<?xml version="1.0" encoding="UTF-8"?><rss xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:atom="http://www.w3.org/2005/Atom" version="2.0"><channel><title><![CDATA[生产知识库最该先监控什么？]]></title><description><![CDATA[<p dir="auto">我们知识库准备给全公司试用。除了接口错误率，还应该监控什么？</p>
]]></description><link>https://localaihub.com/topic/77/生产知识库最该先监控什么</link><generator>RSS for Node</generator><lastBuildDate>Wed, 03 Jun 2026 19:15:55 GMT</lastBuildDate><atom:link href="https://localaihub.com/topic/77.rss" rel="self" type="application/rss+xml"/><pubDate>Mon, 04 May 2026 10:57:00 GMT</pubDate><ttl>60</ttl><item><title><![CDATA[Reply to 生产知识库最该先监控什么？ on Tue, 05 May 2026 16:01:00 GMT]]></title><description><![CDATA[<p dir="auto">旧帖补后续：我们上线第一周最大问题不是模型，是 17% 的文档同步失败。监控救了一命。</p>
]]></description><link>https://localaihub.com/post/438</link><guid isPermaLink="true">https://localaihub.com/post/438</guid><dc:creator><![CDATA[周五下午]]></dc:creator><pubDate>Tue, 05 May 2026 16:01:00 GMT</pubDate></item><item><title><![CDATA[Reply to 生产知识库最该先监控什么？ on Tue, 05 May 2026 13:41:00 GMT]]></title><description><![CDATA[<p dir="auto">对。别把所有拒答都当坏事，也别把所有回答都当成功。</p>
]]></description><link>https://localaihub.com/post/437</link><guid isPermaLink="true">https://localaihub.com/post/437</guid><dc:creator><![CDATA[nora]]></dc:creator><pubDate>Tue, 05 May 2026 13:41:00 GMT</pubDate></item><item><title><![CDATA[Reply to 生产知识库最该先监控什么？ on Tue, 05 May 2026 12:00:00 GMT]]></title><description><![CDATA[<p dir="auto">所以要把“拒答正确”也纳入监控。</p>
]]></description><link>https://localaihub.com/post/436</link><guid isPermaLink="true">https://localaihub.com/post/436</guid><dc:creator><![CDATA[周五下午]]></dc:creator><pubDate>Tue, 05 May 2026 12:00:00 GMT</pubDate></item><item><title><![CDATA[Reply to 生产知识库最该先监控什么？ on Tue, 05 May 2026 10:19:00 GMT]]></title><description><![CDATA[<p dir="auto">不一定。能明确拒答、提示缺资料，也算健康行为。</p>
]]></description><link>https://localaihub.com/post/435</link><guid isPermaLink="true">https://localaihub.com/post/435</guid><dc:creator><![CDATA[林小北]]></dc:creator><pubDate>Tue, 05 May 2026 10:19:00 GMT</pubDate></item><item><title><![CDATA[Reply to 生产知识库最该先监控什么？ on Tue, 05 May 2026 09:26:00 GMT]]></title><description><![CDATA[<p dir="auto">用户问了知识库没有的东西，算失败吗？</p>
]]></description><link>https://localaihub.com/post/434</link><guid isPermaLink="true">https://localaihub.com/post/434</guid><dc:creator><![CDATA[小树]]></dc:creator><pubDate>Tue, 05 May 2026 09:26:00 GMT</pubDate></item><item><title><![CDATA[Reply to 生产知识库最该先监控什么？ on Tue, 05 May 2026 08:53:00 GMT]]></title><description><![CDATA[<p dir="auto">引用点击率挺有意思。用户经常点引用，说明他们在核查；引用点不开就是严重问题。</p>
]]></description><link>https://localaihub.com/post/433</link><guid isPermaLink="true">https://localaihub.com/post/433</guid><dc:creator><![CDATA[阿白]]></dc:creator><pubDate>Tue, 05 May 2026 08:53:00 GMT</pubDate></item><item><title><![CDATA[Reply to 生产知识库最该先监控什么？ on Tue, 05 May 2026 05:51:00 GMT]]></title><description><![CDATA[<p dir="auto">延迟也要按阶段。总耗时高时，你要知道是向量库慢还是 LLM 慢。</p>
