AI Stock Analysis System

AI-powered stock analysis system for A-shares / Hong Kong / US stocks

Analyze your watchlist daily → generate a decision dashboard → push to multiple channels (Telegram/Discord/Slack/Email/WeChat Work/Feishu)

Zero-cost deployment · Runs on GitHub Actions · No server required

Quick Start · Key Features · Sample Output · Full Guide · FAQ · Contributing · All Docs

English | 简体中文 | 繁體中文

💖 Sponsors

Easily scrape real-time financial news data from search engines - SerpApi

✨ Key Features

Module	Feature	Description
AI	Decision Dashboard	One-sentence conclusion + precise entry/exit levels + action checklist
Analysis	Multi-dimensional Analysis	Technicals + chip distribution + sentiment + real-time quotes
Market	Global Markets	A-shares, Hong Kong stocks, US stocks
Search	Smart Autocomplete (MVP)	[Beta] Home search supports code/name/pinyin/aliases; the local index now covers A-shares, Hong Kong, and US stocks and can be refreshed from Tushare or AkShare data
Review	Market Review	Daily overview, sectors, northbound capital flow
Intel	Announcement + Capital Flow Intelligence	IntelAgent now also pulls listed-company announcements (SSE/SZSE/CNINFO) and A-share main-force capital flow, and exposes `capital_flow_signal` (`inflow/outflow/neutral/not_available`) for flow direction context
Backtest	AI Backtest Validation	Auto-evaluate historical analysis accuracy, with a 1-day next-session validation view for AI prediction vs actual move and accuracy
Agent Q&A	Strategy Chat	Multi-turn strategy chat with 11 built-in trading strategies (internally loaded as skills) (Web/Bot/API)
Notifications	Multi-channel Push	Telegram, Discord, Slack, Email, WeChat Work, Feishu, etc.
Automation	Scheduled Runs	GitHub Actions scheduled execution, no server required

The Backtest page now includes a 1-day next-session validation view. You can filter by stock code and analysis date range to compare the original AI prediction with the next trading day close and inspect the filtered accuracy rate. This is based on historical analysis plus eval_window_days=1 backtest data, not real trade execution logs.

Tech Stack & Data Sources

Type	Supported
LLMs	Gemini (free), OpenAI-compatible, DeepSeek, Qwen, Claude, Ollama
Market Data	AkShare, Tushare, Pytdx, Baostock, YFinance, Longbridge (primary for US/HK when configured)
News Search	Tavily, Anspire、SerpAPI, Bocha, Brave, MiniMax

Longbridge-first (US/HK only): With LONGBRIDGE_APP_KEY / LONGBRIDGE_APP_SECRET / LONGBRIDGE_ACCESS_TOKEN set, daily bars and realtime quotes for US & HK stocks are fetched from Longbridge first; YFinance / AkShare are used for fallback or field merge when Longbridge fails or returns incomplete fields. If Longbridge is not configured, it is not called — US/HK still use YFinance / AkShare as before. US market indices (e.g. SPX) always prefer YFinance (indices are not supported on Longbridge). A-share routing is unchanged. See .env.example and the full guide.

Built-in Trading Rules

Rule	Description
No chasing highs	Auto warn when deviation > 5%
Trend trading	Bull alignment: MA5 > MA10 > MA20
Precise levels	Entry, stop loss, target
Checklist	Each condition marked as Pass / Watch / Fail

🚀 Quick Start

Option 1: GitHub Actions (Recommended, Zero Cost)

No server needed, runs automatically every day!

1. Fork this repository

Click the Fork button in the upper right corner

2. Configure Secrets

Go to your forked repo → Settings → Secrets and variables → Actions → New repository secret

AI Model Configuration (Choose one)

For detailed configuration, see LLM Config Guide. The default path is: pick a provider, add the API key, then optionally pin a primary model. Use channels only when you need multi-provider routing or fallbacks; advanced YAML routing is optional for expert setups.

