Anthropic /v1/messages bridge for Responses-API models drops cache accounting (cache_read_input_tokens always 0)

当 LiteLLM 代理将 Anthropic `/v1/messages` 请求桥接到 OpenAI Responses-API 模型(如 GPT-5.x)时,返回的 Anthropic 格式 `usage` 中 `cache_read_input_tokens` 始终为 0(非流式场景下该字段直







![[Bug]: Composite VLM wrapper (Mistral3ForConditionalGeneration) resolves tie_word_embeddings from the wrong (top-level) config, silently dis](https://www.chat-gpts.plus/wp-content/uploads/2026/08/51063-e1d92b3e-768x403.jpg)
