# milvus - Doramagic AI Context Pack

> 定位：安装前体验与判断资产。它帮助宿主 AI 有一个好的开始，但不代表已经安装、执行或验证目标项目。

## 充分原则

- **充分原则，不是压缩原则**：AI Context Pack 应该充分到让宿主 AI 在开工前理解项目价值、能力边界、使用入口、风险和证据来源；它可以分层组织，但不以最短摘要为目标。
- **压缩策略**：只压缩噪声和重复内容，不压缩会影响判断和开工质量的上下文。

## 给宿主 AI 的使用方式

你正在读取 Doramagic 为 milvus 编译的 AI Context Pack。请把它当作开工前上下文：帮助用户理解适合谁、能做什么、如何开始、哪些必须安装后验证、风险在哪里。不要声称你已经安装、运行或执行了目标项目。

## Claim 消费规则

- **事实来源**：Repo Evidence + Claim/Evidence Graph；Human Wiki 只提供显著性、术语和叙事结构。
- **事实最低状态**：`supported`
- `supported`：可以作为项目事实使用，但回答中必须引用 claim_id 和证据路径。
- `weak`：只能作为低置信度线索，必须要求用户继续核实。
- `inferred`：只能用于风险提示或待确认问题，不能包装成项目事实。
- `unverified`：不得作为事实使用，应明确说证据不足。
- `contradicted`：必须展示冲突来源，不得替用户强行选择一个版本。

## 它最适合谁

- **AI 研究者或研究型 Agent 构建者**：README 明确围绕研究、实验或论文工作流展开。 证据：`README.md` Claim：`clm_0002` supported 0.86

## 它能做什么

- **命令行启动或安装流程**（需要安装后验证）：项目文档中存在可执行命令，真实使用需要在本地或宿主环境中运行这些命令。 证据：`DEVELOPMENT.md`, `docs/design-docs/design_docs/20250825-geometry.md`, `docs/design-docs/design_docs/cdc/20260304-cdc-per-cluster-mtls-user-guide.md`, `scripts/download_milvus_proto.sh` 等 Claim：`clm_0001` supported 0.86

## 怎么开始

- `curl --cacert /certs/ca-dev1.pem \` 证据：`docs/design-docs/design_docs/cdc/20260304-cdc-per-cluster-mtls-user-guide.md` Claim：`clm_0003` unverified 0.25
- `curl --request POST \` 证据：`docs/design-docs/design_docs/20250825-geometry.md` Claim：`clm_0004` supported 0.86
- `pip install conan==2.25.1` 证据：`DEVELOPMENT.md` Claim：`clm_0005` unverified 0.25
- `pipx install conan==2.25.1` 证据：`DEVELOPMENT.md` Claim：`clm_0006` unverified 0.25
- `pipx install conan==1.66.0 --suffix=-1` 证据：`DEVELOPMENT.md` Claim：`clm_0007` unverified 0.25
- `pip install -r requirements.txt` 证据：`DEVELOPMENT.md` Claim：`clm_0008` unverified 0.25
- `git clone "$PROTO_REPO"` 证据：`scripts/download_milvus_proto.sh` Claim：`clm_0009` unverified 0.25
- `git clone https://github.com/wjakob/tbb.git && \` 证据：`scripts/install_deps_embd.sh` Claim：`clm_0010` unverified 0.25
- `curl https://sh.rustup.rs -sSf | sh -s -- --default-toolchain=${RUST_VERSION} -y || {` 证据：`scripts/install_deps.sh` Claim：`clm_0011` unverified 0.25
- `curl -sfL https://raw.githubusercontent.com/milvus-io/milvus/master/scripts/standalone_embed.sh -o standalone_embed_latest.sh && \` 证据：`scripts/standalone_embed.sh` Claim：`clm_0012` unverified 0.25

## 继续前判断卡

- **当前建议**：先做角色匹配试用
- **为什么**：这个项目更像角色库，核心风险是选错角色或把角色文案当执行能力；先用 Prompt Preview 试角色匹配，再决定是否沙盒导入。

### 30 秒判断

- **现在怎么做**：先做角色匹配试用
- **最小安全下一步**：先用 Prompt Preview 试角色匹配；满意后再隔离导入
- **先别相信**：角色质量和任务匹配不能直接相信。
- **继续会触碰**：角色选择偏差、命令执行、宿主 AI 配置

### 现在可以相信

- **适合人群线索：AI 研究者或研究型 Agent 构建者**（supported）：有 supported claim 或项目证据支撑，但仍不等于真实安装效果。 证据：`README.md` Claim：`clm_0002` supported 0.86
- **能力存在：命令行启动或安装流程**（supported）：可以相信项目包含这类能力线索；是否适合你的具体任务仍要试用或安装后验证。 证据：`DEVELOPMENT.md`, `docs/design-docs/design_docs/20250825-geometry.md`, `docs/design-docs/design_docs/cdc/20260304-cdc-per-cluster-mtls-user-guide.md`, `scripts/download_milvus_proto.sh` 等 Claim：`clm_0001` supported 0.86
- **存在 Quick Start / 安装命令线索**（supported）：可以相信项目文档出现过启动或安装入口；不要因此直接在主力环境运行。 证据：`docs/design-docs/design_docs/20250825-geometry.md` Claim：`clm_0004` supported 0.86

### 现在还不能相信

- **角色质量和任务匹配不能直接相信。**（unverified）：角色库证明有很多角色，不证明每个角色都适合你的具体任务，也不证明角色能产生高质量结果。
- **不能把角色文案当成真实执行能力。**（unverified）：安装前只能判断角色描述和任务画像是否匹配，不能证明它能在宿主 AI 里完成任务。
- **真实输出质量不能在安装前相信。**（unverified）：Prompt Preview 只能展示引导方式，不能证明真实项目中的结果质量。
- **宿主 AI 版本兼容性不能在安装前相信。**（unverified）：Claude、Cursor、Codex、Gemini 等宿主加载规则和版本差异必须在真实环境验证。
- **不会污染现有宿主 AI 行为，不能直接相信。**（inferred）：Skill、plugin、AGENTS/CLAUDE/GEMINI 指令可能改变宿主 AI 的默认行为。 证据：`CLAUDE.md`
- **可安全回滚不能默认相信。**（unverified）：除非项目明确提供卸载和恢复说明，否则必须先在隔离环境验证。
- **真实安装后是否与用户当前宿主 AI 版本兼容？**（unverified）：兼容性只能通过实际宿主环境验证。
- **项目输出质量是否满足用户具体任务？**（unverified）：安装前预览只能展示流程和边界，不能替代真实评测。

### 继续会触碰什么

- **角色选择偏差**：用户对任务应该由哪个专家角色处理的判断。 原因：选错角色会让 AI 从错误专业视角回答，浪费时间或误导决策。
- **命令执行**：包管理器、网络下载、本地插件目录、项目配置或用户主目录。 原因：运行第一条命令就可能产生环境改动；必须先判断是否值得跑。 证据：`DEVELOPMENT.md`, `docs/design-docs/design_docs/20250825-geometry.md`, `docs/design-docs/design_docs/cdc/20260304-cdc-per-cluster-mtls-user-guide.md`, `scripts/download_milvus_proto.sh` 等
- **宿主 AI 配置**：Claude/Codex/Cursor/Gemini/OpenCode 等宿主的 plugin、Skill 或规则加载配置。 原因：宿主配置会改变 AI 后续工作方式，可能和用户已有规则冲突。 证据：`CLAUDE.md`
- **本地环境或项目文件**：安装结果、插件缓存、项目配置或本地依赖目录。 原因：安装前无法证明写入范围和回滚方式，需要隔离验证。 证据：`DEVELOPMENT.md`, `docs/design-docs/design_docs/20250825-geometry.md`, `docs/design-docs/design_docs/cdc/20260304-cdc-per-cluster-mtls-user-guide.md`, `scripts/download_milvus_proto.sh` 等
- **宿主 AI 上下文**：AI Context Pack、Prompt Preview、Skill 路由、风险规则和项目事实。 原因：导入上下文会影响宿主 AI 后续判断，必须避免把未验证项包装成事实。