Secret Name	Description	Required
`GEMINI_API_KEY`	Get free API key from Google AI Studio	✅*
`OPENAI_API_KEY`	OpenAI-compatible API Key (supports DeepSeek, Qwen, etc.)	Optional
`OPENAI_BASE_URL`	OpenAI-compatible API endpoint (e.g., `https://api.deepseek.com/v1`)	Optional
`OPENAI_MODEL`	Model name (e.g., `deepseek-chat`)	Optional
`OLLAMA_API_BASE`	Ollama local service address (e.g. `http://localhost:11434`), for local/Docker deployment; do not use `OPENAI_BASE_URL` for Ollama, see LLM Config Guide - Ollama	Optional

*Note: Configure at least one of GEMINI_API_KEY, OPENAI_API_KEY, or OLLAMA_API_BASE (local). Ollama requires OLLAMA_API_BASE; using OPENAI_BASE_URL causes 404.

Notification channels (expand, choose at least one)

Secret Name	Description	Required
`TELEGRAM_BOT_TOKEN`	Telegram Bot Token (Get from @BotFather)	Optional
`TELEGRAM_CHAT_ID`	Telegram Chat ID	Optional
`TELEGRAM_MESSAGE_THREAD_ID`	Telegram Topic ID (For sending to topics)	Optional
`DISCORD_WEBHOOK_URL`	Discord Webhook URL	Optional
`DISCORD_BOT_TOKEN`	Discord Bot Token (choose one with Webhook)	Optional
`DISCORD_MAIN_CHANNEL_ID`	Discord Channel ID (required when using Bot)	Optional
`DISCORD_INTERACTIONS_PUBLIC_KEY`	Discord Public Key (required only for inbound Interaction/Webhook signature verification)	Optional
`SLACK_BOT_TOKEN`	Slack Bot Token (recommended, supports image upload; takes priority over Webhook when both set)	Optional
`SLACK_CHANNEL_ID`	Slack Channel ID (required when using Bot)	Optional
`SLACK_WEBHOOK_URL`	Slack Incoming Webhook URL (text only, no image support)	Optional
`EMAIL_SENDER`	Sender email (e.g., `xxx@qq.com`)	Optional
`EMAIL_PASSWORD`	Email authorization code (not login password)	Optional
`EMAIL_RECEIVERS`	Receiver emails (comma-separated, leave empty to send to yourself)	Optional
`WECHAT_WEBHOOK_URL`	WeChat Work Webhook URL	Optional
`FEISHU_WEBHOOK_URL`	Feishu Webhook URL	Optional
`PUSHPLUS_TOKEN`	PushPlus Token (Get it here, Chinese push service)	Optional
`SERVERCHAN3_SENDKEY`	ServerChan v3 SendKey ((Get it here, Mobile app push notification service) )	Optional
`CUSTOM_WEBHOOK_URLS`	Custom Webhook URLs (supports DingTalk, etc., comma-separated)	Optional
`CUSTOM_WEBHOOK_BEARER_TOKEN`	Bearer token for custom webhooks (if required)	Optional
`SINGLE_STOCK_NOTIFY`	Send notification immediately after each stock	Optional
`REPORT_TYPE`	`simple`, `full`, or `brief` (Docker recommended: `full`)	Optional
`REPORT_LANGUAGE`	Report output language: `zh` (default Chinese) / `en` (English); affects prompt instructions, Markdown templates, notification fallbacks, and fixed labels in the Web report view. The bundled `daily_analysis.yml` already maps this variable, so setting it in Actions Secrets/Variables works out of the box	Optional
`ANALYSIS_DELAY`	Delay between stocks and market review (seconds)	Optional

Note: Configure at least one channel; multiple channels will all receive notifications.

