# feast - 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 feast 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 - **希望把专业流程带进宿主 AI 的用户**：仓库包含 Skill 文档。 证据：`skills/feast-architecture/SKILL.md`, `skills/feast-dev/SKILL.md`, `skills/feast-testing/SKILL.md`, `skills/feast-user-guide/SKILL.md` 等 Claim：`clm_0003` supported 0.86 ## 它能做什么 - **AI Skill / Agent 指令资产库**（可做安装前预览）：项目包含可被宿主 AI 读取的 Skill 或 Agent 指令文件，可用于把专业流程带入 Claude、Codex、Cursor 等宿主。 证据：`skills/feast-architecture/SKILL.md`, `skills/feast-dev/SKILL.md`, `skills/feast-testing/SKILL.md`, `skills/feast-user-guide/SKILL.md` 等 Claim：`clm_0001` supported 0.86 - **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：`README.md` Claim：`clm_0002` supported 0.86 ## 怎么开始 - `pip install feast` 证据：`README.md` Claim：`clm_0004` supported 0.86 ## 继续前判断卡 - **当前建议**：先做权限沙盒试用 - **为什么**：项目存在安装命令、宿主配置或本地写入线索，不建议直接进入主力环境，应先在隔离环境试装。 ### 30 秒判断 - **现在怎么做**：先做权限沙盒试用 - **最小安全下一步**：先跑 Prompt Preview；若仍要安装，只在隔离环境试装 - **先别相信**：工具权限边界不能在安装前相信。 - **继续会触碰**：命令执行、宿主 AI 配置、本地环境或项目文件 ### 现在可以相信 - **适合人群线索：希望把专业流程带进宿主 AI 的用户**（supported）：有 supported claim 或项目证据支撑，但仍不等于真实安装效果。 证据：`skills/feast-architecture/SKILL.md`, `skills/feast-dev/SKILL.md`, `skills/feast-testing/SKILL.md`, `skills/feast-user-guide/SKILL.md` 等 Claim：`clm_0003` supported 0.86 - **能力存在：AI Skill / Agent 指令资产库**（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：`skills/feast-architecture/SKILL.md`, `skills/feast-dev/SKILL.md`, `skills/feast-testing/SKILL.md`, `skills/feast-user-guide/SKILL.md` 等 Claim：`clm_0001` 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：`README.md` Claim：`clm_0002` 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：`README.md` Claim：`clm_0004` supported 0.86 ### 现在还不能相信 - **工具权限边界不能在安装前相信。**（unverified）：MCP/tool 类项目通常会触碰文件、网络、浏览器或外部 API，必须真实检查权限和日志。 - **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：`AGENTS.md`, `CLAUDE.md`, `skills/SKILL.md`, `skills/feast-architecture/SKILL.md` 等 - **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。 - **安装命令是否需要网络、权限或全局写入？**（unverified）：这影响企业环境和个人环境的安装风险。 证据：`README.md` ### 继续会触碰什么 - **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：`README.md` - **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：`AGENTS.md`, `CLAUDE.md`, `skills/SKILL.md`, `skills/feast-architecture/SKILL.md` 等 - **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：`README.md` - **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**：用安装前交互式试用判断工作方式是否匹配，不需要授权或改环境。（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。 - **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_0005` 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：`README.md` Claim：`clm_0006` 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 - **AI Skill / Agent 指令资产库**：先基于 role_skill_index / evidence_index 帮用户挑选可用角色、Skill 或工作流。 边界：可做安装前 Prompt 体验。 证据：`skills/feast-architecture/SKILL.md`, `skills/feast-dev/SKILL.md`, `skills/feast-testing/SKILL.md`, `skills/feast-user-guide/SKILL.md` 等 Claim：`clm_0001` supported 0.86 - **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：`README.md` Claim：`clm_0002` supported 0.86 ### 上下文规模 - 文件总数：2434 - 重要文件覆盖：40/2434 - 证据索引条目：83 - 角色 / Skill 条目：5 ### 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 feast 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 feast 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 feast 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 5 个角色 / Skill / 项目文档条目。 - **feast-user-guide**（skill）：Guide for working with Feast Feature Store — defining features, configuring feature store.yaml, retrieving features online/offline, using the CLI, and building RAG retrieval pipelines. Use when the user asks about creating entities, feature views, on-demand feature views, stream feature views, feature services, data sources, feature store.yaml configuration, feast apply/materialize commands, online or historical fea… 激活提示：当用户任务与“feast-user-guide”描述的流程高度相关时，先用它做安装前体验，再决定是否安装。 