### 最小安全下一步

- **先跑 Prompt Preview**：先用交互式试用验证任务画像和角色匹配，不要先导入整套角色库。（适用：任何项目都适用，尤其是输出质量未知时。）
- **只在隔离目录或测试账号试装**：避免安装命令污染主力宿主 AI、真实项目或用户主目录。（适用：存在命令执行、插件配置或本地写入线索时。）
- **先备份宿主 AI 配置**：Skill、plugin、规则文件可能改变 Claude/Cursor/Codex 的默认行为。（适用：存在插件 manifest、Skill 或宿主规则入口时。）
- **安装后只验证一个最小任务**：先验证加载、兼容、输出质量和回滚，再决定是否深用。（适用：准备从试用进入真实工作流时。）

### 退出方式

- **保留安装前状态**：记录原始宿主配置和项目状态，后续才能判断是否可恢复。
- **准备移除宿主 plugin / Skill / 规则入口**：如果试装后行为异常，可以把宿主 AI 恢复到试装前状态。
- **保留原始角色选择记录**：如果输出偏题，可以回到任务画像阶段重新选择角色，而不是继续沿着错误角色推进。
- **记录安装命令和写入路径**：没有明确卸载说明时，至少要知道哪些目录或配置需要手动清理。
- **如果没有回滚路径，不进入主力环境**：不可回滚是继续前阻断项，不应靠信任或运气继续。

## 哪些只能预览

- 解释项目适合谁和能做什么
- 基于项目文档演示典型对话流程
- 帮助用户判断是否值得安装或继续研究

## 哪些必须安装后验证

- 真实安装 Skill、插件或 CLI
- 执行脚本、修改本地文件或访问外部服务
- 验证真实输出质量、性能和兼容性

## 边界与风险判断卡

- **把安装前预览误认为真实运行**：用户可能高估项目已经完成的配置、权限和兼容性验证。 处理方式：明确区分 prompt_preview_can_do 与 runtime_required。 Claim：`clm_0013` inferred 0.45
- **命令执行会修改本地环境**：安装命令可能写入用户主目录、宿主插件目录或项目配置。 处理方式：先在隔离环境或测试账号中运行。 证据：`DEVELOPMENT.md`, `docs/design-docs/design_docs/20250825-geometry.md`, `docs/design-docs/design_docs/cdc/20260304-cdc-per-cluster-mtls-user-guide.md`, `scripts/download_milvus_proto.sh` 等 Claim：`clm_0014` supported 0.86
- **待确认**：真实安装后是否与用户当前宿主 AI 版本兼容？。原因：兼容性只能通过实际宿主环境验证。
- **待确认**：项目输出质量是否满足用户具体任务？。原因：安装前预览只能展示流程和边界，不能替代真实评测。
- **待确认**：安装命令是否需要网络、权限或全局写入？。原因：这影响企业环境和个人环境的安装风险。

## 开工前工作上下文

### 加载顺序

- 先读取 how_to_use.host_ai_instruction，建立安装前判断资产的边界。
- 读取 claim_graph_summary，确认事实来自 Claim/Evidence Graph，而不是 Human Wiki 叙事。
- 再读取 intended_users、capabilities 和 quick_start_candidates，判断用户是否匹配。
- 需要执行具体任务时，优先查 role_skill_index，再查 evidence_index。
- 遇到真实安装、文件修改、网络访问、性能或兼容性问题时，转入 risk_card 和 boundaries.runtime_required。

### 任务路由

- **命令行启动或安装流程**：先说明这是安装后验证能力，再给出安装前检查清单。 边界：必须真实安装或运行后验证。 证据：`DEVELOPMENT.md`, `docs/design-docs/design_docs/20250825-geometry.md`, `docs/design-docs/design_docs/cdc/20260304-cdc-per-cluster-mtls-user-guide.md`, `scripts/download_milvus_proto.sh` 等 Claim：`clm_0001` supported 0.86

### 上下文规模

- 文件总数：4429
- 重要文件覆盖：40/4429
- 证据索引条目：80
- 角色 / Skill 条目：77

### 证据不足时的处理

- **missing_evidence**：说明证据不足，要求用户提供目标文件、README 段落或安装后验证记录；不要补全事实。
- **out_of_scope_request**：说明该任务超出当前 AI Context Pack 证据范围，并建议用户先查看 Human Manual 或真实安装后验证。
- **runtime_request**：给出安装前检查清单和命令来源，但不要替用户执行命令或声称已执行。
- **source_conflict**：同时展示冲突来源，标记为待核实，不要强行选择一个版本。

## Prompt Recipes

### 适配判断

- 目标：判断这个项目是否适合用户当前任务。
- 预期输出：适配结论、关键理由、证据引用、安装前可预览内容、必须安装后验证内容、下一步建议。

```text
请基于 milvus 的 AI Context Pack，先问我 3 个必要问题，然后判断它是否适合我的任务。回答必须包含：适合谁、能做什么、不能做什么、是否值得安装、证据来自哪里。所有项目事实必须引用 evidence_refs、source_paths 或 claim_id。
```

### 安装前体验

- 目标：让用户在安装前感受核心工作流，同时避免把预览包装成真实能力或营销承诺。
- 预期输出：一段带边界标签的体验剧本、安装后验证清单和谨慎建议；不含真实运行承诺或强营销表述。

```text
请把 milvus 当作安装前体验资产，而不是已安装工具或真实运行环境。

请严格输出四段：
1. 先问我 3 个必要问题。
2. 给出一段“体验剧本”：用 [安装前可预览]、[必须安装后验证]、[证据不足] 三种标签展示它可能如何引导工作流。
3. 给出安装后验证清单：列出哪些能力只有真实安装、真实宿主加载、真实项目运行后才能确认。
4. 给出谨慎建议：只能说“值得继续研究/试装”“先补充信息后再判断”或“不建议继续”，不得替项目背书。

硬性边界：
- 不要声称已经安装、运行、执行测试、修改文件或产生真实结果。
- 不要写“自动适配”“确保通过”“完美适配”“强烈建议安装”等承诺性表达。
- 如果描述安装后的工作方式，必须使用“如果安装成功且宿主正确加载 Skill，它可能会……”这种条件句。
- 体验剧本只能写成“示例台词/假设流程”：使用“可能会询问/可能会建议/可能会展示”，不要写“已写入、已生成、已通过、正在运行、正在生成”。
- Prompt Preview 不负责给安装命令；如用户准备试装，只能提示先阅读 Quick Start 和 Risk Card，并在隔离环境验证。
- 所有项目事实必须来自 supported claim、evidence_refs 或 source_paths；inferred/unverified 只能作风险或待确认项。

