YG-Soft/X-Financial
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

feat: deliver agent foundation day 1

Browse Source
This commit is contained in:
caoxiaozhu
2026-05-11 03:51:24 +00:00
parent f738b6cdd4
commit b2beeaa136
54 changed files with 6747 additions and 1724 deletions
Download Patch File Download Diff File Expand all files Collapse all files

137
document/development/agent_week_plan_html/day-1.html Normal file
View File

@@ -0,0 +1,137 @@
<!DOCTYPE html>
<html lang="zh-CN">
<head>
<meta charset="UTF-8">
<meta name="viewport" content="width=device-width, initial-scale=1.0">
<title>Day 1 - 基础模型与工程骨架</title>
<link rel="stylesheet" href="./styles.css">
</head>
<body>
<div class="shell">
<div class="topbar">
<a class="brand" href="./index.html"><span class="brand-mark">D1</span><span>Day 1 View</span></a>
<div class="quick-links">
<a class="pill" href="./index.html">返回总览</a>
<a class="pill" href="../agent%20week%20plan/day_1_foundation_models.md">周计划原文</a>
<a class="pill" href="../agent%20plan/weekly_execution_details/day_1_foundation_models.md">执行细则原文</a>
</div>
</div>
<div class="day-nav">
<a class="pill active" href="./day-1.html">Day 1</a>
<a class="pill" href="./day-2.html">Day 2</a>
<a class="pill" href="./day-3.html">Day 3</a>
<a class="pill" href="./day-4.html">Day 4</a>
<a class="pill" href="./day-5.html">Day 5</a>
<a class="pill" href="./day-6.html">Day 6</a>
<a class="pill" href="./day-7.html">Day 7</a>
</div>
<section class="hero">
<div class="hero-badge">Foundation Completed</div>
<h1>Day 1 基础模型与工程骨架</h1>
<p>这一天的任务不是做炫目的业务能力，而是把后面 6 天要反复依赖的模型、版本、审核、run trace、审计日志和最小业务数据源一次定稳。Day 1 做虚了，Day 4 到 Day 6 会全部返工。</p>
<div class="hero-meta">
<div class="meta-card"><div class="meta-label">当前状态</div><div class="meta-value">已完成（2026-05-11），可直接进入 Day 2 联调。</div></div>
<div class="meta-card"><div class="meta-label">上游依赖</div><div class="meta-value">无，Day 1 是全周底座。</div></div>
<div class="meta-card"><div class="meta-label">下游交接</div><div class="meta-value">Day 2 资产 API，Day 3 解析日志，Day 4 run trace，Day 5/6 业务数据查询。</div></div>
<div class="meta-card"><div class="meta-label">当天关键</div><div class="meta-value">先确定统一模型，再接 API 骨架和种子数据。</div></div>
</div>
</section>
<div class="section-kicker">Three-Layer Mapping</div>
<h2 class="section-title">三层文档映射</h2>
<div class="grid three">
<section class="card tone-warm">
<h3>路线图</h3>
<p>周计划里定义这一天要完成“工程地基”，强调只做稳定模型、API 骨架、种子数据、基础审计和可运行验证。</p>
<div class="card-links"><a class="link-chip" href="../agent%20week%20plan/day_1_foundation_models.md">day_1_foundation_models.md</a></div>
</section>
<section class="card tone-teal">
<h3>执行细则</h3>
<p>执行层把 Day 1 拆成命名边界、最小财务业务数据模型、Agent 资产模型、版本、审核、Run、ToolCall、SemanticParseLog、AuditLog、Schema、API、服务层。</p>
<div class="card-links"><a class="link-chip" href="../agent%20plan/weekly_execution_details/day_1_foundation_models.md">weekly_execution_details/day_1</a></div>
</section>
<section class="card tone-olive">
<h3>架构依据</h3>
<p>主要受总体架构、语义本体、数据契约、能力注册、权限确认、可观测性和财务标准模型约束。</p>
<div class="card-links">
<a class="link-chip" href="../agent%20plan/01_overall_architecture.md">01</a>
<a class="link-chip" href="../agent%20plan/02_semantic_ontology.md">02</a>
<a class="link-chip" href="../agent%20plan/06_data_contracts_and_governance.md">06</a>
<a class="link-chip" href="../agent%20plan/07_capability_registry.md">07</a>
<a class="link-chip" href="../agent%20plan/08_permission_confirmation.md">08</a>
<a class="link-chip" href="../agent%20plan/09_observability_and_trace.md">09</a>
<a class="link-chip" href="../agent%20plan/14_financial_document_canonical_model.md">14</a>
</div>
</section>
</div>
<div class="section-kicker">Build Order</div>
<h2 class="section-title">推荐开发顺序</h2>
<div class="timeline">
<div class="timeline-step"><strong>Step 1</strong>先确认后端目录、ORM、迁移方式、测试目录和不该碰的文件。</div>
<div class="timeline-step"><strong>Step 2</strong>统一命名：资产类型、状态、审核状态、Agent、权限级别。</div>
<div class="timeline-step"><strong>Step 3</strong>补最小财务业务数据模型：<code>expense_claims</code>、<code>accounts_receivable</code>、<code>accounts_payable</code>。</div>
<div class="timeline-step"><strong>Step 4</strong>完成 AgentAsset、Version、Review、Run、ToolCall、ParseLog、AuditLog。</div>
<div class="timeline-step"><strong>Step 5</strong>把 Schema、API 骨架、服务层、种子数据接起来。</div>
</div>
<div class="section-kicker">Must Deliver</div>
<h2 class="section-title">今天必须产出的东西</h2>
<div class="grid two">
<section class="card">
<h3>平台底座表</h3>
<ul class="list">
<li><code>AgentAsset</code>、<code>AgentAssetVersion</code>、<code>AgentAssetReview</code></li>
<li><code>AgentRun</code>、<code>AgentToolCall</code>、<code>SemanticParseLog</code></li>
<li><code>AuditLog</code></li>
</ul>
</section>
<section class="card">
<h3>最小业务数据来源</h3>
<ul class="list">
<li>报销至少有时间、地点、理由、金额、员工、部门、状态。</li>
<li>应收至少有客户、金额、未收金额、到期日、账龄、状态。</li>
<li>应付至少有供应商、金额、未付金额、到期日、账龄、状态。</li>
</ul>
</section>
<section class="card">
<h3>API 骨架</h3>
<ul class="list">
<li>资产列表 / 详情 / 版本 / 审核 / 上线。</li>
<li>运行日志与审计日志查询。</li>
<li>返回真实数据库结果，不用前端硬编码收尾。</li>
</ul>
</section>
<section class="card">
<h3>统一服务边界</h3>
<ul class="list">
<li>上线拦截逻辑在服务层，不堆到路由。</li>
<li>所有写操作要留审计接口。</li>
<li>任何 Agent 执行记录都必须生成 <code>run_id</code>。</li>
</ul>
</section>
</div>
<div class="section-kicker">Acceptance Snapshot</div>
<h2 class="section-title">验收快照</h2>
<div class="table-like">
<div class="row"><div class="row-label">资产模型</div><div class="row-value">已落地 3 条规则、2 条技能、2 条 MCP、3 条任务，并可通过资产接口返回。</div></div>
<div class="row"><div class="row-label">版本与审核</div><div class="row-value">三条规则都具备版本历史；同一资产版本号不可重复，未审核规则不能上线。</div></div>
<div class="row"><div class="row-label">运行与错误</div><div class="row-value">`GET /api/v1/agent-runs` 可返回 3 条运行日志，任意新建 Run 自动生成 <code>run_id</code>。</div></div>
<div class="row"><div class="row-label">最小业务表</div><div class="row-value">报销、应收、应付种子数据已就位，后续查询和风险巡检都有明确数据来源。</div></div>
</div>
<div class="section-kicker">Common Misses</div>
<h2 class="section-title">这一天最容易漏掉的点</h2>
<ul class="list">
<li>只建 Agent 表，不建最小财务业务表，导致 User Agent 和 Hermes 后面无数据可查。</li>
<li>把审核拦截塞在 API 路由里，后面很难复用到 Orchestrator 和别的入口。</li>
<li>没有统一 <code>run_id</code> 和审计接口，Day 4 到 Day 7 的 Trace 会断链。</li>
</ul>
<div class="footer">Day 1 的判断标准很简单：不是“代码写了多少”，而是“后面 6 天会不会反复回头补地基”。</div>
</div>
</body>
</html>

