provider-config

# Provider Configuration Skill Configure and integrate AI providers into the mimo2codex proxy. ## When to use Trigger this skill when: - User wants to add a new provider (e.g., "接入MiniMax", "add a new provider") - User provides provider configuration JSON - User asks to configure a provider (e.g., "按要求配置") - User has provider connectivity issues - User asks about provider compatibility ## Workflow ### 1. Understand Provider Requirements Gather from user: - **Base URL**: API endpoint (e.g., `https://api.example.com/v1`) - **API Key**: Authentication method - **Models**: Which models to expose - **Wire API**: OpenAI-compatible? Custom format? - **Special features**: Vision, tools, streaming, etc. ### 2. Check Existing Provider Patterns Look at `src/providers/` for examples: - `mimo.ts` - MiMo provider (built-in) - `deepseek.ts` - DeepSeek provider (built-in) - `registry.ts` - Provider registration - `types.ts` - Provider type definitions ### 3. Create Provider Configuration For generic OpenAI-compatible providers: ```typescript // In src/providers/registry.ts or config { id: "provider-name", shortcut: "provider-name", displayName: "Provider Display Name", baseUrl: "https://api.example.com/v1", models: ["model-1", "model-2"], wireApi: "openai-chat", // or "custom" apiKeyEnv: "PROVIDER_API_KEY", } ``` ### 4. Handle Provider Quirks Common compatibility issues: - **`max_tokens` vs `max_completion_tokens`**: Some providers use different field names - **Image input**: Not all providers support vision - **Tool calling**: Some providers have limited tool support - **Streaming**: Check SSE format compatibility - **Rate limiting**: Implement retry logic if needed ### 5. Test Provider Integration ```bash # Test with direct API call curl -X POST "$BASE_URL/chat/completions" \ -H "Authorization: Bearer $API_KEY" \ -H "Content-Type: application/json" \ -d '{"model":"model-name","messages":[{"role":"user","content":"Hello"}]}' # Test through mimo2codex curl -X POST http://localhost:8788/v1/chat/completions \ -H "Authorization: Bearer your-key" \ -H "Content-Type: application/json" \ -d '{"model":"provider-name/model-name","messages":[{"role":"user","content":"Hello"}]}' ``` ### 6. Document Provider Add to provider documentation: - Base URL and auth method - Supported models and capabilities - Known quirks or limitations - Example configuration ## Example Usage ``` User: 按要求配置：{ "providers": [ { "id": "sensenova", "shortcut": "sensenova", "displayName": "商汤日日新", "baseUrl": "..." } ] } Agent workflow: 1. Parse the provider configuration JSON 2. Check if provider needs special handling 3. Add to registry.ts or config 4. Test the integration 5. Document any quirks found ``` ## Provider Compatibility Matrix | Provider | Wire API | Vision | Tools | Streaming | Notes | |----------|----------|--------|-------|-----------|-------| | MiMo | openai-chat | ✅ | ✅ | ✅ | Built-in, `max_completion_tokens` | | DeepSeek | openai-chat | ❌ | ✅ | ✅ | Built-in, `reasoning_content` | | MiniMax | openai-chat | ✅ | ✅ | ✅ | May need `dropResponseFormat` | | Generic | openai-chat | ? | ? | ? | Depends on implementation | ## Common Configuration Patterns ### For OpenAI-Compatible APIs ```json { "id": "my-provider", "shortcut": "my-provider", "displayName": "My AI Provider", "baseUrl": "https://api.my-provider.com/v1", "models": ["gpt-4", "gpt-3.5-turbo"], "wireApi": "openai-chat", "apiKeyEnv": "MY_PROVIDER_API_KEY" } ``` ### For Custom APIs ```json { "id": "custom-ai", "shortcut": "custom-ai", "displayName": "Custom AI", "baseUrl": "https://custom-ai.example.com/api", "models": ["custom-model-v1"], "wireApi": "custom", "apiKeyEnv": "CUSTOM_AI_KEY", "transformRequest": "custom-transform.ts", "transformResponse": "custom-transform.ts" } ``` ## Troubleshooting ### "Unauthorized" / 401 Errors - Check API key is correct - Verify API key has proper permissions - Check if key needs specific header format ### "Model Not Found" / 404 Errors - Verify model name matches provider's expected format - Check if model is available in your region - Confirm model is enabled in your account ### Streaming Issues - Check if provider supports SSE format - Verify `stream: true` in request - Look for newline-delimited JSON issues ### Tool Calling Issues - Verify provider supports function calling - Check tool schema format compatibility - Look for `tool_choice` support ## Key Commands Reference ```bash # Print current config node dist/cli.js print-config # Print cc-switch config node dist/cli.js print-cc-switch # Test provider directly curl -X POST "$BASE_URL/chat/completions" \ -H "Authorization: Bearer $API_KEY" \ -H "Content-Type: application/json" \ -d '{"model":"model-name","messages":[{"role":"user","content":"test"}]}' # Check mimo2codex logs tail -f ~/.mimo2codex/logs/proxy.log ``` ## Don't Use This Skill For - General API calls (use mimo_chat.py from mimoskill) - PR review (use pr-review skill) - Issue investigation (use issue-investigation skill) - Codebase analysis (use codebase-analysis skill)