```

### 角色 / Skill 选择

- 目标：从项目里的角色或 Skill 中挑选最匹配的资产。
- 预期输出：候选角色或 Skill 列表，每项包含适用场景、证据路径、风险边界和是否需要安装后验证。

```text
请读取 role_skill_index，根据我的目标任务推荐 3-5 个最相关的角色或 Skill。每个推荐都要说明适用场景、可能输出、风险边界和 evidence_refs。
```

### 风险预检

- 目标：安装或引入前识别环境、权限、规则冲突和质量风险。
- 预期输出：环境、权限、依赖、许可、宿主冲突、质量风险和未知项的检查清单。

```text
请基于 risk_card、boundaries 和 quick_start_candidates，给我一份安装前风险预检清单。不要替我执行命令，只说明我应该检查什么、为什么检查、失败会有什么影响。
```

### 宿主 AI 开工指令

- 目标：把项目上下文转成一次对话开始前的宿主 AI 指令。
- 预期输出：一段边界明确、证据引用明确、适合复制给宿主 AI 的开工前指令。

```text
请基于 milvus 的 AI Context Pack，生成一段我可以粘贴给宿主 AI 的开工前指令。这段指令必须遵守 not_runtime=true，不能声称项目已经安装、运行或产生真实结果。
```


## 角色 / Skill 索引

- 共索引 77 个角色 / Skill / 项目文档条目。

- **What is Milvus?**（project_doc）：🐦 Milvus https://milvus.io/ is a high-performance vector database built for scale. It powers AI applications by efficiently organizing and searching vast amounts of unstructured data, such as text, images, and multi-modal information. 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`README.md`
- **Go MilvusClient**（project_doc）：! license https://img.shields.io/hexpm/l/plug.svg?color=green https://github.com/milvus-io/milvus/blob/master/LICENSE ! Go Reference https://pkg.go.dev/badge/github.com/milvus-io/milvus/client/v2.svg https://pkg.go.dev/github.com/milvus-io/milvus/client/v2 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`client/README.md`
- **Install git-hooks**（project_doc）：If you want to use git hooks, you need to install hooks first! 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`githooks/README.md`
- **Compile and install milvus cluster**（project_doc）：Generate the go files from proto file 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`scripts/README.md`
- **Milvus Development Tools**（project_doc）：mgit.py - Intelligent Git Workflow Tool 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`tools/README.md`
- **Run Milvus standalone through binary files**（project_doc）：Run Milvus standalone through binary files 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`deployments/binary/README.md`
- **README**（project_doc）：For better tracking and debugging Milvus, the script export-milvus-log.sh is provided for exporting all Milvus logs at once. For those pods that have been restarted, this script can export the logs of the running pods and the logs of the previously pods. 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`deployments/export-log/README.md`
- **README**（project_doc）：Milvus 2.2 has changed the meta structure for segment index. To upgrade a Milvus cluster of 2.1.x version you have installed, run this script to migrate the meta and upgrade the Milvus image version. 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`deployments/migrate-meta/README.md`
- **Milvus Metrics Dashboard**（project_doc）：Milvus outputs a list of detailed time-series metrics during runtime. You can use Prometheus https://prometheus.io/ and Grafana https://grafana.com/ to visualize the metrics. This topic introduces the monitoring metrics displayed in the Grafana Milvus Dashboard. 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`deployments/monitor/grafana/README.md`
- **Milvus offline installation**（project_doc）：Milvus installation may fail when images are not properly loaded from public Docker registries. To pull all images and save them into a directory that can be moved to the target host and loaded manually, perform the following procedures: 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`deployments/offline/README.md`
- **README**（project_doc）：Milvus 2.2.3 supports rolling update. This script helps you to perform a rolling update with zero downtime. 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`deployments/upgrade/README.md`
- **Basic usage**（project_doc）：! Build Status https://travis-ci.org/joboccara/NamedType.svg?branch=master https://travis-ci.org/joboccara/NamedType ! GitHub https://img.shields.io/github/license/joboccara/pipes 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`internal/core/thirdparty/NamedType/README.md`
- **Data Coordinator**（project_doc）：Data cooridnator datacoord for short is the component to organize DataNodes and segments allocations. 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`internal/datacoord/README.md`
- **Session Package**（project_doc）：session package contains the worker manager/nodes abstraction for datanodes and indexnodes. 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`internal/datacoord/session/README.md`
- **Data Node**（project_doc）：DataNode is the component to write insert and delete messages into persistent blob storage, for example MinIO or S3. 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`internal/datanode/README.md`
- **pkoracle package**（project_doc）：This package defines the interface and implementations for segments bloom filter sets of flushcommon metacache. 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`internal/flushcommon/metacache/pkoracle/README.md`
- **Generate Parser with Antlr4**（project_doc）：Please follow install antlr4 https://github.com/antlr/antlr4/blob/master/doc/go-target.md to install the antlr tool. 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`internal/parser/planparserv2/README.md`
- **Expression Rewriter planparserv2/rewriter**（project_doc）：Expression Rewriter planparserv2/rewriter 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`internal/parser/planparserv2/rewriter/README.md`
- **Summary**（project_doc）：this package contains privilege related components for proxy. 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`internal/proxy/privilege/README.md`
- **ShardClient Package**（project_doc）：The shardclient package provides client-side connection management and load balancing for communicating with QueryNode shards in the Milvus distributed architecture. It manages QueryNode client connections, caches shard leader information, and implements intelligent request routing strategies. 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`internal/proxy/shardclient/README.md`
- **WAL**（project_doc）：wal package is the basic defination of wal interface of milvus streamingnode. wal use github.com/milvus-io/milvus/pkg/streaming/walimpls to implement the final wal service. 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`internal/streamingnode/server/wal/README.md`
- **mlog - Context-Aware Logging Library**（project_doc）：mlog - Context-Aware Logging Library 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`pkg/mlog/README.md`
- **Milvus**（project_doc）：Vector database. Go + C++ internal/core/ + Rust tantivy . pkg has its own go.mod module: github.com/milvus-io/milvus/pkg/v3 . Run go get from pkg/ when adding dependencies there, not from root. 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`CLAUDE.md`
- **Contributing to Milvus**（project_doc）：Contributions to Milvus are welcome from everyone. We strive to make the contribution process simple and straightforward. Up-to-date information can be found at milvus.io https://milvus.io/ . 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`CONTRIBUTING.md`
- **Channel Model**（project_doc）：The WAL is partitioned into three channel types: PChannel physical , VChannel logical , and CChannel control . 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`docs/agent_guides/streaming-system/channel/channel.md`
- **Broadcaster**（project_doc）：Executes cross-PChannel atomic broadcast for DDL/DCL messages with resource locking, ACK tracking, and callback execution. Singleton running inside StreamingCoord. 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`docs/agent_guides/streaming-system/coordination/broadcaster.md`
- **Channel Management**（project_doc）：Singletons running inside StreamingCoord that manage PChannel/VChannel/CChannel ../channel/channel.md assignment and metadata. 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`docs/agent_guides/streaming-system/coordination/channel_management.md`
- **Alias Messages**（project_doc）：Messages managing collection aliases. All are CChannel-only broadcasts serialized at the database level. 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`docs/agent_guides/streaming-system/message/message-semantic-alias.md`
- **Cluster Messages**（project_doc）：Messages operating at cluster scope — global barriers and cluster-wide configuration. 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`docs/agent_guides/streaming-system/message/message-semantic-cluster.md`
- **Collection Messages**（project_doc）：Messages operating on collections, partitions, segments, indexes, snapshots, imports, and DML. 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`docs/agent_guides/streaming-system/message/message-semantic-collection.md`
- **Database Messages**（project_doc）：Messages managing database lifecycle. All are CChannel-only broadcasts serialized via ExclusiveDBName. 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`docs/agent_guides/streaming-system/message/message-semantic-database.md`
- **RBAC Messages**（project_doc）：Messages managing users, roles, privileges, and privilege groups. All are CChannel-only broadcasts serialized via ExclusivePrivilege resource key. 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`docs/agent_guides/streaming-system/message/message-semantic-rbac.md`
- **TimeTick Message**（project_doc）：A system-generated WAL message acting as a visibility barrier : when a consumer sees TimeTick with timestamp T, all messages with TimeTick < T are committed and safe to consume. No future message will have TimeTick < T. 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`docs/agent_guides/streaming-system/message/message-semantic-time-tick.md`
- **Transaction Messages**（project_doc）：Messages managing transactions on a single VChannel. 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`docs/agent_guides/streaming-system/message/message-semantic-txn.md`
- **Message Model**（project_doc）：Every WAL entry is a Message — the fundamental data unit flowing through all WAL components. A message consists of a typed payload protobuf-encoded header + body and key-value properties map string string , reserved keys prefixed with . 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`docs/agent_guides/streaming-system/message/message.md`
- **Replication & CDC**（project_doc）：Milvus supports multi-cluster WAL replication via a star topology: one PRIMARY cluster origin of all writes and one or more SECONDARY clusters replicas receiving WAL messages . Replication operates per-PChannel. 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`docs/agent_guides/streaming-system/replication/replicate.md`
- **StreamingClient**（project_doc）：In-process singleton library accessed via streaming.WAL providing the client-side API for WAL operations: Append, Read, Broadcast, and Txn. 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`docs/agent_guides/streaming-system/streaming-client/streaming-client.md`
- **Milvus Streaming System**（project_doc）：How to use this knowledge base : This README provides the architecture overview of the WAL system. Each component name is a link to its detailed doc. When your task involves a specific component, read the linked doc to get implementation details, interfaces, and code locations before making changes. 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`docs/agent_guides/streaming-system/streaming-system.md`
- **Lock**（project_doc）：The Lock interceptor enforces exclusive/shared access on VChannel or PChannel scope during append, preventing concurrent conflicting operations. 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`docs/agent_guides/streaming-system/wal/lock.md`
- **RecoveryStorage**（project_doc）：Persists WAL consumer state to the catalog etcd and object storage. Core invariant : from any WAL position + the corresponding persisted state, RecoveryStorage can replay the WAL forward and recover a fully consistent in-memory state. 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`docs/agent_guides/streaming-system/wal/recovery-storage.md`