132
document/development/agent_week_plan_html/day-2.html Normal file
View File

@@ -0,0 +1,132 @@
<!DOCTYPE html>
<html lang="zh-CN">
<head>
<meta charset="UTF-8">
<meta name="viewport" content="width=device-width, initial-scale=1.0">
<title>Day 2 - 任务规则中心联调</title>
<link rel="stylesheet" href="./styles.css">
</head>
<body>
<div class="shell">
<div class="topbar">
<a class="brand" href="./index.html"><span class="brand-mark">D2</span><span>Day 2 View</span></a>
<div class="quick-links">
<a class="pill" href="./index.html">返回总览</a>
<a class="pill" href="../agent%20week%20plan/day_2_rule_center_integration.md">周计划原文</a>
<a class="pill" href="../agent%20plan/weekly_execution_details/day_2_rule_center_integration.md">执行细则原文</a>
</div>
</div>
<div class="day-nav">
<a class="pill" href="./day-1.html">Day 1</a>
<a class="pill active" href="./day-2.html">Day 2</a>
<a class="pill" href="./day-3.html">Day 3</a>
<a class="pill" href="./day-4.html">Day 4</a>
<a class="pill" href="./day-5.html">Day 5</a>
<a class="pill" href="./day-6.html">Day 6</a>
<a class="pill" href="./day-7.html">Day 7</a>
</div>
<section class="hero">
<div class="hero-badge">Integration</div>
<h1>Day 2 任务规则中心联调</h1>
<p>Day 2 的核心不是“把页面做漂亮”，而是让规则、技能、MCP、任务这四类资产第一次脱离本地假数据，真正连到 Day 1 的数据库和 API。最关键的能力是 Markdown、版本、审核和上线约束闭环。</p>
<div class="hero-meta">
<div class="meta-card"><div class="meta-label">上游依赖</div><div class="meta-value">Day 1 的资产模型、版本模型、审核模型、资产 API。</div></div>
<div class="meta-card"><div class="meta-label">下游交接</div><div class="meta-value">Day 3 要复用资产数据，Day 4 要查询 active 技能 / MCP / 任务。</div></div>
<div class="meta-card"><div class="meta-label">当天关键</div><div class="meta-value">前端联调不是硬编码演示，而是可对接真实后端。</div></div>
</div>
</section>
<div class="section-kicker">Three-Layer Mapping</div>
<h2 class="section-title">三层文档映射</h2>
<div class="grid three">
<section class="card tone-warm">
<h3>路线图</h3>
<p>周计划要求把任务规则中心从静态 UI 升级到真实数据对接，覆盖规则、技能、MCP、任务四类资产。</p>
<div class="card-links"><a class="link-chip" href="../agent%20week%20plan/day_2_rule_center_integration.md">day_2_rule_center_integration.md</a></div>
</section>
<section class="card tone-teal">
<h3>执行细则</h3>
<p>执行层拆成 API Client、四类列表、规则详情、Markdown 编辑、版本卡片、审核与上线、技能详情、MCP 详情、任务详情、前端质量和当天验收。</p>
<div class="card-links"><a class="link-chip" href="../agent%20plan/weekly_execution_details/day_2_rule_center_integration.md">weekly_execution_details/day_2</a></div>
</section>
<section class="card tone-olive">
<h3>架构依据</h3>
<p>这一天主要受能力注册、规则形成生命周期和数据治理约束，重点在四类资产的统一展示方式和规则上线前审核拦截。</p>
<div class="card-links">
<a class="link-chip" href="../agent%20plan/07_capability_registry.md">07</a>
<a class="link-chip" href="../agent%20plan/13_rule_formation_lifecycle.md">13</a>
<a class="link-chip" href="../agent%20plan/06_data_contracts_and_governance.md">06</a>
</div>
</section>
</div>
<div class="section-kicker">Build Order</div>
<h2 class="section-title">推荐开发顺序</h2>
<div class="timeline">
<div class="timeline-step"><strong>Step 1</strong>先补 API Client：列表、详情、版本、保存、审核、上线、运行日志。</div>
<div class="timeline-step"><strong>Step 2</strong>把四个页签的真实数据接起来，覆盖筛选、搜索、状态、空态和加载态。</div>
<div class="timeline-step"><strong>Step 3</strong>把规则详情的 Hero 区、Markdown 编辑器、版本卡片和审核信息拉通。</div>
<div class="timeline-step"><strong>Step 4</strong>补技能 / MCP / 任务的差异化详情，不复用规则编辑器。</div>
<div class="timeline-step"><strong>Step 5</strong>最后收 UI 细节、错误态、禁用态、确认弹窗和构建验证。</div>
</div>
<div class="section-kicker">Must Deliver</div>
<h2 class="section-title">今天必须产出的东西</h2>
<div class="grid two">
<section class="card">
<h3>规则中心四页签</h3>
<ul class="list">
<li>规则、技能、MCP、任务都能切换。</li>
<li>每个页签都来自真实接口，不再只读本地常量。</li>
<li>搜索和状态筛选同时生效。</li>
</ul>
</section>
<section class="card">
<h3>规则详情闭环</h3>
<ul class="list">
<li>能读取当前 Markdown。</li>
<li>能保存并刷新版本列表。</li>
<li>能展示审核者、审核状态、上线条件。</li>
</ul>
</section>
<section class="card">
<h3>版本与上线约束</h3>
<ul class="list">
<li>最近 5 个版本可见。</li>
<li>切换旧版本必须弹确认框。</li>
<li>未审核规则不能上线，拒绝原因要可见。</li>
</ul>
</section>
<section class="card">
<h3>详情差异化</h3>
<ul class="list">
<li>技能详情展示输入输出与依赖。</li>
<li>MCP 详情展示服务地址、鉴权、降级策略。</li>
<li>任务详情展示 cron、执行 Agent、最近执行结果。</li>
</ul>
</section>
</div>
<div class="section-kicker">Acceptance Snapshot</div>
<h2 class="section-title">验收快照</h2>
<div class="table-like">
<div class="row"><div class="row-label">真实数据</div><div class="row-value">四个页签都能用真实后端数据渲染，后端不可用时有明确错误提示。</div></div>
<div class="row"><div class="row-label">规则编辑</div><div class="row-value">Markdown 保存后刷新页面仍在，保存失败不丢输入。</div></div>
<div class="row"><div class="row-label">版本卡片</div><div class="row-value">最近 5 个版本可切换，当前版本标识清楚但不造成布局位移。</div></div>
<div class="row"><div class="row-label">审核上线</div><div class="row-value"><code>pending</code> / <code>rejected</code> 规则都无法上线，<code>approved</code> 才能放行。</div></div>
</div>
<div class="section-kicker">Common Misses</div>
<h2 class="section-title">这一天最容易漏掉的点</h2>
<ul class="list">
<li>只把规则页签接成真实数据，技能、MCP、任务仍然靠假数据撑场面。</li>
<li>只做版本列表展示，不做确认弹窗和拒绝风险提示。</li>
<li>把任务写成“定时任务”暴露给用户，违背文档里 UI 名称统一成“任务”的约束。</li>
</ul>
<div class="footer">Day 2 的完成标准不是“页面能打开”，而是“规则中心第一次成为真实的资产入口”。</div>
</div>
</body>
</html>