证据：`skills/SKILL.md` - **feast-architecture**（skill）：Internals of the Feast codebase — how each component works, where the key abstractions live, and the data flow through the system. Use when asked how feast apply works, how the registry stores data, how materialization moves data, how get online features retrieves features, how the feature server works, how the Kubernetes operator manages deployments, or when navigating the codebase to understand where to make a cha… 激活提示：当用户任务与“feast-architecture”描述的流程高度相关时，先用它做安装前体验，再决定是否安装。 证据：`skills/feast-architecture/SKILL.md` - **feast-dev**（skill）：Development guide for contributing to the Feast codebase. Covers environment setup, testing, linting, project structure, and PR workflow for feast-dev/feast. 激活提示：当用户任务与“feast-dev”描述的流程高度相关时，先用它做安装前体验，再决定是否安装。 证据：`skills/feast-dev/SKILL.md` - **feast-testing**（skill）：How to test and debug Feast — running targeted tests, writing unit tests for new components, debugging registry and online store issues, and inspecting live feature store state. Use when writing tests for a new feature, debugging a failing test, investigating a runtime error, or verifying that a change works correctly end-to-end. 激活提示：当用户任务与“feast-testing”描述的流程高度相关时，先用它做安装前体验，再决定是否安装。 证据：`skills/feast-testing/SKILL.md` - **feast-user-guide**（skill）：Guide for working with Feast Feature Store — defining features, configuring feature store.yaml, retrieving features online/offline, using the CLI, and building RAG retrieval pipelines. Use when the user asks about creating entities, feature views, on-demand feature views, stream feature views, feature services, data sources, feature store.yaml configuration, feast apply/materialize commands, online or historical fea… 激活提示：当用户任务与“feast-user-guide”描述的流程高度相关时，先用它做安装前体验，再决定是否安装。 证据：`skills/feast-user-guide/SKILL.md` ## 证据索引 - 共索引 83 条证据。 - **Introduction**（documentation）：Feast Fea ture St ore is an open-source https://github.com/feast-dev/feast feature store that helps teams operate production ML systems at scale by allowing them to define, manage, validate, and serve features for production AI/ML. 证据：`docs/README.md` - **Architecture Decision Records ADR**（documentation）：This directory contains Architecture Decision Records ADRs for the Feast project. ADRs document significant architectural decisions made during the development of Feast, providing context, rationale, and consequences for each decision. 证据：`docs/adr/README.md` - **Blog Posts**（documentation）：Welcome to the Feast blog! Here you'll find articles about feature store development, new features, and community updates. 证据：`docs/blog/README.md` - **Architecture**（documentation）：{% content-ref url="overview.md" %} overview.md overview.md {% endcontent-ref %} 证据：`docs/getting-started/architecture/README.md` - **Components**（documentation）：{% content-ref url="registry.md" %} registry.md registry.md {% endcontent-ref %} 证据：`docs/getting-started/components/README.md` - **Concepts**（documentation）：{% content-ref url="overview.md" %} overview.md overview.md {% endcontent-ref %} 证据：`docs/getting-started/concepts/README.md` - **Customizing Feast**（documentation）：Feast is highly pluggable and configurable: 证据：`docs/how-to-guides/customizing-feast/README.md` - **Feast Operator Configuration Guides**（documentation）：Feast Operator Configuration Guides 证据：`docs/how-to-guides/feast-operator/README.md` - **Running Feast with Snowflake/GCP/AWS**（documentation）：Running Feast with Snowflake/GCP/AWS 证据：`docs/how-to-guides/feast-snowflake-gcp-aws/README.md` - **🧠 ComputeEngine WIP**（documentation）：The ComputeEngine is Feast’s pluggable abstraction for executing feature pipelines — including transformations, aggregations, joins, and materializations/get historical features — on a backend of your choice e.g., Spark, PyArrow, Pandas, Ray . 证据：`docs/reference/compute-engine/README.md` - **Data sources**（documentation）：Please see Data Source ../../getting-started/concepts/data-ingestion.md for a conceptual explanation of data sources. 