- **Shard Management**（project_doc）：The Shard interceptor manages per-PChannel collection/partition/segment in-memory metadata and assigns each incoming DML message to a growing segment. All state is purely in-memory, should always keep consistent with underlying-WAL; on WAL open, it is recovered from the RecoveryStorage recovery-storage.md snapshot. 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`docs/agent_guides/streaming-system/wal/shard-management.md`
- **TimeTick & Transaction**（project_doc）：PChannel-level monotonically increasing log sequence number assigned to every WAL message. Defines total order within a PChannel and serves as the MVCC visibility boundary. Only comparable within the same PChannel. 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`docs/agent_guides/streaming-system/wal/timetick_and_txn.md`
- **WAL Backend**（project_doc）：The WAL backend is the durable storage layer for WAL entries. Each PChannel maps 1:1 to a backend topic/partition. The backend is pluggable — implementations share a common WALImpls interface for append, read, and truncate operations. 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`docs/agent_guides/streaming-system/walbackend/walbackend.md`
- **DataNode Recovery Design**（project_doc）：update: 5.21.2021, by Goose https://github.com/XuanYang-cn update: 6.03.2021, by Goose https://github.com/XuanYang-cn update: 6.21.2021, by Goose https://github.com/XuanYang-cn 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`docs/design-docs/design_docs/20210521-datanode_recovery_design.md`
- **DataNode Flowgraph Recovery Design**（project_doc）：update: 6.4.2021, by Goose https://github.com/XuanYang-cn update: 6.21.2021, by Goose https://github.com/XuanYang-cn 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`docs/design-docs/design_docs/20210604-datanode_flowgraph_recovery_design.md`
- **8. IndexCoord Design**（project_doc）：update: 7.31.2021, by Cai.Zhang https://github.com/xiaocai2333 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`docs/design-docs/design_docs/20210731-index_design.md`
- **Flush Collection**（project_doc）：Flush Collection The Flush operation is used to make sure that inserted data will be written into persistent storage. This document will introduce how the Flush operation works in Milvus 2.0 . The following figure shows the execution flow of Flush . 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`docs/design-docs/design_docs/20211109-milvus_flush_collections.md`
- **Drop Collection**（project_doc）：Milvus 2.0 uses Collection to represent a set of data, like Table in traditional database. Users can create or drop Collection . This article introduces the execution path of Drop Collection . At the end of this article, you should know which components are involved in Drop Collection . 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`docs/design-docs/design_docs/20211115-milvus_drop_collection.md`
- **Hybrid Timestamp in Milvus**（project_doc）：In chapter Milvus TimeSync Mechanism ./milvus timesync en.md , we have already known why we need TSO in Milvus. Milvus uses the TiKV's https://github.com/tikv/tikv implementation into TSO. So if you are interested in how TSO is implemented, you can look into the official documentation of TiKV. 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`docs/design-docs/design_docs/20211214-milvus_hybrid_ts.md`
- **Timesync -- All The things you should know**（project_doc）：Timesync -- All The things you should know 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`docs/design-docs/design_docs/20211215-milvus_timesync.md`
- **Create Collection**（project_doc）：Milvus 2.0 uses Collection to represent a set of data, like Table in a traditional database. User can create or drop Collection . This article introduces the execution path of CreateCollection , at the end of this article, you should know which components are involved in CreateCollection . 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`docs/design-docs/design_docs/20211217-milvus_create_collection.md`
- **Support to retrieve the specified entity from a collection**（project_doc）：Support to retrieve the specified entity from a collection 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`docs/design-docs/design_docs/20211221-retrieve_entity.md`
- **What's Knowhere**（project_doc）：Vector index is a time-efficient and space-efficient data structure built on vectors through a certain mathematical model. Through the vector index, we can efficiently query several vectors similar to the target vector. Since accurate retrieval is usually very time-consuming, most of the vector index types of Milvus use ANNS Approximate Nearest Neighbors Search . Compared with accurate retrieval, the core idea of AN… 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`docs/design-docs/design_docs/20211223-knowhere_design.md`
- **DropCollection release resources**（project_doc）：1. DataNode releases the flowgraph of this collection and drops all the data in a buffer. 2. DataCoord has no idea whether a collection is dropped or not. - DataCoord will make DataNode watch DmChannels of dropped collections. - Blob files will never be removed even if the collection is dropped. 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`docs/design-docs/design_docs/20211224-drop_collection_release_resources.md`
- **Create Index**（project_doc）：Index system is the core part of Milvus , which is used to speed up the searches, this document introduces which components are involved in Create Index ,and what these components do. 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`docs/design-docs/design_docs/20211227-milvus_create_index.md`
- **6. Proxy**（project_doc）：As the user access layer of Milvus, Proxy mainly plays a role that does some checks and preprocessing for requests from clients and then forwards these requests to other components, such as Root Coordinator, Data Coordinator, Query Coordinator, Index Coordinator. The below figure shows how Proxy interacts with other components. 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`docs/design-docs/design_docs/20220105-proxy.md`
- **20220105 Query Boolean Expr**（project_doc）：LogicalExpr := LogicalExpr BinaryLogicalOp LogicalExpr UnaryLogicalOp LogicalExpr " " LogicalExpr " " SingleExpr 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`docs/design-docs/design_docs/20220105-query_boolean_expr.md`
- **Root Coordinator recovery on power failure**（project_doc）：Root Coordinator recovery on power failure 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`docs/design-docs/design_docs/20220105-root_coordinator_recovery_on_power_failure.md`
- **MEP: Dynamic Configuration**（project_doc）：ISSUE: https://github.com/milvus-io/milvus/issues/18300 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`docs/design-docs/design_docs/20220725-dynamic-config.md`
- **MEP: Search By Primary Keys**（project_doc）：ISSUE: Feature : Support to search by primary keys 23184 https://github.com/milvus-io/milvus/issues/23184 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`docs/design-docs/design_docs/20230403-search_by_pk.md`
- **MEP: Default Value**（project_doc）：ISSUE: Feature : Support Default Value 23337 https://github.com/milvus-io/milvus/issues/23337 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`docs/design-docs/design_docs/20230405-default_value.md`
- **MEP: Refactor QueryNode v2**（project_doc）：ISSUE: Enhancement : Refactor QueryNode 21624 https://github.com/milvus-io/milvus/issues/21624 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`docs/design-docs/design_docs/20230418-querynode_v2.md`
- **MEP: Add collection level auto compaction config**（project_doc）：MEP: Add collection level auto compaction config 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`docs/design-docs/design_docs/20230511-collection_level_autocompaction_switch.md`
- **MEP: Datanode remove dependency of Datacoord**（project_doc）：MEP: Datanode remove dependency of Datacoord 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`docs/design-docs/design_docs/20230918-datanode_remove_datacoord_dependency.md`
- **JSON Storage Design Document**（project_doc）：Dense Part A set of "core fields" such as primary keys and commonly used metadata that are present in most records. 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`docs/design-docs/design_docs/20250308-json_storage.md`
- **Primary Key Index Design Document**（project_doc）：This document outlines the design of Milvus' primary key indexing system, which enables fast lookups of string or integer primary keys across multiple segments. The index will be loaded in the Delegator and persisted in S3 storage. 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`docs/design-docs/design_docs/20250429-primarykey_index.md`
- **Milvus Row Level Security RLS Design Document**（project_doc）：Milvus Row Level Security RLS Design Document 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`docs/design-docs/design_docs/20250610-rls_design.md`
- **Background**（project_doc）：Milvus is a high-performance vector database widely used in fields such as image retrieval, recommender systems, and semantic search. With the increasing integration of scenarios such as LBS location-based service , MultiModal Machine Learning retrieval, and driverless technology spatial awareness and semantic retrieval requirements, users increasingly need to be able to perform vector searches in "geographical cont… 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`docs/design-docs/design_docs/20250825-geometry.md`
- **Milvus Snapshot Design Document**（project_doc）：Milvus Snapshot mechanism provides complete collection-level data snapshot capabilities, implementing point-in-time backup and restore functionality. The implementation includes the following core components: 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`docs/design-docs/design_docs/20251114-snapshot_design.md`
- **Entity-level TTL Design**（project_doc）：Currently, Milvus supports collection-level TTL for data expiration, but does not support defining an independent expiration time for individual entities rows . As application scenarios become more diverse, for example: 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`docs/design-docs/design_docs/20251221-entity_level_ttl.md`
- **External Table External Collection Design Document**（project_doc）：External Table External Collection Design Document 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`docs/design-docs/design_docs/20260105-external_table.md`
- **Design Document: Add Function Field Feature**（project_doc）：Design Document: Add Function Field Feature 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`docs/design-docs/design_docs/20260129-add-function-field-design.md`
- **Milvus Search Order By Feature Design Document**（project_doc）：Milvus Search Order By Feature Design Document 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`docs/design-docs/design_docs/20260129-search-orderby.md`
- **Truncate Collection Design Document**（project_doc）：Truncate Collection Design Document 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`docs/design-docs/design_docs/20260129-truncate_collection.md`
- **Milvus Embedded Group By - Design Document**（project_doc）：Milvus Embedded Group By - Design Document 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`docs/design-docs/design_docs/20260130-embeded-group-by.md`
- **MEP: Client-Side Telemetry with Heartbeat and Server Command Support**（project_doc）：MEP: Client-Side Telemetry with Heartbeat and Server Command Support 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`docs/design-docs/design_docs/20260131-client_side_telemetry.md`
- **Query ORDER BY**（project_doc）：- Created: 2026-02-03 - Author s : @MrPresent-Han - Status: Under Review - Component: QueryNode / Proxy - Related Issues: N/A - Released: N/A 激活提示：当用户需要理解项目结构、安装方式或边界时参考。 证据：`docs/design-docs/design_docs/20260203-query-orderby.md`

