# instructor - 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 instructor 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

- **正在使用 Claude/Codex/Cursor/Gemini 等宿主 AI 的开发者**：README 或插件配置提到多个宿主 AI。 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：`CLAUDE.md`, `README.md` Claim：`clm_0001` supported 0.86

## 怎么开始

- `pip install instructor` 证据：`README.md` Claim：`clm_0003` supported 0.86, `clm_0005` supported 0.86, `clm_0006` supported 0.86, `clm_0007` supported 0.86
- `pip install -r requirements-doc.txt` 证据：`CLAUDE.md` Claim：`clm_0004` supported 0.86
- `pip install "instructor[anthropic]"` 证据：`CLAUDE.md` Claim：`clm_0005` supported 0.86
- `pip install "instructor[google-generativeai]"` 证据：`CLAUDE.md` Claim：`clm_0006` supported 0.86
- `pip install "instructor[groq]"` 证据：`CLAUDE.md` Claim：`clm_0007` supported 0.86

## 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：正在使用 Claude/Codex/Cursor/Gemini 等宿主 AI 的开发者**（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：`CLAUDE.md`, `README.md` 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：`README.md` Claim：`clm_0003` supported 0.86, `clm_0005` supported 0.86, `clm_0006` supported 0.86, `clm_0007` 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：`CLAUDE.md`, `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：`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：`CLAUDE.md`, `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**：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_0008` 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：`CLAUDE.md`, `README.md` Claim：`clm_0009` 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：`CLAUDE.md`, `README.md` Claim：`clm_0001` supported 0.86

### 上下文规模

- 文件总数：712
- 重要文件覆盖：40/712
- 证据索引条目：79
- 角色 / Skill 条目：79

### 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 instructor 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 instructor 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 instructor 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 79 个角色 / Skill / 项目文档条目。

