# simplemem - 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 simplemem 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_0003` supported 0.86
- **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_0004` supported 0.86
- **希望把专业流程带进宿主 AI 的用户**：仓库包含 Skill 文档。 证据：`simplemem/integrations/simplemem-skill/SKILL.md`, `SKILL/simplemem-skill/SKILL.md` Claim：`clm_0005` 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：`simplemem/integrations/simplemem-skill/SKILL.md`, `SKILL/simplemem-skill/SKILL.md` 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

- `git clone https://github.com/aiming-lab/SimpleMem.git` 证据：`README.md` Claim：`clm_0006` supported 0.86
- `pip install -r requirements.txt` 证据：`README.md` Claim：`clm_0007` supported 0.86
- `pip install -e .                  # default: text + multimodal + evolver` 证据：`README.md` Claim：`clm_0008` supported 0.86
- `pip install -e ".[server]"        # + MCP / HTTP server (mcp, fastapi, ...)` 证据：`README.md` Claim：`clm_0009` supported 0.86
- `pip install -e ".[all]"           # everything, including dev tools` 证据：`README.md` Claim：`clm_0010` supported 0.86

## 继续前判断卡

- **当前建议**：需要管理员/安全审批
- **为什么**：继续前可能涉及密钥、账号、外部服务或敏感上下文，建议先经过管理员或安全审批。

### 30 秒判断

- **现在怎么做**：需要管理员/安全审批
- **最小安全下一步**：先跑 Prompt Preview；若涉及凭证或企业环境，先审批再试装
- **先别相信**：真实输出质量不能在安装前相信。
- **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：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_0003` supported 0.86
- **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_0004` supported 0.86
- **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：`simplemem/integrations/simplemem-skill/SKILL.md`, `SKILL/simplemem-skill/SKILL.md` Claim：`clm_0005` 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：`simplemem/integrations/simplemem-skill/SKILL.md`, `SKILL/simplemem-skill/SKILL.md` 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_0006` supported 0.86

### 现在还不能相信

- **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：`SKILL/simplemem-skill/SKILL.md`, `simplemem/integrations/simplemem-skill/SKILL.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。
- **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：`SKILL/simplemem-skill/SKILL.md`, `simplemem/integrations/simplemem-skill/SKILL.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：`README.md`
- **环境变量 / API Key**：项目入口文档明确出现 API key、token、secret 或账号凭证配置。 原因：如果真实安装需要凭证，应先使用测试凭证并经过权限/合规判断。 证据：`MCP/README.md`, `MCP/run.sh`, `README.md`, `docs/i18n/README.ar.md` 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**：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。）
- **不要使用真实生产凭证**：环境变量/API key 一旦进入宿主或工具链，可能产生账号和合规风险。（适用：出现 API、TOKEN、KEY、SECRET 等环境线索时。）
- **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。
- **准备撤销测试 API key 或 token**：测试凭证泄露或误用时，可以快速止损。
- **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_0011` 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_0012` 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：`simplemem/integrations/simplemem-skill/SKILL.md`, `SKILL/simplemem-skill/SKILL.md` 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：424
- 重要文件覆盖：40/424
- 证据索引条目：80
- 角色 / Skill 条目：2

### 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 simplemem 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 simplemem 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 simplemem 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 2 个角色 / Skill / 项目文档条目。

- **simplemem-skill**（skill）：Store and retrieve conversation memories across sessions. Use when asked to 'remember this', 'save conversation', 'add to memory', 'what did we discuss about...', 'query memories', or 'import chat history'. Also use proactively to preserve important dialogue context and decisions. 激活提示：当用户任务与“simplemem-skill”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：`SKILL/simplemem-skill/SKILL.md`
- **simplemem-skill**（skill）：Store and retrieve conversation memories across sessions. Use when asked to 'remember this', 'save conversation', 'add to memory', 'what did we discuss about...', 'query memories', or 'import chat history'. Also use proactively to preserve important dialogue context and decisions. 激活提示：当用户任务与“simplemem-skill”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：`simplemem/integrations/simplemem-skill/SKILL.md`