## 证据索引

- 共索引 80 条证据。

- **What is Milvus?**（documentation）：🐦 Milvus https://milvus.io/ is a high-performance vector database built for scale. It powers AI applications by efficiently organizing and searching vast amounts of unstructured data, such as text, images, and multi-modal information. 证据：`README.md`
- **Go MilvusClient**（documentation）：! license https://img.shields.io/hexpm/l/plug.svg?color=green https://github.com/milvus-io/milvus/blob/master/LICENSE ! Go Reference https://pkg.go.dev/badge/github.com/milvus-io/milvus/client/v2.svg https://pkg.go.dev/github.com/milvus-io/milvus/client/v2 证据：`client/README.md`
- **Install git-hooks**（documentation）：If you want to use git hooks, you need to install hooks first! 证据：`githooks/README.md`
- **Compile and install milvus cluster**（documentation）：Generate the go files from proto file 证据：`scripts/README.md`
- **Milvus Development Tools**（documentation）：mgit.py - Intelligent Git Workflow Tool 证据：`tools/README.md`
- **Run Milvus standalone through binary files**（documentation）：Run Milvus standalone through binary files 证据：`deployments/binary/README.md`
- **README**（documentation）：For better tracking and debugging Milvus, the script export-milvus-log.sh is provided for exporting all Milvus logs at once. For those pods that have been restarted, this script can export the logs of the running pods and the logs of the previously pods. 证据：`deployments/export-log/README.md`
- **README**（documentation）：Milvus 2.2 has changed the meta structure for segment index. To upgrade a Milvus cluster of 2.1.x version you have installed, run this script to migrate the meta and upgrade the Milvus image version. 证据：`deployments/migrate-meta/README.md`
- **Milvus Metrics Dashboard**（documentation）：Milvus outputs a list of detailed time-series metrics during runtime. You can use Prometheus https://prometheus.io/ and Grafana https://grafana.com/ to visualize the metrics. This topic introduces the monitoring metrics displayed in the Grafana Milvus Dashboard. 证据：`deployments/monitor/grafana/README.md`
- **Milvus offline installation**（documentation）：Milvus installation may fail when images are not properly loaded from public Docker registries. To pull all images and save them into a directory that can be moved to the target host and loaded manually, perform the following procedures: 证据：`deployments/offline/README.md`
- **README**（documentation）：Milvus 2.2.3 supports rolling update. This script helps you to perform a rolling update with zero downtime. 证据：`deployments/upgrade/README.md`
- **Basic usage**（documentation）：! Build Status https://travis-ci.org/joboccara/NamedType.svg?branch=master https://travis-ci.org/joboccara/NamedType ! GitHub https://img.shields.io/github/license/joboccara/pipes 证据：`internal/core/thirdparty/NamedType/README.md`
- **Data Coordinator**（documentation）：Data cooridnator datacoord for short is the component to organize DataNodes and segments allocations. 证据：`internal/datacoord/README.md`
- **Session Package**（documentation）：session package contains the worker manager/nodes abstraction for datanodes and indexnodes. 证据：`internal/datacoord/session/README.md`
- **Data Node**（documentation）：DataNode is the component to write insert and delete messages into persistent blob storage, for example MinIO or S3. 证据：`internal/datanode/README.md`
- **pkoracle package**（documentation）：This package defines the interface and implementations for segments bloom filter sets of flushcommon metacache. 证据：`internal/flushcommon/metacache/pkoracle/README.md`
- **Generate Parser with Antlr4**（documentation）：Please follow install antlr4 https://github.com/antlr/antlr4/blob/master/doc/go-target.md to install the antlr tool. 证据：`internal/parser/planparserv2/README.md`
- **Expression Rewriter planparserv2/rewriter**（documentation）：Expression Rewriter planparserv2/rewriter 证据：`internal/parser/planparserv2/rewriter/README.md`
- **Summary**（documentation）：this package contains privilege related components for proxy. 证据：`internal/proxy/privilege/README.md`
- **ShardClient Package**（documentation）：The shardclient package provides client-side connection management and load balancing for communicating with QueryNode shards in the Milvus distributed architecture. It manages QueryNode client connections, caches shard leader information, and implements intelligent request routing strategies. 证据：`internal/proxy/shardclient/README.md`
- **WAL**（documentation）：wal package is the basic defination of wal interface of milvus streamingnode. wal use github.com/milvus-io/milvus/pkg/streaming/walimpls to implement the final wal service. 证据：`internal/streamingnode/server/wal/README.md`
- **mlog - Context-Aware Logging Library**（documentation）：mlog - Context-Aware Logging Library 证据：`pkg/mlog/README.md`
- **Milvus**（documentation）：Vector database. Go + C++ internal/core/ + Rust tantivy . pkg has its own go.mod module: github.com/milvus-io/milvus/pkg/v3 . Run go get from pkg/ when adding dependencies there, not from root. 证据：`CLAUDE.md`
- **Contributing to Milvus**（documentation）：Contributions to Milvus are welcome from everyone. We strive to make the contribution process simple and straightforward. Up-to-date information can be found at milvus.io https://milvus.io/ . 证据：`CONTRIBUTING.md`
- **License**（source_file）：Apache License Version 2.0, January 2004 http://www.apache.org/licenses/ 证据：`LICENSE`
- **License**（source_file）：Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files the "Software" , to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions: 证据：`internal/core/thirdparty/NamedType/LICENSE`
- **License**（source_file）：Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files the "Software" , to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions: 证据：`internal/core/thirdparty/boost_ext/LICENSE`
- **Channel Model**（documentation）：The WAL is partitioned into three channel types: PChannel physical , VChannel logical , and CChannel control . 证据：`docs/agent_guides/streaming-system/channel/channel.md`
- **Broadcaster**（documentation）：Executes cross-PChannel atomic broadcast for DDL/DCL messages with resource locking, ACK tracking, and callback execution. Singleton running inside StreamingCoord. 证据：`docs/agent_guides/streaming-system/coordination/broadcaster.md`
- **Channel Management**（documentation）：Singletons running inside StreamingCoord that manage PChannel/VChannel/CChannel ../channel/channel.md assignment and metadata. 证据：`docs/agent_guides/streaming-system/coordination/channel_management.md`
- **Alias Messages**（documentation）：Messages managing collection aliases. All are CChannel-only broadcasts serialized at the database level. 证据：`docs/agent_guides/streaming-system/message/message-semantic-alias.md`
- **Cluster Messages**（documentation）：Messages operating at cluster scope — global barriers and cluster-wide configuration. 证据：`docs/agent_guides/streaming-system/message/message-semantic-cluster.md`
- **Collection Messages**（documentation）：Messages operating on collections, partitions, segments, indexes, snapshots, imports, and DML. 证据：`docs/agent_guides/streaming-system/message/message-semantic-collection.md`
- **Database Messages**（documentation）：Messages managing database lifecycle. All are CChannel-only broadcasts serialized via ExclusiveDBName. 证据：`docs/agent_guides/streaming-system/message/message-semantic-database.md`
- **RBAC Messages**（documentation）：Messages managing users, roles, privileges, and privilege groups. All are CChannel-only broadcasts serialized via ExclusivePrivilege resource key. 证据：`docs/agent_guides/streaming-system/message/message-semantic-rbac.md`
- **TimeTick Message**（documentation）：A system-generated WAL message acting as a visibility barrier : when a consumer sees TimeTick with timestamp T, all messages with TimeTick < T are committed and safe to consume. No future message will have TimeTick < T. 证据：`docs/agent_guides/streaming-system/message/message-semantic-time-tick.md`
