README.md

metadata

title: Anthropic Topic Segmentation Microservice
emoji: 🎯
colorFrom: blue
colorTo: purple
sdk: docker
app_file: gradio_app.py
pinned: false
license: mit

🎯 Anthropic Topic Segmentation Microservice

A production-ready microservice that uses Anthropic's Claude models for intelligent topic segmentation and business insight extraction from interview transcripts. Perfect for processing Czech e-commerce conversations and integrating with n8n workflows.

🚀 Live Demo on HuggingFace Spaces

Try the API directly: https://huggingface.co/spaces/Yeetek/anthropic-topic-segmentation

✨ Key Features

• 🤖 Anthropic Integration: Uses Claude-3.5-Sonnet for superior language understanding
• 🌍 Multi-Language: Perfect Czech, Slovak, and English processing
• 🏢 Business Intelligence: 11 specialized business categories
• 📊 Large Transcripts: Handles up to 1,500 sentences with sliding window processing
• 🔄 n8n Compatible: RESTful API with dynamic prompt injection
• 🎯 High Accuracy: 90%+ confidence scores with actionable insights
• 🐳 Docker Ready: Optimized for HuggingFace Spaces deployment

🎯 Perfect for E-commerce Analysis

Successfully processes Czech Shoptet integration discussions, extracting:

• Client Requirements (B2B/B2C differentiated)
• Technical Barriers and implementation challenges
• Solution Benefits and "aha moments"
• Performance Issues and optimization needs

🚀 Quick Start

Option 1: HuggingFace Spaces (Recommended)

1. Fork this Space or create a new one
2. Add your Anthropic API key to Spaces secrets:

   ANTHROPIC_API_KEY = sk-ant-api03-your-key-here
   

3. The Space will automatically build and deploy!

Option 2: Local Docker

# Clone the repository
git clone https://huggingface.co/spaces/Yeetek/anthropic-topic-segmentation
cd anthropic-topic-segmentation

# Create .env file
cp env.example .env
# Edit .env with your Anthropic API key

# Build and run
docker build -t anthropic-topic-segmentation .
docker run -p 7860:7860 --env-file .env anthropic-topic-segmentation

Option 3: Python Development

# Install dependencies
pip install -r requirements.txt

# Set environment variables
export ANTHROPIC_API_KEY=sk-ant-api03-your-key-here

# Run the server
uvicorn app:app --host 0.0.0.0 --port 7860

📡 API Usage

Health Check

curl -X POST https://yeetek-anthropic-topic-segmentation.hf.space/api/health \
  -H "Content-Type: application/json" \
  -d '{"data": []}'

Topic Extraction

curl -X POST https://yeetek-anthropic-topic-segmentation.hf.space/api/segment \
  -H "Content-Type: application/json" \
  -d '{
    "data": [
      "[{\"text\": \"Zákazníci požadují nestandardní úpravy košíku v Shoptetu.\", \"speaker\": \"Client\", \"start_time\": 2.01, \"end_time\": 8.45, \"sentence_index\": 1}]",
      "customer_call",
      "cs", 
      "E-commerce"
    ]
  }'

Interactive Documentation

• Gradio Interface: https://yeetek-anthropic-topic-segmentation.hf.space
• API Documentation: Use the "📖 API Reference" tab in the interface

🔧 n8n Integration

Perfect for workflow automation:

{
  "workflow_name": "Czech E-commerce Analysis",
  "http_request": {
    "method": "POST",
         "url": "https://yeetek-anthropic-topic-segmentation.hf.space/api/segment",
    "body": {
      "sentences": "{{ $json.transcript }}",
      "prompt_config": {
        "template": "customer_call",
        "language": "cs"
      }
    }
  }
}

📊 Business Categories

The system extracts topics into 11 specialized categories:

• 🎯 client_needs_b2b - B2B client requirements
• 🛒 client_needs_b2c - B2C customer needs
• 🚧 solution_barriers - Implementation obstacles
• ⚙️ technical_requirements - Technical specifications
• 💬 customer_feedback - Customer opinions
• 👥 employee_feedback - Internal insights
• ✅ solution_benefits - Positive outcomes
• 💡 aha_moments - Key breakthroughs
• 🏢 company_info - Business context
• 📝 additional_comments - Miscellaneous insights
• 🔄 general - Fallback category

🌍 Language Support

• Czech: Perfect diacritics and business terminology ✅
• Slovak: Similar processing to Czech ✅
• English: Full business context understanding ✅

📈 Performance Benchmarks

• Small Transcripts (10 sentences): ~5-8 seconds
• Medium Transcripts (100 sentences): ~15-30 seconds
• Large Transcripts (640+ sentences): ~30-60 seconds
• Confidence: 90%+ average accuracy
• Memory: Efficient processing with minimal footprint

🔒 Security & Validation

• ✅ Input validation (max 1,500 sentences)
• ✅ Prompt injection protection
• ✅ Safety scoring and content filtering
• ✅ Rate limiting and error handling
• ✅ CORS configuration for web integration

🛠 Configuration

Environment Variables

# Essential Configuration
ANTHROPIC_API_KEY=sk-ant-api03-your-key-here
ANTHROPIC_MODEL=claude-3-5-sonnet-20241022
ENVIRONMENT=production
LOG_LEVEL=INFO
PORT=7860

# Optional Configuration  
CORS_ORIGINS=*
MAX_SENTENCES=1500
REQUEST_TIMEOUT=300

Model Options

• claude-3-5-sonnet-20241022 (Recommended - Best quality)
• claude-3-5-haiku-20241022 (Faster, cost-effective)

📚 Examples

Check the examples/ directory for:

• sample_request.json - Czech e-commerce transcript
• sample_response.json - Expected API response format

🤝 Contributing

1. Fork the repository
2. Create a feature branch
3. Make your changes
4. Add tests if applicable
5. Submit a pull request

📄 License

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

🆘 Support

• Issues: Report bugs or request features
• Documentation: Check /docs endpoint for interactive API docs
• Health Check: Monitor service status at /health

🏆 Success Stories

• ✅ 640-sentence Czech transcripts processed successfully
• ✅ Shoptet e-commerce integration insights extracted
• ✅ 90% confidence business intelligence
• ✅ n8n workflow automation ready
• ✅ Production deployment on HuggingFace Spaces

---

Built with ❤️ using Anthropic Claude 3.5, FastAPI, and optimized for Czech e-commerce use cases.