- **Contributing to Instructor**（project_doc）：Join us in enhancing the Instructor library with evals, report issues, and submit pull requests on GitHub. Collaborate and contribute! 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/contributing.md`
- **CLAUDE.md**（project_doc）：This file provides guidance to Claude Code claude.ai/code when working with code in this repository. 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`
- **Instructor: Structured Outputs for LLMs**（project_doc）：Instructor: Structured Outputs for LLMs 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`
- **Scripts Directory**（project_doc）：This directory contains utility scripts for maintaining and improving the Instructor documentation and project structure. 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`
- **Cursor Rules**（project_doc）：Cursor rules are configuration files that help guide AI-assisted development in the Cursor IDE. They provide structured instructions for how the AI should behave in specific contexts or when working with certain types of 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：`.cursor/rules/readme.md`
- **Batch API Examples**（project_doc）：This directory contains examples and test scripts for Instructor's batch processing capabilities, including both traditional file-based and new in-memory processing. 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：`examples/batch_api/README.md`
- **Instructor Caching Prototype**（project_doc）：This example demonstrates the new built-in caching functionality in Instructor. 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：`examples/caching_prototype/README.md`
- **Introduction**（project_doc）：This is a simple example which shows how to perform Chain Of Density summarization using GPT-3.5 and utilise the generated output to fine-tune a 3.5 model for production usage. All of our data referenced in this file is located here https://huggingface.co/datasets/ivanleomk/gpt4-chain-of-density on hugging face 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：`examples/chain-of-density/Readme.md`
- **Citation with Extraction**（project_doc）：This repository contains a FastAPI application that uses GPT-4 to answer questions based on a given context and extract relevant facts with correct and exact citations. The extracted facts are returned as JSON events using Server-Sent Events SSE . 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：`examples/citation_with_extraction/README.md`
- **FastAPI Code Generator**（project_doc）：Generates FastAPI application code from API path, task name, JSON schema path, and Jinja2 prompt template. Also creates a models.py file for Pydantic models. 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：`examples/codegen-from-schema/readme.md`
- **What to Expect**（project_doc）：What to Expect This script demonstrates how to use the Instructor library for fine-tuning a Python function that performs three-digit multiplication. It uses Pydantic for type validation and logging features to generate a fine-tuning dataset. 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：`examples/distilations/readme.md`
- **Instructor Hooks Example**（project_doc）：This example demonstrates how to use the Hooks system in the Instructor library to monitor, log, and debug your LLM interactions. 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：`examples/hooks/README.md`
- **Instructions**（project_doc）：1. Create a virtual environment and install all of the packages inside requirements.txt 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：`examples/logfire-fastapi/Readme.md`
- **Read first to correctly work with the provided examples**（project_doc）：Read first to correctly work with the provided examples 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：`examples/open_source_examples/README.md`
- **Using llm validator with OpenAI's GPT-3.5 Turbo and Pydantic for Text Validation with Output Examples**（project_doc）：Using llm validator with OpenAI's GPT-3.5 Turbo and Pydantic for Text Validation with Output Examples 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：`examples/validators/readme.md`
- **Providers Directory Structure**（project_doc）：This directory contains implementations for all supported LLM providers in the instructor 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：`instructor/providers/README.md`
- **Contributing to Instructor**（project_doc）：Thank you for considering contributing to Instructor! This document provides guidelines and instructions to help you contribute effectively. 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`
- **Core Provider Tests**（project_doc）：This directory contains unified tests that run across all core providers : OpenAI, Anthropic, Google Gemini , Cohere, xAI, Mistral, Cerebras, Fireworks, Writer, and Perplexity. 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：`tests/llm/test_core_providers/README.md`
- **AGENT.md - Documentation**（project_doc）：Internal guide for maintaining and improving Instructor documentation 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.md`
- **API Docstring Quality Assessment**（project_doc）：This document assesses the quality and completeness of docstrings for all API items referenced in the expanded API documentation. 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/api-docstring-assessment.md`
- **API Reference**（project_doc）：Explore the comprehensive API reference with details on instructors, validation, iteration, and function calls. 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/api.md`
- **Architecture Overview**（project_doc）：Learn about the internal architecture and design decisions of the Instructor 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：`docs/architecture.md`
- **Debugging**（project_doc）：Learn how to debug Instructor applications with hooks, logging, and exception handling. Practical techniques for inspecting inputs, outputs, and retries. 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/debugging.md`
- **Frequently Asked Questions**（project_doc）：Common questions and answers about using Instructor 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/faq.md`
- **Getting Started with Instructor**（project_doc）：A step-by-step guide to getting started with Instructor for structured outputs from LLMs 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/getting-started.md`
- **Getting help with Instructor**（project_doc）：Explore key resources for getting help with Instructor, including Discord, blog, concepts, cookbooks, and GitHub discussions. 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/help.md`