证据：`docs/reference/data-sources/README.md` - **Feature repository**（documentation）：Feast users use Feast to manage two important sets of configuration: 证据：`docs/reference/feature-repository/README.md` - **Feast servers**（documentation）：Feast users can choose to retrieve features from a feature server, as opposed to through the Python SDK. 证据：`docs/reference/feature-servers/README.md` - **Offline stores**（documentation）：Please see Offline Store ../../getting-started/components/offline-store.md for a conceptual explanation of offline stores. 证据：`docs/reference/offline-stores/README.md` - **Online stores**（documentation）：Please see Online Store ../../getting-started/components/online-store.md for an explanation of online stores. 证据：`docs/reference/online-stores/README.md` - **Providers**（documentation）：Please see Provider ../../getting-started/components/provider.md for an explanation of providers. 证据：`docs/reference/providers/README.md` - **Registies**（documentation）：Please see Registry ../../getting-started/components/registry.md for a conceptual explanation of registries. 证据：`docs/reference/registries/README.md` - **Getting started with Feast on Azure**（documentation）：Getting started with Feast on Azure 证据：`docs/tutorials/azure/README.md` - **Sample use-case tutorials**（documentation）：These Feast tutorials showcase how to use Feast to simplify end to end model training / serving. 证据：`docs/tutorials/tutorials-overview/README.md` - **Join us on Slack!**（documentation）：! PyPI - Downloads https://img.shields.io/pypi/dm/feast https://pypi.org/project/feast/ ! GitHub contributors https://img.shields.io/github/contributors/feast-dev/feast https://github.com/feast-dev/feast/graphs/contributors ! unit-tests https://github.com/feast-dev/feast/actions/workflows/unit tests.yml/badge.svg?branch=master&event=pull request https://github.com/feast-dev/feast/actions/workflows/unit tests.yml ! integration-tests-and-build https://github.com/feast-dev/feast/actions/workflows/master only.yml/badge.svg?branch=master&event=push https://github.com/feast-dev/feast/actions/workflows/master only.yml ! linter https://github.com/feast-dev/feast/actions/workflows/linter.yml/badge.s… 证据：`README.md` - **Feast Community**（documentation）：Please see the Community section on Feast.dev https://docs.feast.dev/ for more details on getting involved. 证据：`community/README.md` - **Feast Examples**（documentation）：The following examples illustrate various Feast use cases to enhance understanding of its functionality. 证据：`examples/README.md` - **Build and Run**（documentation）：Update 10/31/2024 This Go feature server code is updated from the Expedia Group's forked Feast branch https://github.com/ExpediaGroup/feast.git on 10/22/2024. Thanks the engineers of the Expedia Groups who contributed and improved the Go Feature Server. 证据：`go/README.md` - **Feast Java components**（documentation）：This repository contains the following Feast components. Feast Serving: A gRPC service used to serve the latest feature values to models. Feast Serving Client: A client used to retrieve features from Feast Serving. 证据：`java/README.md` - **Feast Protos**（documentation）：Shared protobuf files across Feast components. 证据：`protos/README.md` - **Experimental Feast Web UI**（documentation）：! Sample UI https://github.com/feast-dev/feast/blob/master/ui/sample.png 证据：`ui/README.md` - **Feast-Powered AI Agent Example**（documentation）：This example demonstrates an AI agent with persistent memory that uses Feast as both a feature store and a context memory layer through the Model Context Protocol MCP . This demo uses Milvus as the vector-capable online store, but Feast supports multiple vector backends -- including Milvus, Elasticsearch, Qdrant, PGVector, and FAISS -- swappable via configuration. 