Stock List Configuration

Secret Name	Description	Required
`STOCK_LIST`	Watchlist codes, e.g., `600519,AAPL,hk00700`	✅
`TAVILY_API_KEYS`	Tavily Search API (for news)	Recommended
`ANSPIRE_API_KEYS`	Anspire AI Search Specially optimized for Chinese content (effectively enhances A-share analysis)	Optional
`MINIMAX_API_KEYS`	MiniMax Coding Plan Web Search (structured search results)	Optional
`BRAVE_API_KEYS`	Brave Search API (privacy-focused, US stocks optimized)	Optional
`SERPAPI_API_KEYS`	SerpAPI Backup search	Optional
`BOCHA_API_KEYS`	Bocha Search Web Search API (Chinese search optimized, supports AI summaries, multiple keys comma-separated)	Optional
`SEARXNG_BASE_URLS`	SearXNG self-hosted instances (quota-free fallback, enable format: json in settings.yml); when empty the app auto-discovers public instances	Optional
`SEARXNG_PUBLIC_INSTANCES_ENABLED`	Auto-discover public SearXNG instances from `searx.space` when `SEARXNG_BASE_URLS` is empty (default `true`)	Optional
`TUSHARE_TOKEN`	Tushare Pro Token	Optional
`TICKFLOW_API_KEY`	TickFlow API key (CN market review index enhancement; breadth also uses TickFlow when the plan supports universe queries)	Optional
`LONGBRIDGE_APP_KEY`	Longbridge OpenAPI App Key (becomes primary US/HK data source when configured)	Optional
`LONGBRIDGE_APP_SECRET`	Longbridge App Secret	Optional
`LONGBRIDGE_ACCESS_TOKEN`	Longbridge Access Token	Optional
`LONGBRIDGE_STATIC_INFO_TTL_SECONDS`	In-process `static_info` cache TTL in seconds (default `86400`; `0` = no cache)	Optional
`LONGBRIDGE_HTTP_URL`	HTTP API base URL (default `https://openapi.longbridge.com`)	Optional
`LONGBRIDGE_QUOTE_WS_URL`	Quote WebSocket URL (default `wss://openapi-quote.longbridge.com/v2`)	Optional
`LONGBRIDGE_TRADE_WS_URL`	Trade WebSocket URL (default `wss://openapi-trade.longbridge.com/v2`)	Optional
`LONGBRIDGE_REGION`	Override region endpoint; the SDK auto-selects by network (default `hk`); set if wrong (e.g. `cn`, `hk`)	Optional
`LONGBRIDGE_ENABLE_OVERNIGHT`	Overnight session quotes: `true` / `false` (default `false`)	Optional
`LONGBRIDGE_PUSH_CANDLESTICK_MODE`	Candlestick push mode: `realtime` or `confirmed` (default `realtime`)	Optional
`LONGBRIDGE_PRINT_QUOTE_PACKAGES`	Whether to print quote packages on connect (default `false` when unset; set `1`/`true`/`yes` to enable)	Optional
`WECHAT_MSG_TYPE`	WeChat Work message type, default `markdown`, set to `text` for plain markdown text	Optional
`AGENT_MODE`	Enable Agent strategy chat mode (internally normalized as `skill`, `true`/`false`, default `false`)	Optional
`AGENT_LITELLM_MODEL`	Optional Agent-only primary model; when empty it inherits the primary model, and bare names are normalized to `openai/<model>`	Optional
`AGENT_MAX_STEPS`	Max reasoning-step limit for Agent mode (default `10`); at the default each sub-agent keeps its own preset, when raised above the default all sub-agents adopt this value, and when lowered below a sub-agent's preset that sub-agent is capped at this value	Optional
`AGENT_SKILLS`	Comma-separated active strategy-skill ids. Leave empty to use the primary default strategy skill declared in metadata (built-in default: `bull_trend`); use `all` to activate every loaded strategy skill.	Optional
`AGENT_SKILL_DIR`	Custom strategy-skill directory (default built-in `strategies/` compatibility path)	Optional

Stock Code Format

Market	Format	Examples
A-shares	6-digit number	`600519`, `000001`, `300750`
BSE (Beijing)	8/4/92 prefix, 6-digit	`920748`, `838163`, `430047`
HK Stocks	hk + 5-digit number	`hk00700`, `hk09988`
US Stocks	1-5 uppercase letters	`AAPL`, `TSLA`, `GOOGL`