- **Transaction Messages**（documentation）：Messages managing transactions on a single VChannel. 证据：`docs/agent_guides/streaming-system/message/message-semantic-txn.md`
- **Message Model**（documentation）：Every WAL entry is a Message — the fundamental data unit flowing through all WAL components. A message consists of a typed payload protobuf-encoded header + body and key-value properties map string string , reserved keys prefixed with . 证据：`docs/agent_guides/streaming-system/message/message.md`
- **Replication & CDC**（documentation）：Milvus supports multi-cluster WAL replication via a star topology: one PRIMARY cluster origin of all writes and one or more SECONDARY clusters replicas receiving WAL messages . Replication operates per-PChannel. 证据：`docs/agent_guides/streaming-system/replication/replicate.md`
- **StreamingClient**（documentation）：In-process singleton library accessed via streaming.WAL providing the client-side API for WAL operations: Append, Read, Broadcast, and Txn. 证据：`docs/agent_guides/streaming-system/streaming-client/streaming-client.md`
- **Milvus Streaming System**（documentation）：How to use this knowledge base : This README provides the architecture overview of the WAL system. Each component name is a link to its detailed doc. When your task involves a specific component, read the linked doc to get implementation details, interfaces, and code locations before making changes. 证据：`docs/agent_guides/streaming-system/streaming-system.md`
- **Lock**（documentation）：The Lock interceptor enforces exclusive/shared access on VChannel or PChannel scope during append, preventing concurrent conflicting operations. 证据：`docs/agent_guides/streaming-system/wal/lock.md`
- **RecoveryStorage**（documentation）：Persists WAL consumer state to the catalog etcd and object storage. Core invariant : from any WAL position + the corresponding persisted state, RecoveryStorage can replay the WAL forward and recover a fully consistent in-memory state. 证据：`docs/agent_guides/streaming-system/wal/recovery-storage.md`
- **Shard Management**（documentation）：The Shard interceptor manages per-PChannel collection/partition/segment in-memory metadata and assigns each incoming DML message to a growing segment. All state is purely in-memory, should always keep consistent with underlying-WAL; on WAL open, it is recovered from the RecoveryStorage recovery-storage.md snapshot. 证据：`docs/agent_guides/streaming-system/wal/shard-management.md`
- **TimeTick & Transaction**（documentation）：PChannel-level monotonically increasing log sequence number assigned to every WAL message. Defines total order within a PChannel and serves as the MVCC visibility boundary. Only comparable within the same PChannel. 证据：`docs/agent_guides/streaming-system/wal/timetick_and_txn.md`
- **WAL Backend**（documentation）：The WAL backend is the durable storage layer for WAL entries. Each PChannel maps 1:1 to a backend topic/partition. The backend is pluggable — implementations share a common WALImpls interface for append, read, and truncate operations. 证据：`docs/agent_guides/streaming-system/walbackend/walbackend.md`
- **DataNode Recovery Design**（documentation）：update: 5.21.2021, by Goose https://github.com/XuanYang-cn update: 6.03.2021, by Goose https://github.com/XuanYang-cn update: 6.21.2021, by Goose https://github.com/XuanYang-cn 证据：`docs/design-docs/design_docs/20210521-datanode_recovery_design.md`
- **DataNode Flowgraph Recovery Design**（documentation）：update: 6.4.2021, by Goose https://github.com/XuanYang-cn update: 6.21.2021, by Goose https://github.com/XuanYang-cn 证据：`docs/design-docs/design_docs/20210604-datanode_flowgraph_recovery_design.md`
- **8. IndexCoord Design**（documentation）：update: 7.31.2021, by Cai.Zhang https://github.com/xiaocai2333 证据：`docs/design-docs/design_docs/20210731-index_design.md`
- **Flush Collection**（documentation）：Flush Collection The Flush operation is used to make sure that inserted data will be written into persistent storage. This document will introduce how the Flush operation works in Milvus 2.0 . The following figure shows the execution flow of Flush . 证据：`docs/design-docs/design_docs/20211109-milvus_flush_collections.md`
- **Drop Collection**（documentation）：Milvus 2.0 uses Collection to represent a set of data, like Table in traditional database. Users can create or drop Collection . This article introduces the execution path of Drop Collection . At the end of this article, you should know which components are involved in Drop Collection . 证据：`docs/design-docs/design_docs/20211115-milvus_drop_collection.md`
- **Hybrid Timestamp in Milvus**（documentation）：In chapter Milvus TimeSync Mechanism ./milvus timesync en.md , we have already known why we need TSO in Milvus. Milvus uses the TiKV's https://github.com/tikv/tikv implementation into TSO. So if you are interested in how TSO is implemented, you can look into the official documentation of TiKV. 证据：`docs/design-docs/design_docs/20211214-milvus_hybrid_ts.md`
- **Timesync -- All The things you should know**（documentation）：Timesync -- All The things you should know 证据：`docs/design-docs/design_docs/20211215-milvus_timesync.md`
- **Create Collection**（documentation）：Milvus 2.0 uses Collection to represent a set of data, like Table in a traditional database. User can create or drop Collection . This article introduces the execution path of CreateCollection , at the end of this article, you should know which components are involved in CreateCollection . 证据：`docs/design-docs/design_docs/20211217-milvus_create_collection.md`
- **Support to retrieve the specified entity from a collection**（documentation）：Support to retrieve the specified entity from a collection 证据：`docs/design-docs/design_docs/20211221-retrieve_entity.md`
- **What's Knowhere**（documentation）：Vector index is a time-efficient and space-efficient data structure built on vectors through a certain mathematical model. Through the vector index, we can efficiently query several vectors similar to the target vector. Since accurate retrieval is usually very time-consuming, most of the vector index types of Milvus use ANNS Approximate Nearest Neighbors Search . Compared with accurate retrieval, the core idea of ANNS is no longer limited to returning the most accurate result, but only searching for neighbors of the target. ANNS improves retrieval efficiency by sacrificing accuracy within an acceptable range. 证据：`docs/design-docs/design_docs/20211223-knowhere_design.md`
- **DropCollection release resources**（documentation）：1. DataNode releases the flowgraph of this collection and drops all the data in a buffer. 2. DataCoord has no idea whether a collection is dropped or not. - DataCoord will make DataNode watch DmChannels of dropped collections. - Blob files will never be removed even if the collection is dropped. 证据：`docs/design-docs/design_docs/20211224-drop_collection_release_resources.md`
- **Create Index**（documentation）：Index system is the core part of Milvus , which is used to speed up the searches, this document introduces which components are involved in Create Index ,and what these components do. 证据：`docs/design-docs/design_docs/20211227-milvus_create_index.md`
- **6. Proxy**（documentation）：As the user access layer of Milvus, Proxy mainly plays a role that does some checks and preprocessing for requests from clients and then forwards these requests to other components, such as Root Coordinator, Data Coordinator, Query Coordinator, Index Coordinator. The below figure shows how Proxy interacts with other components. 证据：`docs/design-docs/design_docs/20220105-proxy.md`
- **20220105 Query Boolean Expr**（documentation）：LogicalExpr := LogicalExpr BinaryLogicalOp LogicalExpr UnaryLogicalOp LogicalExpr " " LogicalExpr " " SingleExpr 证据：`docs/design-docs/design_docs/20220105-query_boolean_expr.md`
- 其余 20 条证据见 `AI_CONTEXT_PACK.json` 或 `EVIDENCE_INDEX.json`。