证据：`examples/agent_feature_store/README.md` - **Feast Credit Risk Classification End-to-End Example**（documentation）：! Feast Logo https://raw.githubusercontent.com/feast-dev/feast/master/docs/assets/feast logo.png 证据：`examples/credit-risk-end-to-end/README.md` - **Running Feast Java Server with Redis & calling with python with registry in GCP**（documentation）：Running Feast Java Server with Redis & calling with python with registry in GCP 证据：`examples/java-demo/README.md` - **Install and run Feast with Kind**（documentation）：The following notebooks will guide you through an end-to-end journey to install and validate a simple Feast feature store in a Kind Kubernetes cluster: 01-Install.ipynb ./01-Install.ipynb : Install and configure the cluster, then the Feast components. 02-Client.ipynb ./02-Client.ipynb : Validate the feature store with a remote test application runnning on the notebook. 03-Uninstall.ipynb ./03-Uninstall.ipynb : Clear the installed deployments. 证据：`examples/kind-quickstart/README.md` - **Feast MCP Feature Server Example**（documentation）：This example demonstrates how to enable MCP Model Context Protocol support in Feast, allowing AI agents and applications to interact with your features through standardized MCP interfaces. 证据：`examples/mcp_feature_store/README.md` - **Milvus Tutorial with Feast**（documentation）：This tutorial demonstrates how to use Milvus as a vector database backend for Feast. You'll learn how to set up Milvus, create embeddings, store them in Feast, and perform similarity searches. 证据：`examples/online_store/milvus_tutorial/README.md` - **PGVector Tutorial with Feast**（documentation）：This tutorial demonstrates how to use PostgreSQL with the pgvector extension as a vector database backend for Feast. You'll learn how to set up pgvector, create embeddings, store them in Feast, and perform similarity searches. 证据：`examples/online_store/pgvector_tutorial/README.md` - **Feast OpenLineage Integration Example**（documentation）：Feast OpenLineage Integration Example 证据：`examples/openlineage-integration/README.md` - **Installing Feast on Kubernetes with PostgreSQL TLS Demo using feast operator**（documentation）：Installing Feast on Kubernetes with PostgreSQL TLS Demo using feast operator 证据：`examples/operator-postgres-tls-demo/README.md` - **Install and run a Feature Store on Kubernetes with the Feast Operator**（documentation）：Install and run a Feature Store on Kubernetes with the Feast Operator 证据：`examples/operator-quickstart/README.md` - **Feast Operator RBAC with TLS OpenShift**（documentation）：Feast Operator RBAC with TLS OpenShift 证据：`examples/operator-rbac-openshift-tls/README.md` - **Running the Feast RBAC example on Kubernetes using the Feast Operator.**（documentation）：Running the Feast RBAC example on Kubernetes using the Feast Operator. 证据：`examples/operator-rbac/README.md` - **Feast example using Podman and Podman Compose**（documentation）：Feast example using Podman and Podman Compose 证据：`examples/podman_local/README.md` - **Running Feast Python / Go Feature Server with Redis on Kubernetes**（documentation）：Running Feast Python / Go Feature Server with Redis on Kubernetes 证据：`examples/python-helm-demo/README.md` - **🚀 Quickstart: RAG, Milvus, and Docling with Feast**（documentation）：🚀 Quickstart: RAG, Milvus, and Docling with Feast 证据：`examples/rag-docling/README.md` - **End-to-end RAG Fine Tuning example using Feast and Milvus.**（documentation）：End-to-end RAG Fine Tuning example using Feast and Milvus. 证据：`examples/rag-retriever/README.md` - **🚀 Quickstart: Retrieval-Augmented Generation RAG using Feast and Large Language Models LLMs**（documentation）：🚀 Quickstart: Retrieval-Augmented Generation RAG using Feast and Large Language Models LLMs 证据：`examples/rag/README.md` - **RBAC demo**（documentation）：RBAC demo RBAC demo with local environment. 