## 证据索引

- 共索引 80 条证据。

- **💡 Key Idea**（documentation）：EvolveMem: Self-Evolving Memory Architecture via AutoResearch 证据：`EvolveMem/README.md`
- **SimpleMem MCP Server**（documentation）：Production-Ready Memory Service for LLM Agents via Model Context Protocol MCP 证据：`MCP/README.md`
- **🚀 Quick Start**（documentation）：Omni-SimpleMem: Unified Multimodal Memory for Lifelong AI Agents 证据：`OmniSimpleMem/README.md`
- **Efficient Lifelong Memory for LLM Agents — Text & Multimodal**（documentation）：Efficient Lifelong Memory for LLM Agents — Text & Multimodal 证据：`README.md`
- **SimpleMem Skill**（documentation）：A self-contained Claude skill for managing persistent conversational memory using vector-based retrieval. 证据：`SKILL/README.md`
- **🧠 SimpleMem-Cross**（documentation）：Persistent Cross-Conversation Memory for LLM Agents 证据：`cross/README.md`
- **SimpleMem: Efficient Lifelong Memory for LLM Agents**（documentation）：SimpleMem: Efficient Lifelong Memory for LLM Agents 证据：`MCP/reference/README.md`
- **SimpleMem Skill**（documentation）：A self-contained Claude skill for managing persistent conversational memory using vector-based retrieval. 证据：`simplemem/integrations/README.md`
- **SimpleMem: Efficient Lifelong Memory for LLM Agents**（documentation）：SimpleMem: Efficient Lifelong Memory for LLM Agents 证据：`simplemem/integrations/reference/README.md`
- **SimpleMem Skill**（skill_instruction）：Persistent conversational memory across sessions. 证据：`SKILL/simplemem-skill/SKILL.md`
- **SimpleMem Skill**（skill_instruction）：Persistent conversational memory across sessions. 证据：`simplemem/integrations/simplemem-skill/SKILL.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`
- **License**（source_file）：Apache License Version 2.0, January 2004 http://www.apache.org/licenses/ 证据：`OmniSimpleMem/LICENSE`
- **SimpleMem Package Usage Guide**（documentation）：This guide provides comprehensive documentation for using SimpleMem as a pip-installable Python package. 证据：`docs/PACKAGE_USAGE.md`
- **SimpleMem: Text Memory**（documentation）：How the text backend turns raw dialogue into compact, retrievable memory. For the high-level summary and where this fits the unified package, see the main README Overview ../README.md -overview . 证据：`docs/text-memory.md`
- **ذاكرة مدى الحياة الفعّالة لوكلاء نماذج اللغة الكبيرة — النصوص والوسائط المتعددة**（documentation）：ذاكرة مدى الحياة الفعّالة لوكلاء نماذج اللغة الكبيرة — النصوص والوسائط المتعددة 证据：`docs/i18n/README.ar.md`
- **Effizientes lebenslanges Gedächtnis für LLM-Agenten — Text & Multimodal**（documentation）：Effizientes lebenslanges Gedächtnis für LLM-Agenten — Text & Multimodal 证据：`docs/i18n/README.de.md`
- **Memoria Vitalicia Eficiente para Agentes LLM — Texto y Multimodal**（documentation）：Memoria Vitalicia Eficiente para Agentes LLM — Texto y Multimodal 证据：`docs/i18n/README.es.md`
- **Mémoire à long terme efficace pour les agents LLM — Texte & Multimodal**（documentation）：Mémoire à long terme efficace pour les agents LLM — Texte & Multimodal 证据：`docs/i18n/README.fr.md`
- **Memoria a Lungo Termine Efficiente per Agenti LLM — Testo e Multimodale**（documentation）：Memoria a Lungo Termine Efficiente per Agenti LLM — Testo e Multimodale 证据：`docs/i18n/README.it.md`
- **LLMエージェントのための効率的な生涯記憶 — テキスト & マルチモーダル**（documentation）：LLMエージェントのための効率的な生涯記憶 — テキスト & マルチモーダル 证据：`docs/i18n/README.ja.md`
- **LLM 에이전트를 위한 효율적인 평생 기억 — 텍스트 및 멀티모달**（documentation）：LLM 에이전트를 위한 효율적인 평생 기억 — 텍스트 및 멀티모달 证据：`docs/i18n/README.ko.md`
- **Memória Vitalícia Eficiente para Agentes LLM — Texto e Multimodal**（documentation）：Memória Vitalícia Eficiente para Agentes LLM — Texto e Multimodal 证据：`docs/i18n/README.pt-br.md`
- **Эффективная долгосрочная память для агентов LLM — текст и мультимодальность**（documentation）：Эффективная долгосрочная память для агентов LLM — текст и мультимодальность 证据：`docs/i18n/README.ru.md`
- **LLM Ajanları için Verimli Ömür Boyu Bellek — Metin & Çok Modlu**（documentation）：LLM Ajanları için Verimli Ömür Boyu Bellek — Metin & Çok Modlu 证据：`docs/i18n/README.tr.md`
- **Bộ Nhớ Dài Hạn Hiệu Quả cho Các Tác Nhân LLM — Văn Bản & Đa Phương Thức**（documentation）：Bộ Nhớ Dài Hạn Hiệu Quả cho Các Tác Nhân LLM — Văn Bản & Đa Phương Thức 证据：`docs/i18n/README.vi.md`