- **Instructor: Top Multi-Language Library for Structured LLM Outputs**（project_doc）：Get structured, validated data from any LLM with Instructor - the 1 library for LLM data extraction. Supports 15+ providers OpenAI, Anthropic, Google, Ollama, DeepSeek in 6 languages. Built on type-safe schemas with automatic retries, streaming, and nested object 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/index.md`
- **Installation**（project_doc）：Learn how to install Instructor and its dependencies using pip for Python 3.9+. Simple setup guide included. 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/installation.md`
- **Jobs**（project_doc）： 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/jobs.md`
- **Instructor Mode Comparison Guide**（project_doc）：Compare different modes available in Instructor and understand when to use each 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/modes-comparison.md`
- **Instructor Newsletter**（project_doc）：Get notified about AI tips, blog posts, and research. Stay informed with Instructor's latest features and community insights. 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/newsletter.md`
- **Repository Overview**（project_doc）：Learn the structure of the Instructor repository and the purpose of each major directory. 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/repository-overview.md`
- **Start Here: Instructor for Beginners**（project_doc）：A beginner-friendly introduction to using Instructor for structured outputs from LLMs 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/start-here.md`
- **Why use Instructor?**（project_doc）：Discover why Instructor is the simplest, most reliable way to get structured outputs from LLMs. 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/why.md`
- **Subscribe to our Newsletter for Updates and Tips**（project_doc）：Subscribe to our Newsletter for Updates and Tips 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/blog/index.md`
- **AI Engineer Keynote: Pydantic is all you need**（project_doc）：Explore insights on utilizing Pydantic for effective prompt engineering 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/blog/posts/aisummit-2023.md`
- **Structured Outputs for Gemini now supported**（project_doc）：Introducing structured outputs for Gemini tool calling support in the 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/blog/posts/announcing-gemini-tool-calling-support.md`
- **Announcing Responses API support**（project_doc）：Take advantage of OpenAI's latest offerings with the new responses API 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/blog/posts/announcing-instructor-responses-support.md`
- **What is from provider ?**（project_doc）：Switch between different models and providers with a single string! 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/blog/posts/announcing-unified-provider-interface.md`
- **Why should I use prompt caching?**（project_doc）：Discover how prompt caching with Anthropic can improve response times 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/blog/posts/anthropic-prompt-caching.md`
- **Using Anthropic's Web Search with Instructor for Real-Time Data**（project_doc）：Using Anthropic's Web Search with Instructor for Real-Time Data 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/blog/posts/anthropic-web-search-structured.md`
- **Structured Outputs with Anthropic**（project_doc）：Learn how to integrate Anthropic's powerful language models into your projects using Instructor, with step-by-step guidance on installation, client setup, and creating structured outputs with Pydantic models. 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/blog/posts/anthropic.md`
- **Bad Schemas could break your LLM Structured Outputs**（project_doc）：Discover how response models impact LLM performance, focusing on structured 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/blog/posts/bad-schemas-could-break-llms.md`
- **Why Instructor is the Best Library for Structured LLM Outputs**（project_doc）：Discover how the Instructor library simplifies structured LLM outputs 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/blog/posts/best_framework.md`
- **Advanced Caching Strategies for Python LLM Applications Validated & Tested ✅**（project_doc）：Master advanced Python caching strategies for LLM applications using functools, diskcache, and Redis. Learn how to optimize OpenAI API costs, reduce response times, and implement efficient caching for Pydantic models in production environments. 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/blog/posts/caching.md`
- **Smarter Summaries w/ Finetuning GPT-3.5 and Chain of Density**（project_doc）：Learn to implement Chain of Density with GPT-3.5 for improved summarization, 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/blog/posts/chain-of-density.md`
- **PDF Processing with Structured Outputs with Gemini**（project_doc）：Learn how to use Google's Gemini model with Instructor to process PDFs and extract structured 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：`docs/blog/posts/chat-with-your-pdf-with-gemini.md`
- **Verifying LLM Citations with Pydantic**（project_doc）：Explore how Pydantic enhances LLM citation verification, improving data 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/blog/posts/citations.md`
- **Consistent Stories with GPT-4o**（project_doc）：Generating complex DAGS with gpt-4o 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/blog/posts/consistent-stories.md`
- **Free course on Weights and Biases**（project_doc）：Discover a free one-hour course on Weights and Biases covering essential 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/blog/posts/course.md`
- **Instructor Adopting Cursor Rules**（project_doc）：AI-assisted coding is changing how we use version control. Many developers now use what I call "vibe coding" - coding with AI help. This creates new challenges with Git. Today I'll share how we're using Cursor rules in Instructor to solve these problems. 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/blog/posts/cursor-rules.md`
- **Enhancing Python Functions with Instructor: A Guide to Fine-Tuning and Distillation**（project_doc）：Explore Instructor for fine-tuning language models with Python, simplifying 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/blog/posts/distilation-part1.md`
- **Consistent Stories with GPT-4o**（project_doc）：Generating complex DAGS with gpt-4o 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/blog/posts/extract-model-looks.md`