3. Enable Actions

Go to Actions tab → Click I understand my workflows, go ahead and enable them

4. Manual Test

Actions → Daily Stock Analysis → Run workflow → Select mode → Run workflow

5. Done!

The system will:

Run automatically at scheduled time (default: 18:00 Beijing Time)
Send analysis reports to all configured channels
Save reports locally

Resume fetch and --dry-run data-existence checks now resolve the "latest reusable trading day" from each market's local timezone and trading calendar. Weekends and holidays reuse the most recent trading day, intraday runs reuse the last completed trading day, and after market close the run skips only if the current trading day's data is already stored. See Full Guide for the exact rules.

Option 2: Local Deployment

1. Clone Repository

git clone https://github.com/ZhuLinsen/daily_stock_analysis.git
cd daily_stock_analysis

2. Install Dependencies

# Create virtual environment (recommended)
python3 -m venv venv
source venv/bin/activate  # Windows: venv\Scripts\activate

# Install dependencies
pip install -r requirements.txt

3. Configure Environment Variables

# Copy configuration template
cp .env.example .env

# Edit .env file
nano .env  # or use any editor

Configure the following:

# AI Model (Choose one)
GEMINI_API_KEY=your_gemini_api_key_here

# Stock Watchlist (Mixed markets supported)
STOCK_LIST=600519,AAPL,hk00700

# Notification Channel (Choose at least one)
TELEGRAM_BOT_TOKEN=your_bot_token
TELEGRAM_CHAT_ID=your_chat_id

# News Search (Optional)
TAVILY_API_KEYS=your_tavily_key

4. Run

# One-time analysis
python main.py

# Scheduled mode (runs daily at 18:00)
python main.py --schedule

# Analyze specific stocks
python main.py --stocks AAPL,TSLA,GOOGL

# Market review only
python main.py --market-review

API Endpoints

Endpoint	Method	Description
`/`	GET	Configuration page
`/health`	GET	Health check
`/analysis?code=xxx`	GET	Trigger async analysis for a single stock
`/analysis/history`	GET	Query analysis history records
`/tasks`	GET	Query all task statuses
`/task?id=xxx`	GET	Query a single task status

📱 Supported Notification Channels

1. Telegram (Recommended)

Talk to @BotFather → /newbot → get Bot Token
Get Chat ID: send a message to @userinfobot
Configure:

TELEGRAM_BOT_TOKEN=123456789:ABCdefGHIjklMNOpqrsTUVwxyz
TELEGRAM_CHAT_ID=123456789

2. Discord

Webhook:

DISCORD_WEBHOOK_URL=https://discord.com/api/webhooks/xxx/yyy

Bot:

DISCORD_BOT_TOKEN=your_bot_token
DISCORD_MAIN_CHANNEL_ID=your_channel_id

3. Slack

Bot (recommended, supports image upload; takes priority when both set):

SLACK_BOT_TOKEN=xoxb-...
SLACK_CHANNEL_ID=C01234567

Webhook (text only):

SLACK_WEBHOOK_URL=https://hooks.slack.com/services/T.../B.../xxx

4. Email

EMAIL_SENDER=your_email@gmail.com
EMAIL_PASSWORD=your_app_password
EMAIL_RECEIVERS=receiver@example.com  # Optional

5. WeChat Work / Feishu

WeChat Work:

WECHAT_WEBHOOK_URL=https://qyapi.weixin.qq.com/cgi-bin/webhook/send?key=xxx

Feishu:

FEISHU_WEBHOOK_URL=https://open.feishu.cn/open-apis/bot/v2/hook/xxx

6. PushPlus

PUSHPLUS_TOKEN=your_token_here

🎨 Sample Output

Decision Dashboard Format

# 🎯 2026-01-24 Decision Dashboard

> Total **3** stocks analyzed | 🟢Buy:1 🟡Hold:1 🔴Sell:1

## 📊 Analysis Summary

🟢 **AAPL(Apple Inc.)**: Buy | Score 85 | Strong Bullish
🟡 **600519(Kweichow Moutai)**: Hold | Score 65 | Bullish
🔴 **TSLA(Tesla)**: Sell | Score 35 | Bearish