- **面向 LLM 智能体的高效终身记忆 — 文本与多模态**（documentation）：通过语义无损压缩存储、压缩并检索长期记忆。现已支持文本、图像、音频与视频多模态。 证据：`docs/i18n/README.zh-CN.md`
- **SimpleMem Architecture**（documentation）：SimpleMem is a persistent conversational memory system that stores dialogues in a vector database LanceDB for semantic retrieval and question answering. It uses OpenRouter as a unified API gateway for both LLM and embedding services. 证据：`SKILL/simplemem-skill/references/architecture.md`
- **CLI Command Reference**（documentation）：- --table-name TABLE NAME : Use a custom table name instead of the default memory entries 证据：`SKILL/simplemem-skill/references/cli-reference.md`
- **Batch Import Guide**（documentation）：Each line is a JSON object with the following fields: 证据：`SKILL/simplemem-skill/references/import-guide.md`
- **OpenRouter Integration - Quick Reference**（documentation）：OpenRouter Integration - Quick Reference 证据：`SKILL/simplemem-skill/references/openrouter-guide.md`
- **SimpleMem Architecture**（documentation）：SimpleMem is a persistent conversational memory system that stores dialogues in a vector database LanceDB for semantic retrieval and question answering. It uses OpenRouter as a unified API gateway for both LLM and embedding services. 证据：`simplemem/integrations/simplemem-skill/references/architecture.md`
- **CLI Command Reference**（documentation）：- --table-name TABLE NAME : Use a custom table name instead of the default memory entries 证据：`simplemem/integrations/simplemem-skill/references/cli-reference.md`
- **Batch Import Guide**（documentation）：Each line is a JSON object with the following fields: 证据：`simplemem/integrations/simplemem-skill/references/import-guide.md`
- **OpenRouter Integration - Quick Reference**（documentation）：OpenRouter Integration - Quick Reference 证据：`simplemem/integrations/simplemem-skill/references/openrouter-guide.md`
- **Version control and caches**（source_file）：Version control and caches .git pycache .pyc .pyo .venv venv .env 证据：`.dockerignore`
- **SimpleMem MCP Server - Docker environment template**（source_file）：SimpleMem MCP Server - Docker environment template Copy to .env and set values: cp .env.example .env Then run: docker compose --env-file .env up -d 证据：`.env.example`
- **Python cache files**（source_file）：Python cache files pycache / .py cod $py.class .so 证据：`.gitignore`
- **SimpleMem MCP Server**（source_file）：SimpleMem MCP Server FROM python:3.11-slim 证据：`Dockerfile`
- **Requirements**（source_file）：openai =1.0.0 pyyaml =6.0 sentence-transformers =2.2.0 numpy =1.24.0 证据：`EvolveMem/requirements.txt`
- **── Sample loaders per benchmark ─────────────────────────────────────────**（source_file）：logger = logging.getLogger name ⋮---- def load key config path: str = "openai key.yaml" - dict ⋮---- api key = os.environ.get "OPENAI API KEY", "" ⋮---- def make llm call cfg: dict ⋮---- client = OpenAI base url=cfg "base url" , api key=cfg "api key" model = cfg.get "model", "gpt-4o" ⋮---- supports temperature = not model.startswith "gpt-5" and not model.startswith "o" ⋮---- def llm call messages, max tokens: int = 4096, temperature: float = 0.1 ⋮---- kwargs = dict ⋮---- r = client.chat.completions.create kwargs ⋮---- ── Sample loaders per benchmark ───────────────────────────────────────── ⋮---- def load samples args - tuple list BenchmarkSample , object ⋮---- adapter = LoCoMoAdapter path… 证据：`EvolveMem/run_benchmark.py`
- **Load data**（source_file）：API KEY = os.environ.get "OPENAI API KEY", "" API BASE = os.environ.get "OPENAI API BASE", "https://api.openai.com/v1" LLM MODEL = os.environ.get "LLM MODEL", "gpt-4o" ⋮---- client = OpenAI base url=API BASE, api key=API KEY ⋮---- def llm call messages: list dict , max tokens: int = 4096, temperature: float = 0.1 - str ⋮---- r = client.chat.completions.create content = r.choices 0 .message.content ⋮---- def load locomo sample data path: str, sample idx: int = 0 ⋮---- """Load a LoCoMo sample and return sessions, qa pairs, metadata .""" ⋮---- data = json.load f ⋮---- sample = data sample idx qa pairs = sample "qa" conversation = sample "conversation" speaker a = conversation.get "speaker a",… 证据：`EvolveMem/run_evolution.py`