- **Why Image Metadata is useful**（project_doc）：Structured Extraction makes working with images easy, in this post we'll see how to use it to extract metadata from images 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/blog/posts/extracting-model-metadata.md`
- **Simple Synthetic Data Generation**（project_doc）：Learn to generate synthetic data using Pydantic and OpenAI's models 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/blog/posts/fake-data.md`
- **Why Logfire is a perfect fit for FastAPI + Instructor**（project_doc）：Discover how Logfire enhances FastAPI applications with OpenTelemetry 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/blog/posts/full-fastapi-visibility.md`
- **Eliminating Hallucinations with Structured Outputs using Gemini**（project_doc）：Generate accurate citations and eliminate hallucinations with structured outputs using Gemini. 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/blog/posts/generating-pdf-citations.md`
- **Generators and LLM Streaming**（project_doc）：Explore Python generators and their role in enhancing LLM streaming for 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/blog/posts/generator.md`
- **Do I Still Need Instructor with Google's New OpenAI Integration?**（project_doc）：Learn why Instructor remains essential even with Google's new OpenAI-compatible client for Gemini 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/blog/posts/google-openai-client.md`
- **Introducing structured outputs with Cerebras Inference**（project_doc）：Introducing structured outputs with Cerebras Inference 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/blog/posts/introducing-structured-outputs-with-cerebras-inference.md`
- **Should I Be Using Structured Outputs?**（project_doc）：Explore the challenges of OpenAI's Structured Outputs and how 'instructor 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/blog/posts/introducing-structured-outputs.md`
- **Generating Structured Output / JSON from LLMs**（project_doc）：Learn how Pydantic simplifies working with LLMs and structured JSON outputs 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/blog/posts/introduction.md`
- **Instructor Proposal: Integrating Jinja Templating**（project_doc）：Explore the integration of Jinja templating in the Instructor for enhanced 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/blog/posts/jinja-proposal.md`
- **Seamless Support with Langsmith**（project_doc）：Explore how LangSmith enhances OpenAI clients with seamless LLM observability 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/blog/posts/langsmith.md`
- **Mastering Python asyncio.gather and asyncio.as completed for LLM Processing**（project_doc）：Master Python asyncio.gather and asyncio.as completed for efficient concurrent LLM processing with Instructor. Learn async programming patterns, rate limiting, and performance optimization for AI applications. 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/blog/posts/learn-async.md`
- **Building an LLM-based Reranker for your RAG pipeline**（project_doc）：Learn how to use Instructor and Pydantic to create an LLM-based reranker for improving search results relevance. 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/blog/posts/llm-as-reranker.md`
- **Instructor Adopts llms.txt: Making Documentation AI-Friendly**（project_doc）：Instructor Adopts llms.txt: Making Documentation AI-Friendly 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/blog/posts/llms-txt-adoption.md`
- **Instructor Now Supports llms.txt**（project_doc）：We've added automatic llms.txt generation to Instructor's documentation using the mkdocs-llmstxt https://github.com/pawamoy/mkdocs-llmstxt plugin. 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/blog/posts/llms-txt-support.md`
- **Introduction**（project_doc）：Explore Logfire, an observability platform to enhance application performance 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/blog/posts/logfire.md`
- **London Stock Exchange Group Powers Market Surveillance with Instructor**（project_doc）：London Stock Exchange Group uses Instructor in production for AI-powered market surveillance, achieving 100% precision in detecting price-sensitive news 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/blog/posts/lseg-market-surveillance.md`
- **Matching Language in Multilingual Summarization Tasks**（project_doc）：Explore techniques to ensure language models generate summaries that 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/blog/posts/matching-language.md`
- **Why we migrated to uv**（project_doc）：How we migrated from poetry to uv 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/blog/posts/migrating-to-uv.md`
- **Automating llms.txt Generation with mkdocs-llmstxt Plugin**（project_doc）：Automating llms.txt Generation with mkdocs-llmstxt Plugin 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/blog/posts/mkdocs-llmstxt-plugin-integration.md`
- **Structured Outputs with Multimodal Gemini**（project_doc）：Learn how to use Google's Gemini model for multimodal structured extraction of YouTube videos, extracting structured recommendations for tourist destinations. 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/blog/posts/multimodal-gemini.md`
- **Native Caching in Instructor v1.9.1: Zero-Configuration Performance Boost**（project_doc）：Instructor v1.9.1 introduces native caching support for all providers. Learn how to drastically reduce API costs and improve response times with built-in cache adapters. 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/blog/posts/native_caching.md`
- **Structured Output for Open Source and Local LLMs**（project_doc）：Discover how Instructor integrates with OpenAI and local LLMs for structured 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/blog/posts/open_source.md`
- **OpenAI API Model Distillation with Instructor**（project_doc）：Learn how to use OpenAI's API Model Distillation with Instructor to create 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/blog/posts/openai-distilation-store.md`
- **Audio Support in OpenAI's Chat Completions API**（project_doc）：Explore the new audio capabilities in OpenAI's Chat Completions API using the gpt-4o-audio-preview model. 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/blog/posts/openai-multimodal.md`
- **Building a Pairwise LLM Judge with Instructor and Pydantic**（project_doc）：Explore how to use Instructor and Pydantic to create a pairwise LLM judge for evaluating text relevance. 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/blog/posts/pairwise-llm-judge.md`