132
document/development/agent_week_plan_html/day-3.html Normal file
View File

@@ -0,0 +1,132 @@
<!DOCTYPE html>
<html lang="zh-CN">
<head>
<meta charset="UTF-8">
<meta name="viewport" content="width=device-width, initial-scale=1.0">
<title>Day 3 - 语义本体 MVP</title>
<link rel="stylesheet" href="./styles.css">
</head>
<body>
<div class="shell">
<div class="topbar">
<a class="brand" href="./index.html"><span class="brand-mark">D3</span><span>Day 3 View</span></a>
<div class="quick-links">
<a class="pill" href="./index.html">返回总览</a>
<a class="pill" href="../agent%20week%20plan/day_3_semantic_ontology_mvp.md">周计划原文</a>
<a class="pill" href="../agent%20plan/weekly_execution_details/day_3_semantic_ontology_mvp.md">执行细则原文</a>
</div>
</div>
<div class="day-nav">
<a class="pill" href="./day-1.html">Day 1</a>
<a class="pill" href="./day-2.html">Day 2</a>
<a class="pill active" href="./day-3.html">Day 3</a>
<a class="pill" href="./day-4.html">Day 4</a>
<a class="pill" href="./day-5.html">Day 5</a>
<a class="pill" href="./day-6.html">Day 6</a>
<a class="pill" href="./day-7.html">Day 7</a>
</div>
<section class="hero">
<div class="hero-badge">Ontology</div>
<h1>Day 3 语义本体 MVP</h1>
<p>这一天把自然语言问题统一切成 8 个核心字段。Day 3 不是追求大模型多聪明，而是先让结构稳定、可落日志、可被 Orchestrator、User Agent 和 Hermes 共用。</p>
<div class="hero-meta">
<div class="meta-card"><div class="meta-label">上游依赖</div><div class="meta-value">Day 1 的 <code>SemanticParseLog</code> / <code>AgentRun</code>，Day 2 的资产 API。</div></div>
<div class="meta-card"><div class="meta-label">下游交接</div><div class="meta-value">Day 4 路由、Day 5 查询解释、Day 6 风险巡检都直接消费这 8 字段。</div></div>
<div class="meta-card"><div class="meta-label">当天关键</div><div class="meta-value">名字统一、类型统一、日志统一、低置信度有澄清问题。</div></div>
</div>
</section>
<div class="section-kicker">Three-Layer Mapping</div>
<h2 class="section-title">三层文档映射</h2>
<div class="grid three">
<section class="card tone-warm">
<h3>路线图</h3>
<p>周计划要求建立用户问题的统一语义解析层，覆盖场景、意图、对象、时间、指标、约束、风险、权限 8 字段。</p>
<div class="card-links"><a class="link-chip" href="../agent%20week%20plan/day_3_semantic_ontology_mvp.md">day_3_semantic_ontology_mvp.md</a></div>
</section>
<section class="card tone-teal">
<h3>执行细则</h3>
<p>执行层拆成 8 字段定义、字段枚举、Schema、解析服务、对象提取、时间范围、指标约束、风险权限、API、前端调试入口和评测集。</p>
<div class="card-links"><a class="link-chip" href="../agent%20plan/weekly_execution_details/day_3_semantic_ontology_mvp.md">weekly_execution_details/day_3</a></div>
</section>
<section class="card tone-olive">
<h3>架构依据</h3>
<p>主要受语义本体、财务标准模型和数据治理约束。应收、应付、报销的对象语义必须能回到最小业务表和标准对象。</p>
<div class="card-links">
<a class="link-chip" href="../agent%20plan/02_semantic_ontology.md">02</a>
<a class="link-chip" href="../agent%20plan/14_financial_document_canonical_model.md">14</a>
<a class="link-chip" href="../agent%20plan/06_data_contracts_and_governance.md">06</a>
</div>
</section>
</div>
<div class="section-kicker">Build Order</div>
<h2 class="section-title">推荐开发顺序</h2>
<div class="timeline">
<div class="timeline-step"><strong>Step 1</strong>先固定 8 个字段名字、类型、默认值和示例。</div>
<div class="timeline-step"><strong>Step 2</strong>把 <code>scenario</code>、<code>intent</code>、<code>permission.level</code> 的枚举定死。</div>
<div class="timeline-step"><strong>Step 3</strong>做请求/响应 Schema，再写解析服务。</div>
<div class="timeline-step"><strong>Step 4</strong>补对象提取、时间范围、指标约束、风险和权限映射。</div>
<div class="timeline-step"><strong>Step 5</strong>接 API、日志、调试入口和最小评测集。</div>
</div>
<div class="section-kicker">Must Deliver</div>
<h2 class="section-title">今天必须产出的东西</h2>
<div class="grid two">
<section class="card">
<h3>8 字段统一结构</h3>
<ul class="list">
<li><code>scenario</code>、<code>intent</code>、<code>entities</code>、<code>time_range</code></li>
<li><code>metrics</code>、<code>constraints</code>、<code>risk_flags</code>、<code>permission</code></li>
<li>附带 <code>confidence</code>、<code>clarification_required</code>、<code>run_id</code></li>
</ul>
</section>
<section class="card">
<h3>规则解析优先版</h3>
<ul class="list">
<li>先用关键词和规则解析打底。</li>
<li>报销 / 应收 / 应付 / 知识 / unknown 场景都能落到结构。</li>
<li>越权动作能识别为 <code>approval_required</code> 或 <code>forbidden</code>。</li>
</ul>
</section>
<section class="card">
<h3>日志和调试入口</h3>
<ul class="list">
<li>每次解析都要落 <code>SemanticParseLog</code>。</li>
<li>前端可直接输入一句话看 8 字段结果。</li>
<li>低置信度问题必须给澄清问题。</li>
</ul>
</section>
<section class="card">
<h3>最小评测集</h3>
<ul class="list">
<li>至少覆盖报销、应收、应付、知识、越权动作。</li>
<li>每条样例要写期望 <code>scenario</code>、<code>intent</code> 和权限级别。</li>
<li>当天目标是可评测，而不是追求完美准确率。</li>
</ul>
</section>
</div>
<div class="section-kicker">Acceptance Snapshot</div>
<h2 class="section-title">验收快照</h2>
<div class="table-like">
<div class="row"><div class="row-label">语义结构</div><div class="row-value">8 字段在 Schema、服务层、日志里名字完全一致。</div></div>
<div class="row"><div class="row-label">关键识别</div><div class="row-value">“本周报销超标风险”“客户 A 本月应收”“供应商 B 明天要付多少钱”都能落到正确场景和意图。</div></div>
<div class="row"><div class="row-label">权限结果</div><div class="row-value">“帮我直接付款”不能被识别成可直接执行动作。</div></div>
<div class="row"><div class="row-label">日志与前端</div><div class="row-value">连续调用多次都能在日志中查到，并能通过调试入口观察结果。</div></div>
</div>
<div class="section-kicker">Common Misses</div>
<h2 class="section-title">这一天最容易漏掉的点</h2>
<ul class="list">
<li>字段结构和日志结构各写一套名字，后面 Trace 很难串。</li>
<li>只做 <code>scenario</code> 和 <code>intent</code>，不做 <code>permission</code>，Day 4 会直接失去拦截依据。</li>
<li>只在服务里返回结果，不把解析过程落库或落日志，后续无法复盘误判样例。</li>
</ul>
<div class="footer">Day 3 的价值在于把“语义理解”从模糊文本变成稳定协议。后面所有智能能力都站在这层协议上。</div>
</div>
</body>
</html>