## 宿主 AI 必须遵守的规则

- **把本资产当作开工前上下文，而不是运行环境。**：AI Context Pack 只包含证据化项目理解，不包含目标项目的可执行状态。 证据：`README.md`, `client/README.md`, `githooks/README.md`
- **回答用户时区分可预览内容与必须安装后才能验证的内容。**：安装前体验的消费者价值来自降低误装和误判，而不是伪装成真实运行。 证据：`README.md`, `client/README.md`, `githooks/README.md`

## 用户开工前应该回答的问题

- 你准备在哪个宿主 AI 或本地环境中使用它？
- 你只是想先体验工作流，还是准备真实安装？
- 你最在意的是安装成本、输出质量、还是和现有规则的冲突？

## 验收标准

- 所有能力声明都能回指到 evidence_refs 中的文件路径。
- AI_CONTEXT_PACK.md 没有把预览包装成真实运行。
- 用户能在 3 分钟内看懂适合谁、能做什么、如何开始和风险边界。

---

## Doramagic Context Augmentation

下面内容用于强化 Repomix/AI Context Pack 主体。Human Manual 只提供阅读骨架；踩坑日志会被转成宿主 AI 必须遵守的工作约束。

## Human Manual 骨架

使用规则：这里只是项目阅读路线和显著性信号，不是事实权威。具体事实仍必须回到 repo evidence / Claim Graph。

宿主 AI 硬性规则：
- 不得把页标题、章节顺序、摘要或 importance 当作项目事实证据。
- 解释 Human Manual 骨架时，必须明确说它只是阅读路线/显著性信号。
- 能力、安装、兼容性、运行状态和风险判断必须引用 repo evidence、source path 或 Claim Graph。