## 证据索引

- 共索引 79 条证据。

- **Contributing to Instructor**（documentation）：We welcome contributions to Instructor! This page covers the different ways you can help improve the library. 证据：`docs/contributing.md`
- **CLAUDE.md**（documentation）：This file provides guidance to Claude Code claude.ai/code when working with code in this repository. 证据：`CLAUDE.md`
- **Instructor: Structured Outputs for LLMs**（documentation）：Instructor: Structured Outputs for LLMs 证据：`README.md`
- **Scripts Directory**（documentation）：This directory contains utility scripts for maintaining and improving the Instructor documentation and project structure. 证据：`scripts/README.md`
- **Cursor Rules**（documentation）：Cursor rules are configuration files that help guide AI-assisted development in the Cursor IDE. They provide structured instructions for how the AI should behave in specific contexts or when working with certain types of files. 证据：`.cursor/rules/readme.md`
- **Batch API Examples**（documentation）：This directory contains examples and test scripts for Instructor's batch processing capabilities, including both traditional file-based and new in-memory processing. 证据：`examples/batch_api/README.md`
- **Instructor Caching Prototype**（documentation）：This example demonstrates the new built-in caching functionality in Instructor. 证据：`examples/caching_prototype/README.md`
- **Introduction**（documentation）：This is a simple example which shows how to perform Chain Of Density summarization using GPT-3.5 and utilise the generated output to fine-tune a 3.5 model for production usage. All of our data referenced in this file is located here https://huggingface.co/datasets/ivanleomk/gpt4-chain-of-density on hugging face 证据：`examples/chain-of-density/Readme.md`
- **Citation with Extraction**（documentation）：This repository contains a FastAPI application that uses GPT-4 to answer questions based on a given context and extract relevant facts with correct and exact citations. The extracted facts are returned as JSON events using Server-Sent Events SSE . 证据：`examples/citation_with_extraction/README.md`
- **FastAPI Code Generator**（documentation）：Generates FastAPI application code from API path, task name, JSON schema path, and Jinja2 prompt template. Also creates a models.py file for Pydantic models. 证据：`examples/codegen-from-schema/readme.md`
- **What to Expect**（documentation）：What to Expect This script demonstrates how to use the Instructor library for fine-tuning a Python function that performs three-digit multiplication. It uses Pydantic for type validation and logging features to generate a fine-tuning dataset. 证据：`examples/distilations/readme.md`
- **Instructor Hooks Example**（documentation）：This example demonstrates how to use the Hooks system in the Instructor library to monitor, log, and debug your LLM interactions. 证据：`examples/hooks/README.md`
- **Instructions**（documentation）：1. Create a virtual environment and install all of the packages inside requirements.txt 证据：`examples/logfire-fastapi/Readme.md`
- **Read first to correctly work with the provided examples**（documentation）：Read first to correctly work with the provided examples 证据：`examples/open_source_examples/README.md`
- **Using llm validator with OpenAI's GPT-3.5 Turbo and Pydantic for Text Validation with Output Examples**（documentation）：Using llm validator with OpenAI's GPT-3.5 Turbo and Pydantic for Text Validation with Output Examples 证据：`examples/validators/readme.md`
- **Providers Directory Structure**（documentation）：This directory contains implementations for all supported LLM providers in the instructor library. 证据：`instructor/providers/README.md`
- **Contributing to Instructor**（documentation）：Thank you for considering contributing to Instructor! This document provides guidelines and instructions to help you contribute effectively. 证据：`CONTRIBUTING.md`
- **Core Provider Tests**（documentation）：This directory contains unified tests that run across all core providers : OpenAI, Anthropic, Google Gemini , Cohere, xAI, Mistral, Cerebras, Fireworks, Writer, and Perplexity. 证据：`tests/llm/test_core_providers/README.md`
- **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：`LICENSE`
- **AGENT.md - Documentation**（documentation）：Commands - Serve docs locally: uv run mkdocs serve - Build docs: ./build mkdocs.sh or uv run mkdocs build - Install doc deps: uv pip install -e ". docs " - Test examples: uv run pytest docs/ --examples 证据：`docs/AGENT.md`
- **API Docstring Quality Assessment**（documentation）：This document assesses the quality and completeness of docstrings for all API items referenced in the expanded API documentation. 证据：`docs/api-docstring-assessment.md`
- **API Reference**（documentation）：Core modes are the recommended default. Legacy provider-specific modes still work but are deprecated and will show warnings. See the Mode Migration Guide concepts/mode-migration.md for details. 证据：`docs/api.md`
- **Architecture Overview**（documentation）：This page explains the core execution flow and where to plug in or debug. It highlights the minimal sync/async code paths and how streaming, partial, and parallel modes integrate. 证据：`docs/architecture.md`
- **Debugging**（documentation）：This guide shows how to quickly inspect inputs/outputs, capture retries, and reproduce failures when working with Instructor. It focuses on practical techniques using hooks, logging, and exception data. 证据：`docs/debugging.md`
- **Frequently Asked Questions**（documentation）：This page answers common questions about using Instructor with various LLM providers. 证据：`docs/faq.md`
- **Getting Started with Instructor**（documentation）：This guide will walk you through the basics of using Instructor to extract structured data from language models. By the end, you'll understand how to: 证据：`docs/getting-started.md`
- **Getting help with Instructor**（documentation）：If you need help getting started with Instructor or with advanced usage, the following sources may be useful. 证据：`docs/help.md`
- **Instructor: Top Multi-Language Library for Structured LLM Outputs**（documentation）：Instructor: Top Multi-Language Library for Structured LLM Outputs 证据：`docs/index.md`
- **Installation**（documentation）：- openai https://pypi.org/project/openai/ : OpenAI's Python client. - typer https://pypi.org/project/typer/ : Build great CLIs. Easy to code. Based on Python type hints. - docstring-parser https://pypi.org/project/docstring-parser/ : A parser for Python docstrings, to improve the experience of working with docstrings in jsonschema. - pydantic https://pypi.org/project/pydantic/ : Data validation and settings management using python type annotations. 证据：`docs/installation.md`
- **Instructor Mode Comparison Guide**（documentation）：Instructor uses core modes that work across providers. Provider-specific modes still work, but they are deprecated and will show warnings. 证据：`docs/modes-comparison.md`
- **Instructor Newsletter**（documentation）：If you want to be notified of tips, new blog posts, and research, subscribe to our newsletter. Here's what you can expect: 证据：`docs/newsletter.md`
- **Repository Overview**（documentation）：This page explains the layout of the Instructor codebase and what each key directory contains. 证据：`docs/repository-overview.md`