证据：`examples/rbac-local/README.md` - **Feast Deployment with RBAC**（documentation）：Demo Summary This demo showcases how to enable Role-Based Access Control RBAC for Feast using Kubernetes or OIDC https://openid.net/developers/how-connect-works/ Authentication type. The demo steps involve deploying server components registry, offline, online and client examples within a Kubernetes environment. The goal is to ensure secure access control based on user roles and permissions. For understanding the Feast RBAC framework Please read these reference documents. - RBAC Architecture https://docs.feast.dev/v/master/getting-started/architecture/rbac - RBAC Permission https://docs.feast.dev/v/master/getting-started/concepts/permission . - RBAC Authorization Manager https://docs.feast.d… 证据：`examples/rbac-remote/README.md` - **Feast Remote Offline Store Server**（documentation）：This example demonstrates the steps using an Arrow Flight https://arrow.apache.org/blog/2019/10/13/introducing-arrow-flight/ server/client as the remote Feast offline store. 证据：`examples/remote-offline-store/README.md` - **Quickstart: Running Feast example**（documentation）：This quickstart guide will walk you through setting up and using Feast https://feast.dev as a feature store on Red Hat OpenShift AI. By the end of this tutorial, you’ll have the environment configured, sample data loaded, and features retrieved using Feast objects. 证据：`examples/rhoai-quickstart/README.md` - **Feast Python / Go Feature Server Helm Charts**（documentation）：Feast Python / Go Feature Server Helm Charts 证据：`infra/charts/feast-feature-server/README.md` - **Feast Java Helm Charts alpha**（documentation）：This repo contains Helm charts for Feast Java components that are being installed on Kubernetes: Feast root chart : The complete Helm chart containing all Feast components and dependencies. Most users will use this chart, but can selectively enable/disable subcharts using the values.yaml file. Feature Server charts/feature-server : High performant JVM-based implementation of feature server. Transformation Service charts/transformation-service : Transformation server for calculating on-demand features Redis: Optional One of possible options for an online store used by Feature Server 证据：`infra/charts/feast/README.md` - **feature-server**（documentation）：! Version: 0.64.0 https://img.shields.io/badge/Version-0.64.0-informational?style=flat-square ! AppVersion: v0.64.0 https://img.shields.io/badge/AppVersion-v0.64.0-informational?style=flat-square 证据：`infra/charts/feast/charts/feature-server/README.md` - **transformation-service**（documentation）：! Version: 0.64.0 https://img.shields.io/badge/Version-0.64.0-informational?style=flat-square ! AppVersion: v0.64.0 https://img.shields.io/badge/AppVersion-v0.64.0-informational?style=flat-square 证据：`infra/charts/feast/charts/transformation-service/README.md` - **Feast Operator**（documentation）：Feast Operator This is a K8s Operator that can be used to deploy and manage Feast , an open source feature store for machine learning. 证据：`infra/feast-operator/README.md` - **Terraform config for feast on AWS**（documentation）：1. Run aws emr create-default-roles once. 证据：`infra/terraform/aws/README.md` - **Terraform config for Feast on Azure**（documentation）：Terraform config for Feast on Azure 证据：`infra/terraform/azure/README.md` - **Terraform config for feast on GCP**（documentation）：This serves as a guide on how to deploy Feast on GCP. At the end of this guide, we will have provisioned: 1. GKE cluster 2. Feast services running on GKE 3. Google Memorystore Redis as online store 4. Dataproc cluster 4. Kafka running on GKE, exposed to the dataproc cluster via internal load balancer. 证据：`infra/terraform/gcp/README.md` - **Feast Website**（documentation）：Feast Website Running Locally Add blog posts See docs/blog/ for examples 证据：`infra/website/README.md` - **Readme**（documentation）：Feast Data Types for Java ========================= 证据：`java/datatypes/README.md` - **Getting Started Guide for Feast Serving Developers**（documentation）：Getting Started Guide for Feast Serving Developers 证据：`java/serving/README.md` - **DAG**（documentation）：DAG Directed Acyclic Graph DAG representation for feature engineering workflows in Feast. This module is designed to remain abstracted and independent of Feast feature view business logic. 