---

## 🟢 AAPL (Apple Inc.)

### 📰 Key Information
**💭 Sentiment**: Positive news on iPhone 16 sales
**📊 Earnings**: Q1 2024 earnings beat expectations

### 📌 Core Conclusion

**🟢 Buy** | Strong Bullish

> **One-sentence Decision**: Strong technical setup with positive catalyst, ideal entry point

⏰ **Time Sensitivity**: Within this week

| Position | Action |
|----------|--------|
| 🆕 **No Position** | Buy at pullback |
| 💼 **With Position** | Continue holding |

### 📊 Data Perspective

**MA Alignment**: MA5>MA10>MA20 | Bull Trend: ✅ Yes | Trend Strength: 85/100

| Price Metrics | Value |
|--------------|-------|
| Current | $185.50 |
| MA5 | $183.20 |
| MA10 | $180.50 |
| MA20 | $177.80 |
| Bias (MA5) | +1.26% ✅ Safe |
| Support | $183.20 |
| Resistance | $190.00 |

**Volume**: Ratio 1.8 (Moderate increase) | Turnover 2.3%
💡 *Volume confirms bullish momentum*

### 🎯 Action Plan

**📍 Sniper Points**

| Level Type | Price |
|-----------|-------|
| 🎯 Ideal Entry | $183-184 |
| 🔵 Secondary Entry | $180-181 |
| 🛑 Stop Loss | $177 |
| 🎊 Target | $195 |

**💰 Position Sizing**: 20-30% of portfolio
- Entry Plan: Enter in 2-3 batches
- Risk Control: Strict stop loss at $177

**✅ Checklist**

- ✅ Bull trend confirmed
- ✅ Price near MA5 support
- ✅ Volume confirms trend
- ⚠️ Monitor market volatility

---

🔧 Advanced Configuration

Environment Variables

# === Analysis Behavior ===
ANALYSIS_DELAY=10              # Delay between analysis (seconds) to avoid API rate limit
REPORT_TYPE=full               # Report type: simple/full
SINGLE_STOCK_NOTIFY=true       # Push immediately after each stock analysis

# === Schedule ===
SCHEDULE_ENABLED=true          # Enable scheduled task
SCHEDULE_TIME=18:00            # Daily run time (HH:MM, 24-hour format)
MARKET_REVIEW_ENABLED=true     # Enable market review

# === Data Source ===
TUSHARE_TOKEN=your_token       # Tushare Pro (priority data source if configured)

# === System ===
MAX_WORKERS=3                  # Concurrent threads (3 recommended to avoid blocking)
DEBUG=false                    # Enable debug logging

🧩 FastAPI Web Service (Optional)

Enable the FastAPI service for configuration management and triggering analysis when running locally.

Startup Methods

Command	Description
`python main.py --serve`	Start API service + run full analysis once
`python main.py --serve-only`	Start API service only, manually trigger analysis

URL: http://127.0.0.1:8000
API docs: http://127.0.0.1:8000/docs

Features

📝 Configuration Management - View/modify watchlist
🚀 Quick Analysis - Trigger analysis via API
📊 Real-time Progress - Analysis task status updates in real-time, supports parallel tasks
🤖 Agent Strategy Chat - Use /ask, /chat, /history, /strategies, and /research for multi-turn Q&A, history, strategy listing, and deep research (enable with AGENT_MODE=true)
🧩 Intel compatibility - capital_flow_signal is an additive Intel output field; clients that do not consume it can ignore it safely, while existing fields such as risk_alerts and positive_catalysts remain unchanged.
📈 Backtest Validation - Evaluate historical analysis accuracy, query direction win rate and simulated returns