- **Start Here: Instructor for Beginners**（documentation）：Start Here: Instructor for Beginners 证据：`docs/start-here.md`
- **Why use Instructor?**（documentation）：You've built something with an LLM, but 15% of the time it returns garbage. Parsing JSON is a nightmare. Different providers have different APIs. There has to be a better way. 证据：`docs/why.md`
- **Subscribe to our Newsletter for Updates and Tips**（documentation）：Subscribe to our Newsletter for Updates and Tips 证据：`docs/blog/index.md`
- **AI Engineer Keynote: Pydantic is all you need**（documentation）：AI Engineer Keynote: Pydantic is all you need 证据：`docs/blog/posts/aisummit-2023.md`
- **Structured Outputs for Gemini now supported**（documentation）：Structured Outputs for Gemini now supported 证据：`docs/blog/posts/announcing-gemini-tool-calling-support.md`
- **Announcing Responses API support**（documentation）：We're excited to announce Instructor's integration with OpenAI's new Responses API. This integration brings a more streamlined approach to working with structured outputs from OpenAI models. Let's see what makes this integration special and how it can improve your LLM applications. 证据：`docs/blog/posts/announcing-instructor-responses-support.md`
- **What is from provider ?**（documentation）：We are pleased to introduce a significant enhancement to Instructor: the from provider function. While Instructor has always focused on providing robust structured outputs, we've observed that many users work with multiple LLM providers. This often involves repetitive setup for each client. 证据：`docs/blog/posts/announcing-unified-provider-interface.md`
- **Why should I use prompt caching?**（documentation）：Developers often face two key challenges when working with large context - Slow response times and high costs. This is especially true when we're making multiple of these calls over time, severely impacting the cost and latency of our applications. With Anthropic's new prompt caching feature, we can easily solve both of these issues. 证据：`docs/blog/posts/anthropic-prompt-caching.md`
- **Using Anthropic's Web Search with Instructor for Real-Time Data**（documentation）：Using Anthropic's Web Search with Instructor for Real-Time Data 证据：`docs/blog/posts/anthropic-web-search-structured.md`
- **Structured Outputs with Anthropic**（documentation）：A special shoutout to Shreya https://twitter.com/shreyaw for her contributions to the anthropic support. As of now, all features are operational with the exception of streaming support. 证据：`docs/blog/posts/anthropic.md`
- **Bad Schemas could break your LLM Structured Outputs**（documentation）：Bad Schemas could break your LLM Structured Outputs 证据：`docs/blog/posts/bad-schemas-could-break-llms.md`
- **Why Instructor is the Best Library for Structured LLM Outputs**（documentation）：Why Instructor is the Best Library for Structured LLM Outputs 证据：`docs/blog/posts/best_framework.md`
- **Advanced Caching Strategies for Python LLM Applications Validated & Tested ✅**（documentation）：Advanced Caching Strategies for Python LLM Applications Validated & Tested ✅ 证据：`docs/blog/posts/caching.md`
- **Smarter Summaries w/ Finetuning GPT-3.5 and Chain of Density**（documentation）：Smarter Summaries w/ Finetuning GPT-3.5 and Chain of Density 证据：`docs/blog/posts/chain-of-density.md`
- **PDF Processing with Structured Outputs with Gemini**（documentation）：PDF Processing with Structured Outputs with Gemini 证据：`docs/blog/posts/chat-with-your-pdf-with-gemini.md`
- **Verifying LLM Citations with Pydantic**（documentation）：Verifying LLM Citations with Pydantic 证据：`docs/blog/posts/citations.md`
- **Consistent Stories with GPT-4o**（documentation）：Language Models struggle to generate consistent graphs that have a large number of nodes. Often times, this is because the graph itself is too large for the model to handle. This causes the model to generate inconsistent graphs that have invalid and disconnected nodes among other issues. 证据：`docs/blog/posts/consistent-stories.md`
- **Free course on Weights and Biases**（documentation）：I just released a free course on wits and biases. It goes over the material from tutorial ../../tutorials/1-introduction.ipynb . Check it out at wandb.courses https://www.wandb.courses/courses/steering-language-models its free and open to everyone and just under an hour long! 证据：`docs/blog/posts/course.md`
- **Instructor Adopting Cursor Rules**（documentation）：AI-assisted coding is changing how we use version control. Many developers now use what I call "vibe coding" - coding with AI help. This creates new challenges with Git. Today I'll share how we're using Cursor rules in Instructor to solve these problems. 证据：`docs/blog/posts/cursor-rules.md`
- **Enhancing Python Functions with Instructor: A Guide to Fine-Tuning and Distillation**（documentation）：Enhancing Python Functions with Instructor: A Guide to Fine-Tuning and Distillation 证据：`docs/blog/posts/distilation-part1.md`
- **Consistent Stories with GPT-4o**（documentation）：Language Models struggle to generate consistent graphs that have a large number of nodes. Often times, this is because the graph itself is too large for the model to handle. This causes the model to generate inconsistent graphs that have invalid and disconnected nodes among other issues. 证据：`docs/blog/posts/extract-model-looks.md`
- **Why Image Metadata is useful**（documentation）：Multimodal Language Models like gpt-4o excel at processing multimodal, enabling us to extract rich, structured metadata from images. 证据：`docs/blog/posts/extracting-model-metadata.md`
- **Simple Synthetic Data Generation**（documentation）：What that people have been using instructor for is to generate synthetic data rather than extracting data itself. We can even use the J-Schemo extra fields to give specific examples to control how we generate data. 证据：`docs/blog/posts/fake-data.md`
- **Why Logfire is a perfect fit for FastAPI + Instructor**（documentation）：Why Logfire is a perfect fit for FastAPI + Instructor 证据：`docs/blog/posts/full-fastapi-visibility.md`
- **Eliminating Hallucinations with Structured Outputs using Gemini**（documentation）：Eliminating Hallucinations with Structured Outputs using Gemini 证据：`docs/blog/posts/generating-pdf-citations.md`
- **Generators and LLM Streaming**（documentation）：Latency is crucial, especially in eCommerce and newer chat applications like ChatGPT. Streaming is the solution that enables us to enhance the user experience without the need for faster response times. 证据：`docs/blog/posts/generator.md`
- **Do I Still Need Instructor with Google's New OpenAI Integration?**（documentation）：Do I Still Need Instructor with Google's New OpenAI Integration? 证据：`docs/blog/posts/google-openai-client.md`
- **Introducing structured outputs with Cerebras Inference**（documentation）：Introducing structured outputs with Cerebras Inference 证据：`docs/blog/posts/introducing-structured-outputs-with-cerebras-inference.md`
- 其余 19 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/contributing.md`, `CLAUDE.md`, `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/contributing.md`, `CLAUDE.md`, `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。

