📄 中文摘要
Google Gemini 的上下文缓存功能能够在输入后缓存上下文，并在后续请求中重用这些缓存的内容。缓存的令牌使用成本约为标准费用的 25%，从而实现显著的成本节约。该功能的基本规格包括默认缓存有效期为 3600 秒（1 小时）、最小令牌要求为 32,768 令牌，以及与模型名称结合管理的缓存。上下文缓存在大规模数据库分析中表现尤为突出，特别是在使用 FTS5+BM25 构建的 SQLite 数据库时，能够加速关键词提取和高精度答案生成的过程。
通过 Gemini 上下文缓存降低大规模文档分析的 API 成本
📄 English Summary
Reduce API Costs for Large-Scale Document Analysis with Gemini Context Caching
Google Gemini's Context Caching feature allows for caching context after input and reusing it in subsequent requests. Cached tokens can be utilized at approximately 25% of the standard rate, leading to significant cost savings. Basic specifications include a default cache validity period of 3,600 seconds (1 hour), a minimum token requirement of 32,768 tokens, and cache management in conjunction with the model name. Context Caching is particularly effective in large-scale database analysis, especially when analyzing an SQLite DB built with FTS5+BM25, facilitating fast keyword extraction and high-accuracy answer generation.