- **Init**（source_file）：uv venv source .venv/bin/activate uv pip install -r requirements.txt 证据：`MCP/init.sh`
- **Register**（source_file）：API BASE = "http://localhost:8000" ⋮---- def register ⋮---- api key = "ollama-placeholder-key" ⋮---- response = requests.post ⋮---- data = response.json ⋮---- config = { 证据：`MCP/register.py`
- **SimpleMem MCP Server Dependencies**（source_file）：Web Framework fastapi =0.109.0 uvicorn standard =0.27.0 证据：`MCP/requirements.txt`
- **Run**（source_file）：def main ⋮---- parser = argparse.ArgumentParser ⋮---- args = parser.parse args 证据：`MCP/run.py`
- **Run**（source_file）：source .venv/bin/activate export JWT SECRET KEY="your-secure-random-secret-key" export ENCRYPTION KEY="your-32-byte-encryption-key!!" export LLM PROVIDER=ollama export OLLAMA BASE URL=http://localhost:11434/v1 export LLM MODEL=qwen3:4b-instruct export EMBEDDING MODEL=qwen3-embedding:4b export EMBEDDING DIMENSION=2560 env python run.py 证据：`MCP/run.sh`
- **Python**（source_file）：Python pycache / .py cod $py.class .so .egg-info/ dist/ build/ .egg 证据：`OmniSimpleMem/.gitignore`
- **Core dependencies**（source_file）：Core dependencies numpy =1.21.0 openai =1.0.0 pydantic =2.0.0 tiktoken =0.5.0 rank bm25 =0.2.2 sentence-transformers =2.2.0 证据：`OmniSimpleMem/requirements.txt`
- **Setup**（source_file）：readme path = Path file .parent / "README.md" long description = "" ⋮---- long description = readme path.read text encoding="utf-8" 证据：`OmniSimpleMem/setup.py`
- **============================================================================**（source_file）：""" Configuration file - System parameters and LLM settings 证据：`config.py.example`
- **Init**（source_file）：all = 证据：`cross/__init__.py`
- **Context injection optional orchestrator capability**（source_file）：logger = logging.getLogger name ⋮---- class StartSessionRequest BaseModel ⋮---- tenant id: str = Field content session id: str = Field project: str = Field user prompt: Optional str = Field ⋮---- class StartSessionResponse BaseModel ⋮---- memory session id: str = Field ..., description="Unique cross-session memory id" context: str = Field context tokens: int = Field ⋮---- class RecordMessageRequest BaseModel ⋮---- memory session id: str = Field content: str = Field ..., min length=1, description="Message body" role: Optional str = Field ⋮---- class RecordToolUseRequest BaseModel ⋮---- tool name: str = Field ..., min length=1, description="Name of the invoked tool" tool input: str = Field ..… 证据：`cross/api_http.py`
- **Pydantic v2+**（source_file）：logger = logging.getLogger name ⋮---- class MCPToolRegistry ⋮---- def init self, orchestrator: Any - None ⋮---- def get tool definitions self - List Dict str, Any ⋮---- async def call tool self, name: str, arguments: Dict str, Any - Dict str, Any ⋮---- handler = self. tool map.get name ⋮---- result = await await if coro ⋮---- fn = resolve method ⋮---- result = await await if coro fn memory session id=memory session id ⋮---- result = await await if coro fn query=query, top k=top k ⋮---- result = await await if coro fn user prompt=user prompt ⋮---- async def cross session stats self - Dict str, Any ⋮---- result = await await if coro fn ⋮---- def create mcp tools orchestrator: Any - MCPToolReg… 证据：`cross/api_mcp.py`
- **For compatibility with session manager.py which passes memory session id to constructor**（source_file）：class EventKind str, Enum ⋮---- message = "message" tool use = "tool use" file change = "file change" note = "note" system = "system" ⋮---- class RedactionLevel str, Enum ⋮---- none = "none" partial = "partial" full = "full" ⋮---- class SessionEvent BaseModel ⋮---- memory session id: str timestamp: datetime kind: EventKind title: Optional str = None payload json: Optional str = None redaction level: RedactionLevel = RedactionLevel.none ⋮---- class ObservationType str, Enum ⋮---- decision = "decision" bugfix = "bugfix" feature = "feature" refactor = "refactor" discovery = "discovery" change = "change" ⋮---- class CrossObservation BaseModel ⋮---- type: ObservationType title: str narrative: Op… 证据：`cross/collectors.py`