- **项目概览**：importance `high`
  - source_paths: README.md, instructor/__init__.py, mkdocs.yml
- **安装指南**：importance `high`
  - source_paths: pyproject.toml, requirements.txt, docs/installation.md
- **快速开始**：importance `high`
  - source_paths: examples/simple-extraction/user.py, docs/getting-started.md, docs/learning/getting_started/first_extraction.md
- **系统架构**：importance `high`
  - source_paths: instructor/core/client.py, instructor/processing/response.py, instructor/processing/function_calls.py, instructor/processing/validators.py, docs/architecture.md
- **提供商集成**：importance `high`
  - source_paths: instructor/providers/__init__.py, instructor/providers/anthropic/__init__.py, instructor/providers/gemini/__init__.py, instructor/providers/README.md, docs/integrations/index.md
- **响应模型**：importance `high`
  - source_paths: instructor/processing/schema.py, docs/concepts/models.md, docs/learning/getting_started/response_models.md, examples/simple-extraction/maybe_user.py
- **验证机制**：importance `high`
  - source_paths: instructor/dsl/validators.py, instructor/validation/llm_validators.py, instructor/core/retry.py, docs/concepts/validation.md, docs/concepts/reask_validation.md
- **流式输出**：importance `medium`
  - source_paths: instructor/dsl/partial.py, instructor/dsl/iterable.py, docs/concepts/partial.md, docs/concepts/iterable.md, examples/partial_streaming/run.py

