rag-retrieval skillA
rag-retrieval is agent-read markdown (skill) from yuecao365/offercome: 检索增强深挖：文档解析与分块、Embedding 与混合检索、重排与改写、评测集、生产化。.
Indexed from public GitHub and served as immutable, content-addressed versions. Install it pinned to an exact SHA-256 with the mdr CLI, and every file is verified against the hash recorded here before it reaches your agent. The deterministic audit below grades the latest version, and the same file always earns the same grade.
What the file says
## 面试官在意什么 RAG 是大模型应用里最容易做出 demo、最难做到可用的一段。面试官默认候选人的"RAG 项目"是教程改了数据源，所以从三个地方验：有没有检索评测集与召回率数字、能不能把"没检索到"和"检索到了答错"分开排查、知不知道解析与分块才是上限。2026 年长上下文便宜了很多，面试官还会追"为什么不直接把文档塞进上下文"，答不出边界的算没想过。 ## 项目 / 实习怎么深挖 简历上出现下面这类经历时从哪里切、追什么。追到候选人能说出机制、数字的来源与一次真实的故障或取舍才算实；只有框架名与结论、说不出自己那一段的，记为危险信号。通用的追问方法见 project-deep-dive。 - 简历出现 RAG 项目 → 追检索评测集怎么建、召回率与上下文精确率多少、最典型的失败案例与最终归因、为什么不直接长上下文 - 简历出现向量库 → 追索引类型与参数、数据量、更新怎么做、换 embedding 模型时怎么迁移 - 简历出现"准确率提升 X%" → 追指标定义、样本数、是不是自己标的、有没有对照 - 简历出现重排 / 混合检索 / 查询改写 → 追各自解决了哪一类失败、加了多少延迟、有没有关掉过 - 简历出现 PDF / 表格 / 多模态解析 → 追解析质量怎么抽检、复杂版式怎么处理、解析错误占失败的比例 - 简历出现 GraphRAG / 知识图谱 → 追图从哪来、维护成本、和普通检索对比的实测差异 ## 常见失守与危险信号 - 文档解析与分块：把所有 PDF 一律转纯文本；分块只按固定长度；没意识到解析质量是上限 - Embedding 与混合检索：认为 embedding 越大越好；不知道 BM25 与稠密的互补；没测过 top-k 召回率就说效果好 - 重排与查询改写：把"加 rerank"当万能药说不出它解决的是哪一段；改写后没评测 - 检索评测集：从来没有 query-正确文档对；评估靠"看起来不错" - 失败归因：分不清召回失败与生成失败；排查没有顺序 - 生产化：索引更新靠全量重建且没有计划；权限过滤在生成之后做；不知道缓存能省哪一段 ## 常考主题清单 只列名字、阶梯与答实的标志，作"问到哪一层算实"的参考；问哪些、问几道由这份 JD 与这份简历定，不是配额。 ### RAG 链路设计与失败归因 - 阶梯：为什么不直接把文档塞进长上下文 → 分块策略、embedding 模型、向量库索引（HNSW/IVF）各自影响什么 → "检索到了但答错"与"根本没检索到"的区分方法与各自排查顺序 → 混合检索（稠密 + BM25、RRF 融合）、重排、查询改写各自的收益与延迟成本，什么规模值得上 …
Read the whole file at its exact version.
How to install
mdr add yuecao365/offercome/rag-retrieval@git:20260920.a4843f5mdr add yuecao365/offercome/rag-retrieval@sha256:2ef0400a0a589dfePin to a label to follow the author's releases, or to a sha256 to freeze the exact bytes forever. Either way the resolved hash is written to mdr.lock, and mdr install reproduces it on any machine.
[](https://markdownregistry.com/a/art_mhbydhu325lboaji)
1 badge views in 30 days
Versions
Audit of the latest version
- pass: Frontmatter block present
- pass: Frontmatter declares a name
- pass: Frontmatter declares a description
- pass: Size between 200 bytes and 200 KB (7409 bytes)
- pass: No zero-width or bidi control characters
- pass: No instruction hidden inside an HTML comment
- pass: No link to an exfiltration or paste host
- pass: No credential-shaped string
- pass: No instruction to send local credentials anywhere
- pass: No text hidden with inline styles
- pass: No prompt-injection phrasing
- pass: No curl or wget piped into a shell
- pass: No recursive delete of root, home or parent
- pass: No instruction to read or print local credentials
- pass: No base64 blob over 200 characters
- pass: No link to a raw IP address
- pass: No script tag
Source
yuecao365/offercome · 23 stars · license MIT · pushed 2026-09-23 · branch main
API
GET https://markdownregistry.com/api/v1/artifacts/art_mhbydhu325lboaji GET https://markdownregistry.com/api/v1/resolve?ref=yuecao365/offercome/rag-retrieval GET https://markdownregistry.com/api/v1/blob/2ef0400a0a589dfef4210a1a935261cb8b1c52437f443939dd9007db9a52195f
Your agent does the legwork. You hear about the deals worth your word. Hand yours the standing instructions at modelranch.com and it joins the network that reads files like this one.
More from yuecao365/offercome
Every file in yuecao365/offercome