- **Consolidation**（source_file）：logger = logging.getLogger name ⋮---- class ConsolidationPolicy BaseModel ⋮---- max age days: int = 90 ⋮---- decay factor: float = 0.9 ⋮---- merge similarity threshold: float = 0.95 ⋮---- min importance: float = 0.05 ⋮---- max entries per run: int = 1000 ⋮---- class ConsolidationResult BaseModel ⋮---- decayed count: int = 0 merged count: int = 0 pruned count: int = 0 duration seconds: float = 0.0 ⋮---- def cosine similarity vec a: list float , vec b: list float - float ⋮---- dot = sum a b for a, b in zip vec a, vec b mag a = math.sqrt sum a a for a in vec a mag b = math.sqrt sum b b for b in vec b ⋮---- def parse valid from valid from: object - Optional datetime ⋮---- dt = datetime.fromisof… 证据：`cross/consolidation.py`
- **---------------------------------------------------------------------------**（source_file）：logger = logging.getLogger name ⋮---- T = TypeVar "T" ⋮---- def estimate tokens text: str - int ⋮---- def text for summary summary: SessionSummary - str ⋮---- parts: list str = ⋮---- def text for observation obs: CrossObservation - str ⋮---- """Derive a representative text string from a CrossObservation.""" detail = obs.subtitle or obs.narrative or "" ⋮---- def text for entry entry: CrossMemoryEntry - str ⋮---- """Derive a representative text string from a CrossMemoryEntry.""" ⋮---- --------------------------------------------------------------------------- Budget packing ⋮---- """Greedily pack items into a token budget. Parameters ---------- items: Ordered sequence of candidate items highe… 证据：`cross/context_injector.py`
- **Hooks**（source_file）：logger = logging.getLogger name ⋮---- class HookResult BaseModel ⋮---- context bundle: Optional ContextBundle = None finalization report: Optional FinalizationReport = None events recorded: int = Field default=0, ge=0 ⋮---- class SessionHooks ABC ⋮---- class DefaultHooks SessionHooks ⋮---- memory session id = await self. await if coro ⋮---- context bundle = await self. await if coro ⋮---- event = SessionEvent ⋮---- payload = { ⋮---- report = await self. await if coro finalization: Optional FinalizationReport = None ⋮---- finalization = report ⋮---- @staticmethod async def await if coro value: Any - Any ⋮---- class NoOpHooks SessionHooks 证据：`cross/hooks.py`
- **---------------------------------------------------------------------------**（source_file）：logger = logging.getLogger name ⋮---- DEFAULT DB PATH = "~/.simplemem-cross/cross memory.db" DEFAULT LANCEDB PATH = "~/.simplemem-cross/lancedb cross" ⋮---- class CrossMemOrchestrator ⋮---- session record = await asyncio.to thread memory session id = session record.memory session id ⋮---- context bundle = await asyncio.to thread rendered context = self. render context safe context bundle ⋮---- report = await asyncio.to thread ⋮---- results = self.vector store.semantic search ⋮---- bundle = self. build context safe user prompt ⋮---- def get stats self - Dict str, Any ⋮---- stats: Dict str, Any = self.sqlite storage.get stats ⋮---- def close self - None ⋮---- async def aenter self - CrossMemO… 证据：`cross/orchestrator.py`
- **Session Manager**（source_file）：class CollectedEvent ⋮---- slots = "kind", "title", "payload", "timestamp" ⋮---- class EventCollector ⋮---- def init self, memory session id: str - None ⋮---- event = CollectedEvent kind=kind, title=title, payload=payload ⋮---- def flush self - List CollectedEvent ⋮---- events = list self. events ⋮---- @property def event count self - int ⋮---- class ObservationExtractor ⋮---- KIND TO OBS TYPE: Dict str, ObservationType = { ⋮---- observations: List CrossObservation = ⋮---- title = getattr event, "title", None ⋮---- kind value = obs type = self. KIND TO OBS TYPE.get payload = getattr event, "payload", None narrative: Optional str = None ⋮---- narrative = payload.get "content" or payload.get… 证据：`cross/session_manager.py`
- 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：`EvolveMem/README.md`, `MCP/README.md`, `OmniSimpleMem/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：`EvolveMem/README.md`, `MCP/README.md`, `OmniSimpleMem/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。