]]></description><link>https://localaihub.com/post/432</link><guid isPermaLink="true">https://localaihub.com/post/432</guid><dc:creator><![CDATA[MingK]]></dc:creator><pubDate>Tue, 05 May 2026 05:51:00 GMT</pubDate></item><item><title><![CDATA[Reply to 生产知识库最该先监控什么？ on Tue, 05 May 2026 03:08:00 GMT]]></title><description><![CDATA[<p dir="auto">自动指标只能告诉你哪里可能坏，不能替代人工复盘。</p>
]]></description><link>https://localaihub.com/post/431</link><guid isPermaLink="true">https://localaihub.com/post/431</guid><dc:creator><![CDATA[小路灯]]></dc:creator><pubDate>Tue, 05 May 2026 03:08:00 GMT</pubDate></item><item><title><![CDATA[Reply to 生产知识库最该先监控什么？ on Tue, 05 May 2026 00:13:00 GMT]]></title><description><![CDATA[<p dir="auto">我们做了一个“坏答案回放”页面，产品每天挑 10 条看。比看大盘有用。</p>
]]></description><link>https://localaihub.com/post/430</link><guid isPermaLink="true">https://localaihub.com/post/430</guid><dc:creator><![CDATA[米饭]]></dc:creator><pubDate>Tue, 05 May 2026 00:13:00 GMT</pubDate></item><item><title><![CDATA[Reply to 生产知识库最该先监控什么？ on Mon, 04 May 2026 21:18:00 GMT]]></title><description><![CDATA[<p dir="auto">日志别存用户无权内容全文，尤其多租户。可观测不是泄露理由。</p>
]]></description><link>https://localaihub.com/post/429</link><guid isPermaLink="true">https://localaihub.com/post/429</guid><dc:creator><![CDATA[leaf_1997]]></dc:creator><pubDate>Mon, 04 May 2026 21:18:00 GMT</pubDate></item><item><title><![CDATA[Reply to 生产知识库最该先监控什么？ on Mon, 04 May 2026 18:47:00 GMT]]></title><description><![CDATA[<p dir="auto">每个答案记录 request_id、query、retrieved_chunk_ids、model、prompt_version、index_version。注意脱敏。</p>
]]></description><link>https://localaihub.com/post/428</link><guid isPermaLink="true">https://localaihub.com/post/428</guid><dc:creator><![CDATA[rootless]]></dc:creator><pubDate>Mon, 04 May 2026 18:47:00 GMT</pubDate></item><item><title><![CDATA[Reply to 生产知识库最该先监控什么？ on Mon, 04 May 2026 16:24:00 GMT]]></title><description><![CDATA[<p dir="auto">还有数据新鲜度。最新文件没进库，比模型慢更致命。</p>
]]></description><link>https://localaihub.com/post/427</link><guid isPermaLink="true">https://localaihub.com/post/427</guid><dc:creator><![CDATA[林小北]]></dc:creator><pubDate>Mon, 04 May 2026 16:24:00 GMT</pubDate></item><item><title><![CDATA[Reply to 生产知识库最该先监控什么？ on Mon, 04 May 2026 14:56:00 GMT]]></title><description><![CDATA[<p dir="auto">RAG 质量监控至少要有召回命中率抽检、无答案率、引用失败率、用户点踩原因。</p>
]]></description><link>https://localaihub.com/post/426</link><guid isPermaLink="true">https://localaihub.com/post/426</guid><dc:creator><![CDATA[nora]]></dc:creator><pubDate>Mon, 04 May 2026 14:56:00 GMT</pubDate></item><item><title><![CDATA[Reply to 生产知识库最该先监控什么？ on Mon, 04 May 2026 12:47:00 GMT]]></title><description><![CDATA[<p dir="auto">链路拆开：解析、入库、检索、rerank、生成、引用、反馈。只看接口 200 没用。</p>
]]></description><link>https://localaihub.com/post/425</link><guid isPermaLink="true">https://localaihub.com/post/425</guid><dc:creator><![CDATA[小乔同学]]></dc:creator><pubDate>Mon, 04 May 2026 12:47:00 GMT</pubDate></item></channel></rss>