llm-api-optimization

【效率与协作】Use when an LLM-backed application's API calls need to get cheaper, faster or more reliable — prompt caching and the prefix rules that decide whether it hits, batch processing, context engineering and what earns a place in the window, payload format economics, structured outputs and constrained decoding, model routing by task, streaming and latency, …

独立 GitHub 快照 · 自动分类 · 以原作者文档为准