[FEATURE] Integrate Ollama Local AI (llama3.1:8b) Support

[anhmtk]

📋 Description

Hiện tại StillMe chỉ support DeepSeek (cloud) và OpenAI (cloud). Mục tiêu là thêm support cho Ollama local AI (llama3.1:8b) để:

• Giảm cost khi dùng local model
• Faster response cho simple queries
• Privacy-first option (no data sent to cloud)

🎯 Current Status

• ✅ Code đã có trong legacy folder
• ✅ Ollama đã được install và pull model llama3.1:8b
• ❌ Chưa integrate vào backend API
• ❌ Smart router chưa hỗ trợ routing local vs cloud

🛠️ Technical Requirements

• Integrate Ollama client vào backend/api/main.py
• Implement Smart Router logic:
  • Simple queries → Ollama local (llama3.1:8b)
  • Complex queries → DeepSeek cloud
  • Fallback: Ollama → DeepSeek nếu Ollama fail
• Add Ollama health check endpoint
• Update generate_ai_response() để support Ollama

📁 Files to Modify

• backend/api/main.py - Add Ollama integration
• backend/services/ - Create Ollama client service (optional)
• dashboard.py - Update model selection
• .env.example - Add OLLAMA_BASE_URL config

🧪 Testing Requirements

• Test Ollama connection
• Test Smart Router logic
• Test fallback mechanism
• Test với simple vs complex queries

💡 Implementation Ideas

• Use requests hoặc httpx để call Ollama API
• Ollama endpoint: http://localhost:11434/api/generate
• Model: llama3.1:8b
• Smart routing based on:
  • Query complexity (length, keywords)
  • Estimated response time
  • Ollama availability

✅ Acceptance Criteria

• Ollama integration working
• Smart router correctly routes queries
• Fallback mechanism works
• Health check endpoint responds
• Dashboard model selection works

🤝 Willing to Contribute?

• Yes, tôi có thể implement
• Maybe, cần guidance
• No, chỉ muốn suggest

📚 References

• Ollama API: https://github.com/ollama/ollama/blob/main/docs/api.md
• Legacy code: legacy/stillme_core/adapters/ollama_client.py (reference)