README.md

---
title: Trading Analysis Platform
emoji: 📊
colorFrom: blue
colorTo: green
sdk: gradio
sdk_version: "6.3.0"
app_file: deployment/huggingface/app.py
pinned: false
license: mit
---

# 📊 Trading Analysis Platform

AI-powered multi-agent trading analysis platform with **HuggingFace Inference Providers** support for cost-effective LLM access.

## 🚀 Features

- **Multi-Agent Analysis**: 10 specialized AI agents for comprehensive market analysis
  - Technical: Indicator, Pattern, Trend, Decision
  - Fundamental: Fundamentals, Sentiment, News
  - Management: Technical Analyst, Risk Manager, Portfolio Manager

- **HuggingFace Inference Providers**: Access 22+ LLM providers through a single API
  - **Cost Optimization**: Use `:cheapest` routing for 90%+ cost savings
  - **Performance**: Use `:fastest` routing for best quality
  - **Free Tier**: Many providers offer free inference

- **Multi-Provider Support**:
  - HuggingFace Inference Providers (recommended)
  - OpenAI (GPT-4, GPT-3.5)
  - Anthropic (Claude)

- **Cost Tracking**: Real-time API cost monitoring and budget alerts
- **Phase-Based Analysis**: Customize analysis workflow (Technical, Fundamental, Sentiment, Research, Risk, Decision)
- **Investment Styles**: Long-term investing vs Swing trading presets

## 🔑 Setup Instructions

### Required Secrets

Add these in your Space's **Settings → Repository secrets**:

1. **HF_TOKEN** (Required)
   - Your HuggingFace API token
   - Get it from: https://huggingface.co/settings/tokens
   - Needed for Inference Providers access

2. **OPENAI_API_KEY** (Optional)
   - Only needed if using OpenAI provider
   - Get it from: https://platform.openai.com/api-keys

3. **ANTHROPIC_API_KEY** (Optional)
   - Only needed if using Anthropic/Claude
   - Get it from: https://console.anthropic.com/

💡 **For public Spaces**: Use only `HF_TOKEN` with `:cheapest` routing to avoid API costs

### Routing Policy Examples

The platform supports flexible routing policies for cost/performance optimization:

#### Cost-Optimized (Recommended for Public Spaces)
```
Provider: HuggingFace
Routing Policy: :cheapest
```
- Uses free-tier models when available
- Automatically selects lowest-cost provider
- 90%+ cost savings vs. OpenAI

#### Performance-Optimized
```
Provider: HuggingFace
Routing Policy: :fastest
```
- Uses highest-quality models
- Best for critical decisions
- Higher cost but better accuracy

#### Specific Provider
```
Provider: HuggingFace
Routing Policy: groq
```
- Route to specific provider (groq, together, replicate, etc.)
- Consistent performance
- Provider-specific pricing

#### Specific Model
```
Provider: HuggingFace
Routing Policy: meta-llama/Llama-3.3-70B-Instruct
```
- Use exact model
- Predictable results
- Model-specific pricing

### Budget Management

Set budget limits to control costs:

1. Open **Settings** tab
2. Expand **Budget Management**
3. Set budget limit (e.g., $5.00)
4. Configure alert thresholds (75%, 90%, 100%)
5. Enable "Require confirmation at limit"

The system will alert you as you approach your budget and provide cost-reduction tips.

## 📈 Usage

1. **Enter Ticker**: Type a stock symbol (e.g., AAPL, MSFT, BTC-USD)
2. **Select Style**: Choose Long-term or Swing trading
3. **Configure Phases**: Enable desired analysis phases
4. **Run Analysis**: Click "🔍 Analyze" and wait for results
5. **Review Reports**: Check Decision, Technical, Pattern, and Trend analysis tabs
6. **Monitor Costs**: View cost summary at bottom of page

## 🎯 Supported Assets

- **Stocks**: AAPL, MSFT, GOOGL, TSLA, etc.
- **Crypto**: BTC-USD, ETH-USD, etc.
- **Indices**: ^GSPC (S&P 500), ^DJI (Dow Jones)
- **Commodities**: GC=F (Gold), CL=F (Oil)

## 🔒 Privacy & Security

- **Private Spaces**: Set Space to "Private" to restrict access
- **Public Spaces**: Use HuggingFace Inference Providers with `:cheapest` routing
  - Avoid adding OpenAI/Anthropic keys to public Spaces
  - Free-tier models prevent unexpected costs

## 🛠️ Advanced Configuration

### Per-Agent Routing

Configure different routing policies for each agent:

1. Open **Settings** → **Per-Agent Routing Configuration**
2. Set policies per agent:
   - Fast agents (Indicator, Pattern) → `:cheapest`
   - Decision agents (Decision, Portfolio) → `:fastest`
3. Apply presets:
   - **Cost Optimized**: All agents use `:cheapest`
   - **Performance Optimized**: All agents use `:fastest`
   - **Balanced**: Smart mix of cheap and premium

### Export/Import Configuration

Save and share your configuration:

1. **Export**: Click "Export Config" to get JSON
2. **Import**: Paste JSON and click "Import Config"
3. **Share**: Send JSON to team members

## 📊 Cost Comparison

| Provider | Model | Cost per 1M tokens | Free Tier |
|----------|-------|-------------------|-----------|
| HF Inference (:cheapest) | Various | $0.00 - $0.20 | ✅ Yes |
| HF Inference (groq) | Llama 3.3 70B | $0.05 - $0.10 | ✅ Yes |
| OpenAI | GPT-4 | $30.00 | ❌ No |
| OpenAI | GPT-3.5-turbo | $1.50 | ❌ No |
| Anthropic | Claude 3.5 Sonnet | $3.00 | ❌ No |

💡 **Recommendation**: Use `:cheapest` routing for 90%+ cost savings

## 🤝 Contributing

This platform is part of the trading-tools project. For issues or contributions, visit the GitHub repository.

## 📜 License

MIT License - See LICENSE file for details

## 🙏 Credits

Built with:
- **LangGraph** & **LangChain**: Multi-agent orchestration
- **HuggingFace Inference Providers**: Multi-provider LLM access
- **Gradio**: Web interface framework
- **yfinance**: Market data
- **TA-Lib/pandas-ta**: Technical indicators

---

**Note**: This is an AI-powered analysis tool. Always perform your own due diligence before making investment decisions.