Inject cache_control: {"type": "ephemeral"} on the system message and
last tool definition for providers that support prompt caching. Adds
supports_prompt_caching flag to ProviderSpec (enabled for Anthropic only)
and skips caching when routing through a gateway.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
- Add moonshot to ProvidersConfig schema
- Add MOONSHOT_API_BASE environment variable for custom endpoint
- Handle kimi-k2.5 model temperature restriction (must be 1.0)
- Fix is_vllm detection to exclude moonshot provider
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
LiteLLM expects the 'zai/' provider prefix for Zhipu AI (Z.ai) models,
not 'zhipu/'. This was causing 'LLM Provider NOT provided' errors when
users configured models like 'glm-4.7' without an explicit prefix.
According to LiteLLM docs, the correct format is:
- model='zai/glm-4.7' (correct)
- NOT model='zhipu/glm-4.7' (incorrect)
This fix ensures auto-prefixed models use the correct 'zai/' format.
Fixes: Error when using Zhipu AI models with shorthand names like 'glm-4.7'
- Update configuration schema to include Gemini provider
- Modify API key retrieval priority to include Gemini
- Enhance CLI status command to display Gemini API status
- Update LiteLLMProvider to support Gemini integration
- Add vllm provider configuration in config schema
- Auto-detect vLLM endpoints and use hosted_vllm/ prefix for LiteLLM
- Pass api_base directly to acompletion for custom endpoints
- Add vLLM status display in CLI status command
- Add vLLM setup documentation in README
