master

Roman-Andrfeat(llm-client): thread cached input tokens through Usage into trace spans

02471552创建于 6月14日历史提交

文件	最后提交记录	最后更新时间
convert.rs	refactor: update reasoning effort handling in providers and documentation - Enhanced provider logic to conditionally include reasoning effort fields based on the selected effort level. - Updated README to clarify the behavior of provider mapping with respect to reasoning effort settings. - Added tests to ensure that reasoning effort fields are omitted when set to 'off' for OpenAI, Anthropic, and Gemini providers.	2 个月前
mod.rs	feat(llm-client): thread cached input tokens through Usage into trace spans Account for cache-read tokens end to end: - Usage carries cached_tokens (OpenAI prompt_tokens_details.cached_tokens; Anthropic cache_read_input_tokens), threaded into LLM_CALL / TURN trace spans. - Anthropic marks cache_control ephemeral on the stable system-prefix block ONLY, never the volatile Context block, so the breakpoint is read back across turns instead of being rewritten every turn. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>	1 个月前
types.rs	feat: add reasoning effort configuration to CLI and related components - Introduced ReasoningEffort enum to manage reasoning effort levels (Off, High, Max). - Updated CLI commands to accept reasoning_effort as an argument. - Enhanced configuration structures and session management to include reasoning effort. - Modified various components to utilize the new reasoning effort settings, ensuring consistent behavior across the application. - Added default values and serialization support for reasoning effort in relevant structs.	2 个月前