## Repo Inspection Evidence / 源码检查证据

- repo_clone_verified: true
- repo_inspection_verified: true
- repo_commit: `5e8e2d57e791ed505c9637c0e215b10a5441b66a`
- inspected_files: `pyproject.toml`, `README.md`, `uv.lock`, `requirements.txt`, `docs/why.md`, `docs/architecture.md`, `docs/contributing.md`, `docs/newsletter.md`, `docs/debugging.md`, `docs/index.md`, `docs/api.md`, `docs/repository-overview.md`, `docs/getting-started.md`, `docs/api-docstring-assessment.md`, `docs/installation.md`, `docs/modes-comparison.md`, `docs/start-here.md`, `docs/faq.md`, `docs/help.md`, `docs/jobs.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：Documentation (at least Google-related) is an outdated mess.

- 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：Documentation (at least Google-related) is an outdated mess.
- Host AI rule: 来源问题仍为 open，Pack Agent 需要复核是否仍影响当前版本。
- Why it matters: 可能影响升级、迁移或版本选择。
- Evidence: community_evidence:github | cevd_b87b003416cd4308bc863f0cd66a68fd | https://github.com/567-labs/instructor/issues/2289 | 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: 来源证据：v1.13.0

- 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：v1.13.0
- 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: 可能影响升级、迁移或版本选择。
- Evidence: community_evidence:github | cevd_f3c1e8416a744b51b5691317c10bc5bf | https://github.com/567-labs/instructor/releases/tag/v1.13.0 | 来源类型 github_release 暴露的待验证使用条件。
- 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：v1.12.0

- Trigger: GitHub 社区证据显示该项目存在一个配置相关的待验证问题：v1.12.0
- 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: 可能影响升级、迁移或版本选择。
- Evidence: community_evidence:github | cevd_9b8236f58f8641c586777618a838409f | https://github.com/567-labs/instructor/releases/tag/v1.12.0 | 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 4: 来源证据：v1.14.0

- Trigger: GitHub 社区证据显示该项目存在一个配置相关的待验证问题：v1.14.0
- 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: 可能增加新用户试用和生产接入成本。
- Evidence: community_evidence:github | cevd_053ef3382ace48778d05ef006d87cead | https://github.com/567-labs/instructor/releases/tag/v1.14.0 | 来源类型 github_release 暴露的待验证使用条件。
- 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: 来源证据：v1.14.3

- Trigger: GitHub 社区证据显示该项目存在一个配置相关的待验证问题：v1.14.3
- 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_d18c003929614c7097d16742ec94cc8c | https://github.com/567-labs/instructor/releases/tag/v1.14.3 | 来源类型 github_release 暴露的待验证使用条件。
- 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: 来源证据：v1.14.4

- Trigger: GitHub 社区证据显示该项目存在一个配置相关的待验证问题：v1.14.4
- 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_76987134f9a747d685958a5e98f3b51a | https://github.com/567-labs/instructor/releases/tag/v1.14.4 | 来源类型 github_release 暴露的待验证使用条件。
- 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: 来源证据：v1.15.0

- Trigger: GitHub 社区证据显示该项目存在一个配置相关的待验证问题：v1.15.0
- 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: 可能增加新用户试用和生产接入成本。
- Evidence: community_evidence:github | cevd_430369db61e440e5a4b575e2b3618464 | https://github.com/567-labs/instructor/releases/tag/v1.15.0 | 来源类型 github_release 暴露的待验证使用条件。
- 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: 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:653589102 | https://github.com/567-labs/instructor | 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 9: 来源证据：v1.14.2

- Trigger: GitHub 社区证据显示该项目存在一个运行相关的待验证问题：v1.14.2
- 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_fd561faf78f147518bcda7a0370a9a4f | https://github.com/567-labs/instructor/releases/tag/v1.14.2 | 来源讨论提到 node 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 10: 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:653589102 | https://github.com/567-labs/instructor | 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。