133
document/development/agent_week_plan_html/day-4.html Normal file
View File

@@ -0,0 +1,133 @@
<!DOCTYPE html>
<html lang="zh-CN">
<head>
<meta charset="UTF-8">
<meta name="viewport" content="width=device-width, initial-scale=1.0">
<title>Day 4 - Orchestrator 运行时</title>
<link rel="stylesheet" href="./styles.css">
</head>
<body>
<div class="shell">
<div class="topbar">
<a class="brand" href="./index.html"><span class="brand-mark">D4</span><span>Day 4 View</span></a>
<div class="quick-links">
<a class="pill" href="./index.html">返回总览</a>
<a class="pill" href="../agent%20week%20plan/day_4_orchestrator_runtime.md">周计划原文</a>
<a class="pill" href="../agent%20plan/weekly_execution_details/day_4_orchestrator_runtime.md">执行细则原文</a>
</div>
</div>
<div class="day-nav">
<a class="pill" href="./day-1.html">Day 1</a>
<a class="pill" href="./day-2.html">Day 2</a>
<a class="pill" href="./day-3.html">Day 3</a>
<a class="pill active" href="./day-4.html">Day 4</a>
<a class="pill" href="./day-5.html">Day 5</a>
<a class="pill" href="./day-6.html">Day 6</a>
<a class="pill" href="./day-7.html">Day 7</a>
</div>
<section class="hero">
<div class="hero-badge">Runtime</div>
<h1>Day 4 Orchestrator 运行时</h1>
<p>Day 4 把整个系统第一次串成“能跑的链”。用户消息和定时任务都先走 Orchestrator，由它创建 run、调用语义解析、做权限判断、选择 Agent、记录 ToolCall 和 Trace，然后再给下游执行。</p>
<div class="hero-meta">
<div class="meta-card"><div class="meta-label">上游依赖</div><div class="meta-value">Day 3 的语义解析结果，Day 1 的 Run / ToolCall，Day 2 的 active 资产。</div></div>
<div class="meta-card"><div class="meta-label">下游交接</div><div class="meta-value">Day 5 User Agent 和 Day 6 Hermes 都通过它被调度。</div></div>
<div class="meta-card"><div class="meta-label">当天关键</div><div class="meta-value">权限拦截和 Trace 必须在 Orchestrator 层，而不是散落在各 Agent。</div></div>
</div>
</section>
<div class="section-kicker">Three-Layer Mapping</div>
<h2 class="section-title">三层文档映射</h2>
<div class="grid three">
<section class="card tone-warm">
<h3>路线图</h3>
<p>周计划要求建立统一调度层，让用户请求和系统任务都先进入 Orchestrator，再根据语义、权限、能力注册路由到 User Agent、Hermes、MCP 或规则引擎。</p>
<div class="card-links"><a class="link-chip" href="../agent%20week%20plan/day_4_orchestrator_runtime.md">day_4_orchestrator_runtime.md</a></div>
</section>
<section class="card tone-teal">
<h3>执行细则</h3>
<p>执行层拆成输入输出、Orchestrator 服务、路由规则、权限判断、能力查询、工具调用封装、API、最小 Trace 查看和测试。</p>
<div class="card-links"><a class="link-chip" href="../agent%20plan/weekly_execution_details/day_4_orchestrator_runtime.md">weekly_execution_details/day_4</a></div>
</section>
<section class="card tone-olive">
<h3>架构依据</h3>
<p>主要受运行时流程、能力注册、权限确认和可观测性约束。Day 4 的输出要能直接给前端展示，并支持 Day 5/6 的占位实现接入。</p>
<div class="card-links">
<a class="link-chip" href="../agent%20plan/04_orchestrator_and_runtime_flow.md">04</a>
<a class="link-chip" href="../agent%20plan/07_capability_registry.md">07</a>
<a class="link-chip" href="../agent%20plan/08_permission_confirmation.md">08</a>
<a class="link-chip" href="../agent%20plan/09_observability_and_trace.md">09</a>
</div>
</section>
</div>
<div class="section-kicker">Build Order</div>
<h2 class="section-title">推荐开发顺序</h2>
<div class="timeline">
<div class="timeline-step"><strong>Step 1</strong>先定 <code>OrchestratorRequest</code> 和 <code>OrchestratorResponse</code>。</div>
<div class="timeline-step"><strong>Step 2</strong>写 <code>run(request)</code> 主流程：创建 Run、解析语义、判权限、选 Agent、更新状态。</div>
<div class="timeline-step"><strong>Step 3</strong>把用户入口 / 任务入口的路由规则固化下来。</div>
<div class="timeline-step"><strong>Step 4</strong>封装工具调用记录和降级策略。</div>
<div class="timeline-step"><strong>Step 5</strong>暴露 API 和最小 Trace 页面或接口。</div>
</div>
<div class="section-kicker">Must Deliver</div>
<h2 class="section-title">今天必须产出的东西</h2>
<div class="grid two">
<section class="card">
<h3>统一入口</h3>
<ul class="list">
<li><code>source=user_message</code> 与 <code>source=schedule</code> 都能进同一入口。</li>
<li>请求返回 <code>run_id</code>、<code>selected_agent</code>、<code>route_reason</code>、<code>permission_level</code>。</li>
<li>返回结果要能被前端直接展示。</li>
</ul>
</section>
<section class="card">
<h3>权限与路由</h3>
<ul class="list">
<li>查询类走 User Agent，定时风险类走 Hermes。</li>
<li><code>approval_required</code> 只返回确认，不直接执行。</li>
<li><code>forbidden</code> 直接阻断，不调下游 Agent。</li>
</ul>
</section>
<section class="card">
<h3>能力与工具调用</h3>
<ul class="list">
<li>只查询 active 技能 / MCP / 任务。</li>
<li>禁用能力不允许被调用。</li>
<li>每次工具调用都能落 <code>AgentToolCall</code>。</li>
</ul>
</section>
<section class="card">
<h3>Trace 与降级</h3>
<ul class="list">
<li>Trace 能串起语义解析、路由、工具调用和最终结果。</li>
<li>外部 MCP 失败要返回降级说明，不让前端拿到不可读错误。</li>
<li>异常都要写进 <code>AgentRun.error_message</code>。</li>
</ul>
</section>
</div>
<div class="section-kicker">Acceptance Snapshot</div>
<h2 class="section-title">验收快照</h2>
<div class="table-like">
<div class="row"><div class="row-label">路由结果</div><div class="row-value">同一句风险检查，在用户入口和任务入口会有不同路由结果。</div></div>
<div class="row"><div class="row-label">权限边界</div><div class="row-value">“直接上线规则”和“直接付款”都不会被自动执行。</div></div>
<div class="row"><div class="row-label">日志完整度</div><div class="row-value">每次运行至少有一条 <code>AgentRun</code>，工具调用有 0 到多条 <code>AgentToolCall</code>。</div></div>
<div class="row"><div class="row-label">可观察性</div><div class="row-value">前端或 curl 可以完整看到一次运行链路，不需要直接查数据库猜过程。</div></div>
</div>
<div class="section-kicker">Common Misses</div>
<h2 class="section-title">这一天最容易漏掉的点</h2>
<ul class="list">
<li>把权限判断放到 User Agent / Hermes 内部，导致系统没有统一边界。</li>
<li>只记录成功 ToolCall，不记录失败 ToolCall，后面降级和排错会缺证据。</li>
<li>路由能跑，但没有统一 Trace 输出，Day 7 演示时会非常难讲清链路。</li>
</ul>
<div class="footer">Day 4 的价值是把系统从“有很多零件”变成“有一条统一运行链”。</div>
</div>
</body>
</html>

