISK_DISCUSS_ROUNDS
• Check the per-node Xs timing in the output to find the slowest nodes

---

🤝 Contributing

Contributions, issues, and ideas are all welcome.

1. Fork the repo
2. Create a branch (git checkout -b feature/AmazingFeature)
3. Commit (git commit -m 'Add some AmazingFeature')
4. Push (git push origin feature/AmazingFeature)
5. Open a Pull Request

Code style: format with Black, lint with Ruff, add type hints where useful.

---

🙏 Acknowledgments

• AKShare — Financial data APIs
• LangChain — Agent framework
• LangGraph — Multi-agent graph orchestration
• TradingAgents — Multi-agent financial research architecture reference
• Rich — Terminal UI
• Typer — CLI framework

---

⬆ Back to top

中文 | English

Made with ❤️ by OpenFR Team