- **Milvus 概述**：importance `high`
  - source_paths: README.md, docs/developer_guides/chap01_system_overview.md
- **系统架构**：importance `high`
  - source_paths: internal/proxy/proxy.go, internal/rootcoord/root_coord.go, internal/datacoord/server.go, internal/querycoordv2/server.go, docs/developer_guides/chap01_system_overview.md
- **集合与 Schema 设计**：importance `high`
  - source_paths: internal/rootcoord/create_collection_task.go, internal/metastore/model/collection.go, internal/metastore/model/field.go, internal/metastore/model/partition.go
- **数据插入与查询流程**：importance `high`
  - source_paths: internal/proxy/task_insert.go, internal/proxy/task_search.go, internal/proxy/task_query.go, internal/querynodev2/handlers.go, docs/developer_guides/how-guarantee-ts-works.md
- **向量索引类型**：importance `high`
  - source_paths: client/index/hnsw.go, client/index/ivf.go, client/index/scann.go, client/index/disk_ann.go, client/index/sparse.go
- **协调节点详解**：importance `medium`
  - source_paths: internal/rootcoord/root_coord.go, internal/datacoord/server.go, internal/querycoordv2/server.go, internal/datacoord/services.go, internal/querycoordv2/services.go
- **执行节点详解**：importance `medium`
  - source_paths: internal/datanode/data_node.go, internal/querynodev2/server.go, internal/querynodev2/segments/segment.go, internal/querynodev2/segments/manager.go, internal/datanode/services.go
- **Go SDK (client/v2) 使用指南**：importance `high`
  - source_paths: client/milvusclient/client.go, client/milvusclient/collection.go, client/milvusclient/write.go, client/milvusclient/read.go, client/milvusclient/rbac.go

## Repo Inspection Evidence / 源码检查证据

- repo_clone_verified: true
- repo_inspection_verified: true
- repo_commit: `d8de464ff6ac9cca79e2688597c09f07c763bed5`
- inspected_files: `README.md`, `docker-compose.yml`, `docs/jaeger_guides/opentracing_user_guide.md`, `docs/developer_guides/chap01_system_overview.md`, `docs/developer_guides/appendix_c_system_configurations.md`, `docs/developer_guides/chap04_message_stream.md`, `docs/developer_guides/appendix_d_error_code.md`, `docs/developer_guides/chap07_query_coordinator.md`, `docs/developer_guides/chap09_data_coord.md`, `docs/developer_guides/how_to_develop_with_local_milvus_proto.md`, `docs/developer_guides/chap08_binlog.md`, `docs/developer_guides/developer_guides.md`, `docs/developer_guides/chap02_schema.md`, `docs/developer_guides/how-guarantee-ts-works-cn.md`, `docs/developer_guides/proxy-reduce.md`, `docs/developer_guides/chap06_root_coordinator.md`, `docs/developer_guides/appendix_a_basic_components.md`, `docs/developer_guides/appendix_b_api_reference.md`, `docs/developer_guides/proxy-reduce-cn.md`, `docs/developer_guides/chap03_index_service.md`

宿主 AI 硬性规则：
- 没有 repo_clone_verified=true 时，不得声称已经读过源码。
- 没有 repo_inspection_verified=true 时，不得把 README/docs/package 文件判断写成事实。
- 没有 quick_start_verified=true 时，不得声称 Quick Start 已跑通。

## Doramagic Pitfall Constraints / 踩坑约束

这些规则来自 Doramagic 发现、验证或编译过程中的项目专属坑点。宿主 AI 必须把它们当作工作约束，而不是普通说明文字。

### Constraint 1: 来源证据：Milvus standalone crashed

- Trigger: GitHub 社区证据显示该项目存在一个安装相关的待验证问题：Milvus standalone crashed
- Host AI rule: 来源显示可能已有修复、规避或版本变化，说明书中必须标注适用版本。
- Why it matters: 可能阻塞安装或首次运行。
- Evidence: community_evidence:github | cevd_f0bbd70e35ef4c65a374c802b7614b46 | https://github.com/milvus-io/milvus/issues/28583 | 来源讨论提到 linux 相关条件，需在安装/试用前复核。
- Hard boundary: 不要把这个坑点包装成已解决、已验证或可忽略，除非后续验证证据明确证明它已经关闭。

### Constraint 2: 能力判断依赖假设

- Trigger: README/documentation is current enough for a first validation pass.
- Host AI rule: 将假设转成下游验证清单。
- Why it matters: 假设不成立时，用户拿不到承诺的能力。
- Evidence: capability.assumptions | github_repo:208728772 | https://github.com/milvus-io/milvus | README/documentation is current enough for a first validation pass.
- Hard boundary: 不要把这个坑点包装成已解决、已验证或可忽略，除非后续验证证据明确证明它已经关闭。

### Constraint 3: 运行可能依赖外部服务

- Trigger: 项目说明出现 external service/cloud/webhook/database 等运行依赖关键词。
- Host AI rule: 确认是否有离线 demo、mock 数据或可替代服务。
- Why it matters: 本地安装成功不等于能力可用，外部服务不可用会阻断体验。
- Evidence: packet_text.keyword_scan | github_repo:208728772 | https://github.com/milvus-io/milvus | matched external service / cloud / webhook / database keyword
- Hard boundary: 不要把这个坑点包装成已解决、已验证或可忽略，除非后续验证证据明确证明它已经关闭。

### Constraint 4: 维护活跃度未知

- Trigger: 未记录 last_activity_observed。
- Host AI rule: 补 GitHub 最近 commit、release、issue/PR 响应信号。
- Why it matters: 新项目、停更项目和活跃项目会被混在一起，推荐信任度下降。
- Evidence: evidence.maintainer_signals | github_repo:208728772 | https://github.com/milvus-io/milvus | last_activity_observed missing
- Hard boundary: 不要把这个坑点包装成已解决、已验证或可忽略，除非后续验证证据明确证明它已经关闭。

### Constraint 5: 下游验证发现风险项

- Trigger: no_demo
- Host AI rule: 进入安全/权限治理复核队列。
- Why it matters: 下游已经要求复核，不能在页面中弱化。
- Evidence: downstream_validation.risk_items | github_repo:208728772 | https://github.com/milvus-io/milvus | no_demo; severity=medium
- Hard boundary: 不要把这个坑点包装成已解决、已验证或可忽略，除非后续验证证据明确证明它已经关闭。

### Constraint 6: 存在评分风险

- Trigger: no_demo
- Host AI rule: 把风险写入边界卡，并确认是否需要人工复核。
- Why it matters: 风险会影响是否适合普通用户安装。
- Evidence: risks.scoring_risks | github_repo:208728772 | https://github.com/milvus-io/milvus | no_demo; severity=medium
- Hard boundary: 不要把这个坑点包装成已解决、已验证或可忽略，除非后续验证证据明确证明它已经关闭。

### Constraint 7: issue/PR 响应质量未知

- Trigger: issue_or_pr_quality=unknown。
- Host AI rule: 抽样最近 issue/PR，判断是否长期无人处理。
- Why it matters: 用户无法判断遇到问题后是否有人维护。
- Evidence: evidence.maintainer_signals | github_repo:208728772 | https://github.com/milvus-io/milvus | issue_or_pr_quality=unknown
- Hard boundary: 不要把这个坑点包装成已解决、已验证或可忽略，除非后续验证证据明确证明它已经关闭。

### Constraint 8: 发布节奏不明确

- Trigger: release_recency=unknown。
- Host AI rule: 确认最近 release/tag 和 README 安装命令是否一致。
- Why it matters: 安装命令和文档可能落后于代码，用户踩坑概率升高。
- Evidence: evidence.maintainer_signals | github_repo:208728772 | https://github.com/milvus-io/milvus | release_recency=unknown
- Hard boundary: 不要把这个坑点包装成已解决、已验证或可忽略，除非后续验证证据明确证明它已经关闭。