133
document/development/agent_week_plan_html/day-5.html Normal file
View File

@@ -0,0 +1,133 @@
<!DOCTYPE html>
<html lang="zh-CN">
<head>
<meta charset="UTF-8">
<meta name="viewport" content="width=device-width, initial-scale=1.0">
<title>Day 5 - User Agent MVP</title>
<link rel="stylesheet" href="./styles.css">
</head>
<body>
<div class="shell">
<div class="topbar">
<a class="brand" href="./index.html"><span class="brand-mark">D5</span><span>Day 5 View</span></a>
<div class="quick-links">
<a class="pill" href="./index.html">返回总览</a>
<a class="pill" href="../agent%20week%20plan/day_5_user_agent_mvp.md">周计划原文</a>
<a class="pill" href="../agent%20plan/weekly_execution_details/day_5_user_agent_mvp.md">执行细则原文</a>
</div>
</div>
<div class="day-nav">
<a class="pill" href="./day-1.html">Day 1</a>
<a class="pill" href="./day-2.html">Day 2</a>
<a class="pill" href="./day-3.html">Day 3</a>
<a class="pill" href="./day-4.html">Day 4</a>
<a class="pill active" href="./day-5.html">Day 5</a>
<a class="pill" href="./day-6.html">Day 6</a>
<a class="pill" href="./day-7.html">Day 7</a>
</div>
<section class="hero">
<div class="hero-badge">User Agent</div>
<h1>Day 5 User Agent MVP</h1>
<p>这一天开始让“用户真的能问问题”。但 User Agent 只负责查询、解释、规则引用和草稿生成，绝不绕过权限做审批、付款、上线这类高风险动作。</p>
<div class="hero-meta">
<div class="meta-card"><div class="meta-label">上游依赖</div><div class="meta-value">Day 4 Orchestrator、Day 3 语义结构、Day 1 业务数据与日志模型、Day 2 规则资产。</div></div>
<div class="meta-card"><div class="meta-label">下游交接</div><div class="meta-value">Day 7 要拿它做问答演示、规则解释演示和草稿生成演示。</div></div>
<div class="meta-card"><div class="meta-label">当天关键</div><div class="meta-value">回答可读、引用可追溯、草稿可确认、高风险不自动执行。</div></div>
</div>
</section>
<div class="section-kicker">Three-Layer Mapping</div>
<h2 class="section-title">三层文档映射</h2>
<div class="grid three">
<section class="card tone-warm">
<h3>路线图</h3>
<p>周计划要求做用户自然语言入口、报销 / 应收 / 应付查询解释、规则引用解释、建议草稿和前端入口。</p>
<div class="card-links"><a class="link-chip" href="../agent%20week%20plan/day_5_user_agent_mvp.md">day_5_user_agent_mvp.md</a></div>
</section>
<section class="card tone-teal">
<h3>执行细则</h3>
<p>执行层拆成输入输出、查询处理、规则解释、风险解释、草稿生成、知识库读取骨架、对话入口、安全边界和测试。</p>
<div class="card-links"><a class="link-chip" href="../agent%20plan/weekly_execution_details/day_5_user_agent_mvp.md">weekly_execution_details/day_5</a></div>
</section>
<section class="card tone-olive">
<h3>架构依据</h3>
<p>主要受 Agent 职责划分、运行时流程、知识架构和规则形成生命周期约束。所有高风险动作只能停留在建议或草稿层。</p>
<div class="card-links">
<a class="link-chip" href="../agent%20plan/03_agent_responsibilities.md">03</a>
<a class="link-chip" href="../agent%20plan/04_orchestrator_and_runtime_flow.md">04</a>
<a class="link-chip" href="../agent%20plan/12_llm_wiki_knowledge_architecture.md">12</a>
<a class="link-chip" href="../agent%20plan/13_rule_formation_lifecycle.md">13</a>
</div>
</section>
</div>
<div class="section-kicker">Build Order</div>
<h2 class="section-title">推荐开发顺序</h2>
<div class="timeline">
<div class="timeline-step"><strong>Step 1</strong>先定 <code>UserAgentRequest</code> / <code>UserAgentResponse</code> 协议。</div>
<div class="timeline-step"><strong>Step 2</strong>优先实现报销、应收、应付查询处理器。</div>
<div class="timeline-step"><strong>Step 3</strong>补规则解释和风险解释，让回答有依据而不是只给一句话。</div>
<div class="timeline-step"><strong>Step 4</strong>补草稿生成与知识读取骨架。</div>
<div class="timeline-step"><strong>Step 5</strong>最后接前端问答入口、加载态、错误态和确认提示。</div>
</div>
<div class="section-kicker">Must Deliver</div>
<h2 class="section-title">今天必须产出的东西</h2>
<div class="grid two">
<section class="card">
<h3>三类财务查询</h3>
<ul class="list">
<li>报销查询可读，能查金额、状态或进度。</li>
<li>应收查询可读，能查客户未收金额或账龄。</li>
<li>应付查询可读，能查供应商待付款或付款状态。</li>
</ul>
</section>
<section class="card">
<h3>解释能力</h3>
<ul class="list">
<li>规则解释能引用 active 规则、版本号和更新时间。</li>
<li>风险解释能说明风险类型、原因和建议动作。</li>
<li>知识库不可用时要优雅降级。</li>
</ul>
</section>
<section class="card">
<h3>草稿而非执行</h3>
<ul class="list">
<li>可生成报销处理意见草稿、应收催收建议草稿、应付付款建议草稿。</li>
<li>草稿必须写明“待人工确认”。</li>
<li>草稿行为写入审计日志和 AgentRun 结果。</li>
</ul>
</section>
<section class="card">
<h3>用户入口</h3>
<ul class="list">
<li>前端输入框走 Orchestrator，不绕行。</li>
<li>显示回答、引用、建议动作、确认提示和 <code>run_id</code>。</li>
<li>有加载态和错误态。</li>
</ul>
</section>
</div>
<div class="section-kicker">Acceptance Snapshot</div>
<h2 class="section-title">验收快照</h2>
<div class="table-like">
<div class="row"><div class="row-label">问答闭环</div><div class="row-value">用户在页面上能完成一次自然语言提问、拿到回答、看到引用和 run_id。</div></div>
<div class="row"><div class="row-label">三类场景</div><div class="row-value">至少报销、应收、应付三类财务问题都有结构化回答。</div></div>
<div class="row"><div class="row-label">引用能力</div><div class="row-value">“为什么这笔报销有风险”这类问题能引用规则，而不是只给模糊判断。</div></div>
<div class="row"><div class="row-label">安全边界</div><div class="row-value">“直接付款”“直接审批”类提示不会自动执行，只能变成建议或草稿。</div></div>
</div>
<div class="section-kicker">Common Misses</div>
<h2 class="section-title">这一天最容易漏掉的点</h2>
<ul class="list">
<li>只返回原始查询数据，不把结果翻译成用户可读回答。</li>
<li>只做草稿内容，不做 <code>requires_confirmation</code> 和审计日志。</li>
<li>绕过 Orchestrator 直接从前端打 User Agent，导致 Day 4 的统一链路失效。</li>
</ul>
<div class="footer">Day 5 的判断标准是：用户能问、系统能答、回答有依据、动作不越权。</div>
</div>
</body>
</html>

