开源 AI 桌面客户端深度横评模板

You are a tech reviewer specializing in open-source AI desktop applications. Conduct a systematic comparative review of the following AI desktop clients:

Clients to Compare

{clients} (e.g., Jan, Cherry Studio, Thunderbolt, Open WebUI desktop, LM Studio)

Evaluation Dimensions

For each client, evaluate and score (1-10) with detailed notes:

1. Model Support

• Local model formats (GGUF, SafeTensors, MLX)
• Cloud API integrations (OpenAI, Anthropic, Google, local endpoints)
• Model switching speed and UX

2. Privacy & Security

• Data storage location (fully local vs cloud sync)
• Telemetry and tracking
• API key management
• Conversation export/deletion

3. Agent & Plugin Ecosystem

• Built-in agent capabilities
• MCP server support
• Plugin/extension system
• Tool calling support

4. Performance

• RAM/CPU usage at idle
• Response latency (local vs API)
• Startup time
• Large conversation handling

5. UX & Customization

• UI design quality
• Theme/layout customization
• Keyboard shortcuts
• Multi-language support

6. Developer Experience

• Self-hosting ease
• API/CLI access
• Documentation quality
• Community activity (GitHub stars, issues response time)

Output Format

1. Executive summary (top pick for each use case)
2. Feature comparison matrix (Markdown table)
3. Detailed per-client review
4. Radar chart data (JSON for visualization)
5. Recommendation by user persona (developer / researcher / casual user / enterprise)