- **项目概述与系统架构**：importance `high`
  - source_paths: README.md, simplemem/__init__.py, simplemem/router.py, simplemem_router.py, simplemem/text/system.py
- **核心 API、记忆管线与检索机制**：importance `high`
  - source_paths: simplemem/core/memory_builder.py, simplemem/core/hybrid_retriever.py, simplemem/core/answer_generator.py, simplemem/core/models/memory_entry.py, simplemem/multimodal/orchestrator.py
- **MCP 服务器、Docker 部署与后端集成**：importance `high`
  - source_paths: MCP/README.md, MCP/run.py, MCP/server/http_server.py, MCP/server/mcp_handler.py, MCP/server/auth/token_manager.py
- **常见问题、复现指南、安全与社区议题**：importance `high`
  - source_paths: test_locomo10.py, test_ref/load_dataset.py, test_ref/test_advanced.py, test_ref/utils.py, EvolveMem/run_benchmark.py

## Repo Inspection Evidence / 源码检查证据

- repo_clone_verified: true
- repo_inspection_verified: true
- repo_commit: `60a48e83a7fef10d386e1f438589047d3a4257bc`
- inspected_files: `Dockerfile`, `README.md`, `docker-compose.yml`, `requirements.txt`, `docs/PACKAGE_USAGE.md`, `docs/i18n/README.ar.md`, `docs/i18n/README.de.md`, `docs/i18n/README.es.md`, `docs/i18n/README.fr.md`, `docs/i18n/README.it.md`, `docs/i18n/README.ja.md`, `docs/i18n/README.ko.md`, `docs/i18n/README.pt-br.md`, `docs/i18n/README.ru.md`, `docs/i18n/README.tr.md`, `docs/i18n/README.vi.md`, `docs/i18n/README.zh-CN.md`, `docs/text-memory.md`, `examples/quickstart.py`

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：How can I reproduce the locomo results on GPT-4o

- Trigger: GitHub 社区证据显示该项目存在一个安全/权限相关的待验证问题：How can I reproduce the locomo results on GPT-4o
- Why it matters: 可能影响授权、密钥配置或安全边界。
- Evidence: community_evidence:github | https://github.com/aiming-lab/SimpleMem/issues/58 | 来源类型 github_issue 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 2: 来源证据：Question about the evaluation protocol used for MemGallery

- 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：Question about the evaluation protocol used for MemGallery
- 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/aiming-lab/SimpleMem/issues/64 | 来源类型 github_issue 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：Sponsored partnership

- 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：Sponsored partnership
- 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/aiming-lab/SimpleMem/issues/63 | 来源类型 github_issue 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 4: 可能修改宿主 AI 配置

- Trigger: 项目面向 Claude/Cursor/Codex/Gemini/OpenCode 等宿主，或安装命令涉及用户配置目录。
- Host AI rule: 列出会写入的配置文件、目录和卸载/回滚步骤。
- Why it matters: 安装可能改变本机 AI 工具行为，用户需要知道写入位置和回滚方法。
- Evidence: capability.host_targets | https://github.com/aiming-lab/SimpleMem | host_targets=mcp_host, claude, cursor, chatgpt
- 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: 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/aiming-lab/SimpleMem | 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 6: 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/aiming-lab/SimpleMem | 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/aiming-lab/SimpleMem | 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 8: 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/aiming-lab/SimpleMem | 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 9: 来源证据：Missing OmniSimpleMem package config module: omni_memory.core.config

- Trigger: GitHub 社区证据显示该项目存在一个安全/权限相关的待验证问题：Missing OmniSimpleMem package config module: omni_memory.core.config
- 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: 可能影响授权、密钥配置或安全边界。
- Evidence: community_evidence:github | https://github.com/aiming-lab/SimpleMem/issues/60 | 来源类型 github_issue 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 10: 来源证据：Request for more evaluation details for Table 3 on LoCoMo

- Trigger: GitHub 社区证据显示该项目存在一个安全/权限相关的待验证问题：Request for more evaluation details for Table 3 on LoCoMo
- 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: 可能影响授权、密钥配置或安全边界。
- Evidence: community_evidence:github | https://github.com/aiming-lab/SimpleMem/issues/47 | 来源类型 github_issue 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。