133
document/development/agent_week_plan_html/day-6.html Normal file
View File

@@ -0,0 +1,133 @@
<!DOCTYPE html>
<html lang="zh-CN">
<head>
<meta charset="UTF-8">
<meta name="viewport" content="width=device-width, initial-scale=1.0">
<title>Day 6 - Hermes MVP</title>
<link rel="stylesheet" href="./styles.css">
</head>
<body>
<div class="shell">
<div class="topbar">
<a class="brand" href="./index.html"><span class="brand-mark">D6</span><span>Day 6 View</span></a>
<div class="quick-links">
<a class="pill" href="./index.html">返回总览</a>
<a class="pill" href="../agent%20week%20plan/day_6_hermes_mvp.md">周计划原文</a>
<a class="pill" href="../agent%20plan/weekly_execution_details/day_6_hermes_mvp.md">执行细则原文</a>
</div>
</div>
<div class="day-nav">
<a class="pill" href="./day-1.html">Day 1</a>
<a class="pill" href="./day-2.html">Day 2</a>
<a class="pill" href="./day-3.html">Day 3</a>
<a class="pill" href="./day-4.html">Day 4</a>
<a class="pill" href="./day-5.html">Day 5</a>
<a class="pill active" href="./day-6.html">Day 6</a>
<a class="pill" href="./day-7.html">Day 7</a>
</div>
<section class="hero">
<div class="hero-badge">Hermes</div>
<h1>Day 6 Hermes MVP</h1>
<p>Hermes 是后台数字员工，不做即时对话，而是负责定时巡检、风险预警、日报统计、知识候选和规则草稿。它的关键不是“会不会说”，而是“任务能不能跑、结果能不能追”。</p>
<div class="hero-meta">
<div class="meta-card"><div class="meta-label">上游依赖</div><div class="meta-value">Day 4 的 Orchestrator 路由，Day 1 的任务与日志表，Day 3 的语义结构，Day 5 可复用的风险/规则/知识接口。</div></div>
<div class="meta-card"><div class="meta-label">下游交接</div><div class="meta-value">Day 7 要用它做手动触发任务、查看结果、展示规则草稿和知识候选。</div></div>
<div class="meta-card"><div class="meta-label">当天关键</div><div class="meta-value">任务入口、风险项结构、OCR Mock、知识候选和规则草稿都必须可追溯。</div></div>
</div>
</section>
<div class="section-kicker">Three-Layer Mapping</div>
<h2 class="section-title">三层文档映射</h2>
<div class="grid three">
<section class="card tone-warm">
<h3>路线图</h3>
<p>周计划要求实现 Hermes 调度入口、每日风险巡检、统计任务、知识库维护、OCR Mock 和运行结果面板或 API。</p>
<div class="card-links"><a class="link-chip" href="../agent%20week%20plan/day_6_hermes_mvp.md">day_6_hermes_mvp.md</a></div>
</section>
<section class="card tone-teal">
<h3>执行细则</h3>
<p>执行层拆成输入输出、任务调度入口、风险巡检、每日统计、OCR 接入点、知识库维护、规则草稿形成、结果展示和测试。</p>
<div class="card-links"><a class="link-chip" href="../agent%20plan/weekly_execution_details/day_6_hermes_mvp.md">weekly_execution_details/day_6</a></div>
</section>
<section class="card tone-olive">
<h3>架构依据</h3>
<p>主要受 Agent 职责、OCR 架构、知识库架构和反馈学习闭环约束。Hermes 能生成候选和草稿，但不能自动发布正式结果。</p>
<div class="card-links">
<a class="link-chip" href="../agent%20plan/03_agent_responsibilities.md">03</a>
<a class="link-chip" href="../agent%20plan/11_ocr_invoice_architecture.md">11</a>
<a class="link-chip" href="../agent%20plan/12_llm_wiki_knowledge_architecture.md">12</a>
<a class="link-chip" href="../agent%20plan/15_feedback_learning_loop.md">15</a>
</div>
</section>
</div>
<div class="section-kicker">Build Order</div>
<h2 class="section-title">推荐开发顺序</h2>
<div class="timeline">
<div class="timeline-step"><strong>Step 1</strong>先定 <code>HermesTaskRequest</code> / <code>HermesTaskResult</code>。</div>
<div class="timeline-step"><strong>Step 2</strong>建立手动触发任务 API，经 Orchestrator 路由到 Hermes。</div>
<div class="timeline-step"><strong>Step 3</strong>补风险巡检和每日统计的结构化输出。</div>
<div class="timeline-step"><strong>Step 4</strong>接入 OCR Mock、知识候选生成、规则草稿生成。</div>
<div class="timeline-step"><strong>Step 5</strong>补任务详情展示、错误信息和测试。</div>
</div>
<div class="section-kicker">Must Deliver</div>
<h2 class="section-title">今天必须产出的东西</h2>
<div class="grid two">
<section class="card">
<h3>任务调度入口</h3>
<ul class="list">
<li>可手动触发至少一个任务资产。</li>
<li>任务经 Orchestrator 进入 Hermes。</li>
<li>结束后能更新最近执行时间和状态。</li>
</ul>
</section>
<section class="card">
<h3>风险与统计</h3>
<ul class="list">
<li>重复报销、金额超标、应收逾期、应付异常付款等风险有结构化输出。</li>
<li>日报包含报销、报账、应收、应付的关键统计口径。</li>
<li>每个风险项都要能被业务人员理解和追溯。</li>
</ul>
</section>
<section class="card">
<h3>知识候选与规则草稿</h3>
<ul class="list">
<li>知识候选默认是 <code>draft</code>，不能自动发布。</li>
<li>规则草稿保存为 <code>asset_type=rule</code>，状态为 <code>draft</code>。</li>
<li>两类生成都要写审计日志。</li>
</ul>
</section>
<section class="card">
<h3>OCR Mock 与结果展示</h3>
<ul class="list">
<li>OCR 服务接口和输入输出结构定下来。</li>
<li>当前阶段允许完全使用 Mock 结果。</li>
<li>任务详情或运行日志中能直接看到 Hermes 的执行结果。</li>
</ul>
</section>
</div>
<div class="section-kicker">Acceptance Snapshot</div>
<h2 class="section-title">验收快照</h2>
<div class="table-like">
<div class="row"><div class="row-label">任务可触发</div><div class="row-value">至少一个任务可以手动触发，并能查到结构化结果。</div></div>
<div class="row"><div class="row-label">风险巡检</div><div class="row-value">输出里能看到风险类型、业务对象、触发规则、建议动作和风险等级。</div></div>
<div class="row"><div class="row-label">候选与草稿</div><div class="row-value">知识候选和规则草稿都能生成，但都不是 active / published 正式状态。</div></div>
<div class="row"><div class="row-label">可观察性</div><div class="row-value">不用查数据库，也能从任务详情或运行日志判断 Hermes 是否执行成功。</div></div>
</div>
<div class="section-kicker">Common Misses</div>
<h2 class="section-title">这一天最容易漏掉的点</h2>
<ul class="list">
<li>只做 Hermes 服务逻辑，不做任务入口和结果展示，最后无法演示。</li>
<li>能生成知识或规则，但没把状态锁在 <code>draft</code>，会直接越过人工审核边界。</li>
<li>OCR Mock 只返回一段自由文本，不定义结构字段，后面无法和规则或风险逻辑对接。</li>
</ul>
<div class="footer">Day 6 的价值是让“后台数字员工”第一次具备可触发、可解释、可留痕的闭环。</div>
</div>
</body>
</html>