证据：`sdk/python/feast/infra/compute_engines/dag/README.md` - **MongoDB Offline Store**（documentation）：This offline store lets you train models and run batch scoring directly from it. All feature views share a single collection. Reads use MongoDB aggregation pipelines with a compound index, so per-entity cost is O log n observations regardless of collection size. 证据：`sdk/python/feast/infra/offline_stores/contrib/mongodb_offline_store/README.md` - 其余 23 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：`docs/README.md`, `docs/adr/README.md`, `docs/blog/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：`docs/README.md`, `docs/adr/README.md`, `docs/blog/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。 - **Overview & Core Architecture**：importance `high` - source_paths: README.md, sdk/python/feast/feature_store.py, sdk/python/feast/feature_view.py, sdk/python/feast/entity.py, sdk/python/feast/feature_service.py - **Data Infrastructure, Stores & Retrieval**：importance `high` - source_paths: sdk/python/feast/infra/registry/registry.py, sdk/python/feast/infra/registry/sql.py, sdk/python/feast/infra/online_stores/redis.py, sdk/python/feast/infra/online_stores/dynamodb.py, sdk/python/feast/infra/offline_stores/snowflake.py - **Feature Servers, Deployment & Operations**：importance `high` - source_paths: sdk/python/feast/feature_server.py, sdk/python/feast/offline_server.py, sdk/python/feast/registry_server.py, go/main.go, go/internal/feast/server/grpc_server.go - **Web UI, Extensibility & Integrations**：importance `medium` - source_paths: ui/src/FeastUI.tsx, ui/src/pages/RegistryVisualizationTab.tsx, ui/src/pages/feature-views/FeatureViewInstance.tsx, ui/src/pages/entities/EntityInstance.tsx, ui/src/pages/label-views/LabelViewInstance.tsx ## Repo Inspection Evidence / 源码检查证据 - repo_clone_verified: true - repo_inspection_verified: true - repo_commit: `f0c5be8a66f3d3f2c7a6e560b926b25d53f3042f` - inspected_files: `README.md`, `pyproject.toml`, `docs/README.md`, `docs/SUMMARY.md`, `docs/adr/ADR-0001-feature-services.md`, `docs/adr/ADR-0002-component-refactor.md`, `docs/adr/ADR-0003-on-demand-transformations.md`, `docs/adr/ADR-0004-entity-join-key-mapping.md`, `docs/adr/ADR-0005-stream-transformations.md`, `docs/adr/ADR-0006-kubernetes-operator.md`, `docs/adr/ADR-0007-unified-feature-transformations.md`, `docs/adr/ADR-0008-feature-view-versioning.md`, `docs/adr/ADR-0009-contribution-extensibility.md`, `docs/adr/ADR-0010-vector-database-integration.md`, `docs/adr/ADR-0011-data-quality-monitoring.md`, `docs/adr/ADR-0012-label-view.md`, `docs/adr/ADR-TEMPLATE.md`, `docs/adr/README.md`, `docs/adr/rfc-feature-view-versioning.md`, `docs/blog/README.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：get_online_features on remote stores drops all join keys but the first - 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：get_online_features on remote stores drops all join keys but the first - Why it matters: 可能增加新用户试用和生产接入成本。 - Evidence: community_evidence:github | https://github.com/feast-dev/feast/issues/6488 | 来源讨论提到 python 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 | https://github.com/feast-dev/feast | 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: 来源证据：0.63.0 -> 0.64.0: UI breaks in demo - Trigger: GitHub 社区证据显示该项目存在一个维护/版本相关的待验证问题：0.63.0 -> 0.64.0: UI breaks in demo - Why it matters: 可能增加新用户试用和生产接入成本。 - Evidence: community_evidence:github | https://github.com/feast-dev/feast/issues/6539 | 来源讨论提到 macos 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 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 | https://github.com/feast-dev/feast | 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。 - Trigger: no_demo - Evidence: downstream_validation.risk_items | https://github.com/feast-dev/feast | 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 - Why it matters: 风险会影响是否适合普通用户安装。 - Evidence: risks.scoring_risks | https://github.com/feast-dev/feast | 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 | https://github.com/feast-dev/feast | 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 | https://github.com/feast-dev/feast | 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。