API Endpoints

Endpoint	Method	Description
`/api/v1/analysis/analyze`	POST	Trigger stock analysis
`/api/v1/analysis/tasks`	GET	Query task list
`/api/v1/analysis/status/{task_id}`	GET	Query task status
`/api/v1/history`	GET	Query analysis history
`/api/v1/backtest/run`	POST	Trigger backtest
`/api/v1/backtest/results`	GET	Query backtest results (paginated)
`/api/v1/backtest/performance`	GET	Get overall backtest performance
`/api/v1/backtest/performance/{code}`	GET	Get per-stock backtest performance
`/api/v1/agent/skills`	GET	Get available built-in/custom strategy skills
`/api/v1/agent/chat/stream`	POST (SSE)	Stream multi-turn Agent strategy chat
`/api/health`	GET	Health check

Note: POST /api/v1/analysis/analyze supports only one stock when async_mode=false; batch stock_codes requires async_mode=true. The async 202 response returns a single task_id for one stock, or an accepted / duplicates summary for batch requests.

For detailed instructions, see Full Guide - API Service

🔎 Smart Search Autocomplete (MVP)

The home analysis input now behaves more like a search box, reducing the need to memorize exact symbols.

Multi-signal matching: supports stock code, company name, pinyin abbreviation, and aliases (for example gzmt -> 贵州茅台, tencent -> 腾讯控股, aapl -> Apple Inc.).
Multi-market coverage: the local index now covers A-shares, Hong Kong stocks, and US stocks. It can be regenerated from either Tushare or AkShare source data when needed.
Graceful fallback:
- If the index is outdated, missing a newly listed symbol, or fails to load, the UI falls back to plain manual input without blocking analysis.
- If no suggestion matches, pressing Enter still submits the original input directly.

Tip: to refresh the index, run python3 scripts/fetch_tushare_stock_list.py to update the stock-list CSV files, then run python3 scripts/generate_index_from_csv.py to rebuild the static index.

📖 Documentation

☕ Support the Project

Alipay	WeChat Pay	Xiaohongshu

🤝 Contributing

Contributions are welcome! Please:

Fork this repository
Create your feature branch (git checkout -b feature/AmazingFeature)
Commit your changes (git commit -m 'Add some AmazingFeature')
Push to the branch (git push origin feature/AmazingFeature)
Open a Pull Request

See CONTRIBUTING.md for details.

📄 License

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

⭐ Star History

Made with ❤️ by AI enthusiasts | Star ⭐ this repo if you find it useful!

⚠️ Disclaimer

This tool is for informational and educational purposes only. The analysis results are generated by AI and should not be considered as investment advice. Stock market investments carry risk, and you should:

Do your own research before making investment decisions
Understand that past performance does not guarantee future results
Only invest money you can afford to lose
Consult with a licensed financial advisor for personalized advice

The developers of this tool are not liable for any financial losses resulting from the use of this software.

🙏 Acknowledgments

AkShare - Stock data source
Google Gemini - AI analysis engine
Tavily - News search API
All contributors who helped improve this project

📞 Contact

GitHub Issues: Report bugs or request features
Discussions: Join discussions
Email: zhuls345@gmail.com

FilesExpand file tree

README_EN.md

Latest commit

History

README_EN.md

File metadata and controls

AI Stock Analysis System

💖 Sponsors

✨ Key Features

Tech Stack & Data Sources

Built-in Trading Rules

🚀 Quick Start

Option 1: GitHub Actions (Recommended, Zero Cost)

1. Fork this repository

2. Configure Secrets

3. Enable Actions

4. Manual Test

5. Done!

Option 2: Local Deployment

1. Clone Repository

2. Install Dependencies

3. Configure Environment Variables

4. Run

API Endpoints

📱 Supported Notification Channels

1. Telegram (Recommended)

2. Discord

3. Slack

4. Email

5. WeChat Work / Feishu

6. PushPlus

🎨 Sample Output

Decision Dashboard Format

🔧 Advanced Configuration

Environment Variables

🧩 FastAPI Web Service (Optional)

Startup Methods

Features

API Endpoints

🔎 Smart Search Autocomplete (MVP)

📖 Documentation

☕ Support the Project

🤝 Contributing

📄 License

⭐ Star History

⚠️ Disclaimer

🙏 Acknowledgments

📞 Contact