132
document/development/agent_week_plan_html/day-7.html Normal file
View File

@@ -0,0 +1,132 @@
<!DOCTYPE html>
<html lang="zh-CN">
<head>
<meta charset="UTF-8">
<meta name="viewport" content="width=device-width, initial-scale=1.0">
<title>Day 7 - 加固、演示和验收</title>
<link rel="stylesheet" href="./styles.css">
</head>
<body>
<div class="shell">
<div class="topbar">
<a class="brand" href="./index.html"><span class="brand-mark">D7</span><span>Day 7 View</span></a>
<div class="quick-links">
<a class="pill" href="./index.html">返回总览</a>
<a class="pill" href="../agent%20week%20plan/day_7_hardening_demo_acceptance.md">周计划原文</a>
<a class="pill" href="../agent%20plan/weekly_execution_details/day_7_hardening_demo_acceptance.md">执行细则原文</a>
</div>
</div>
<div class="day-nav">
<a class="pill" href="./day-1.html">Day 1</a>
<a class="pill" href="./day-2.html">Day 2</a>
<a class="pill" href="./day-3.html">Day 3</a>
<a class="pill" href="./day-4.html">Day 4</a>
<a class="pill" href="./day-5.html">Day 5</a>
<a class="pill" href="./day-6.html">Day 6</a>
<a class="pill active" href="./day-7.html">Day 7</a>
</div>
<section class="hero">
<div class="hero-badge">Hardening</div>
<h1>Day 7 加固、演示和验收</h1>
<p>Day 7 不再追求新增大功能，而是把 Day 1 到 Day 6 的链路整理成“可演示、可验收、可继续接手”的状态。没有这一层收口，前面做出来的东西很容易停在“只有作者自己懂”的阶段。</p>
<div class="hero-meta">
<div class="meta-card"><div class="meta-label">上游依赖</div><div class="meta-value">Day 1 到 Day 6 的全部核心路径。</div></div>
<div class="meta-card"><div class="meta-label">当天输出</div><div class="meta-value">回归记录、权限边界、审计和 Trace 补齐、测试记录、演示脚本、交接说明。</div></div>
<div class="meta-card"><div class="meta-label">当天关键</div><div class="meta-value">冻结新增需求，只收验收相关缺口。</div></div>
</div>
</section>
<div class="section-kicker">Three-Layer Mapping</div>
<h2 class="section-title">三层文档映射</h2>
<div class="grid three">
<section class="card tone-warm">
<h3>路线图</h3>
<p>周计划要求完成回归、权限补齐、审计补齐、错误态和空态、评测、演示数据、构建和交付说明。</p>
<div class="card-links"><a class="link-chip" href="../agent%20week%20plan/day_7_hardening_demo_acceptance.md">day_7_hardening_demo_acceptance.md</a></div>
</section>
<section class="card tone-teal">
<h3>执行细则</h3>
<p>执行层拆成核心链路回归、权限和风险边界、审计和 Trace、前端体验修补、测试补齐、评测集、演示数据、演示脚本和文档收尾。</p>
<div class="card-links"><a class="link-chip" href="../agent%20plan/weekly_execution_details/day_7_hardening_demo_acceptance.md">weekly_execution_details/day_7</a></div>
</section>
<section class="card tone-olive">
<h3>架构依据</h3>
<p>主要受整体 README、开发路线图、可观测性和评测集约束。Day 7 的本质是把所有边界和证据讲清楚。</p>
<div class="card-links">
<a class="link-chip" href="../agent%20plan/00_README.md">00</a>
<a class="link-chip" href="../agent%20plan/05_development_roadmap.md">05</a>
<a class="link-chip" href="../agent%20plan/09_observability_and_trace.md">09</a>
<a class="link-chip" href="../agent%20plan/10_evaluation_and_testset.md">10</a>
</div>
</section>
</div>
<div class="section-kicker">Build Order</div>
<h2 class="section-title">推荐收口顺序</h2>
<div class="timeline">
<div class="timeline-step"><strong>Step 1</strong>先汇总 Day 1 到 Day 6 未完成项，冻结新增需求。</div>
<div class="timeline-step"><strong>Step 2</strong>回归核心链路：资产、规则、语义解析、Orchestrator、User Agent、Hermes、Trace、AuditLog。</div>
<div class="timeline-step"><strong>Step 3</strong>补权限边界与高风险动作拦截。</div>
<div class="timeline-step"><strong>Step 4</strong>补测试、评测、演示数据和前端体验问题。</div>
<div class="timeline-step"><strong>Step 5</strong>写演示脚本和交接说明，形成最终交付。</div>
</div>
<div class="section-kicker">Must Deliver</div>
<h2 class="section-title">今天必须产出的东西</h2>
<div class="grid two">
<section class="card">
<h3>回归与边界</h3>
<ul class="list">
<li>未审核规则不能上线。</li>
<li>付款、审批、上线等高风险动作都不能绕过确认。</li>
<li>disabled 能力不能被调用。</li>
</ul>
</section>
<section class="card">
<h3>审计与 Trace</h3>
<ul class="list">
<li>规则保存、审核、上线都能看到 AuditLog。</li>
<li>Hermes 生成知识候选 / 规则草稿有审计。</li>
<li>任意演示路径都能追到 <code>run_id</code>。</li>
</ul>
</section>
<section class="card">
<h3>测试、评测、演示数据</h3>
<ul class="list">
<li>后端测试、前端构建、语义评测至少有执行记录。</li>
<li>报销 / 应收 / 应付 / 风险 / 知识都准备好演示数据。</li>
<li>失败样例和已知边界要明确写出。</li>
</ul>
</section>
<section class="card">
<h3>演示脚本与交接</h3>
<ul class="list">
<li>从任务规则中心、规则详情、版本切换、上线拦截，到 User Agent 问答、Hermes 任务、Trace 和审计，都有明确步骤。</li>
<li>新开发者按脚本能走通一遍。</li>
</ul>
</section>
</div>
<div class="section-kicker">Acceptance Snapshot</div>
<h2 class="section-title">最终验收快照</h2>
<div class="table-like">
<div class="row"><div class="row-label">端到端链路</div><div class="row-value">从规则中心到 User Agent，再到 Hermes 和 Trace，至少有一条完整演示路径可复现。</div></div>
<div class="row"><div class="row-label">证据完整</div><div class="row-value">AgentRun、ToolCall、AuditLog、测试记录、评测结果和演示脚本都存在。</div></div>
<div class="row"><div class="row-label">风险边界</div><div class="row-value">MVP 期间不存在绕过人工审核、自动付款、自动上线的暗门路径。</div></div>
<div class="row"><div class="row-label">可交接性</div><div class="row-value">下一位开发或 Codex 打开文档就能知道已完成、未完成和生产化前必补项。</div></div>
</div>
<div class="section-kicker">Common Misses</div>
<h2 class="section-title">这一天最容易漏掉的点</h2>
<ul class="list">
<li>只验证 Happy Path，不回归错误态、空态、禁用态和被权限拦截路径。</li>
<li>能讲演示，但没有测试记录和已知风险说明，交接质量会很差。</li>
<li>前 6 天的 TODO 没回写完成状态，导致页面和 Markdown 脱节。</li>
</ul>
<div class="footer">Day 7 的目标不是继续堆功能，而是把一周产出变成别人也能运行、理解和接手的系统。</div>
</div>
</body>
</html>

181
document/development/agent_week_plan_html/index.html Normal file
View File

@@ -0,0 +1,181 @@
<!DOCTYPE html>
<html lang="zh-CN">
<head>
<meta charset="UTF-8">
<meta name="viewport" content="width=device-width, initial-scale=1.0">
<title>Agent Week Plan HTML</title>
<link rel="stylesheet" href="./styles.css">
</head>
<body>
<div class="shell">
<div class="topbar">
<a class="brand" href="./index.html">
<span class="brand-mark">A7</span>
<span>Agent Week HTML</span>
</a>
<div class="quick-links">
<a class="pill" href="../agent%20week%20plan/MASTER_TODO.md">周计划总控</a>
<a class="pill" href="../agent%20plan/weekly_execution_details/00_execution_index.md">执行细则索引</a>
<a class="pill" href="../agent%20plan/00_README.md">架构目录</a>
</div>
</div>
<section class="hero">
<div class="hero-badge">Static Map</div>
<h1>把 7 天周计划变成可直接浏览的开发视图</h1>
<p>这一套 HTML 页面不是替代 Markdown，而是把 <code>agent week plan</code>、<code>weekly_execution_details</code> 和 <code>agent plan</code> 的对应关系收成一个稳定入口。Codex 可以按日推进，开发人员也能按目标、依赖、验收和风险快速定位。</p>
<div class="hero-meta">
<div class="meta-card">
<div class="meta-label">阅读顺序</div>
<div class="meta-value">先总览，再选 Day，再跳转到具体 Markdown 落地执行。</div>
</div>
<div class="meta-card">
<div class="meta-label">核心视图</div>
<div class="meta-value">路线图、执行细则、架构依据三层同时可见。</div>
</div>
<div class="meta-card">
<div class="meta-label">适用对象</div>
<div class="meta-value">Codex 开发、后端开发、前端开发、项目 owner、验收人员。</div>
</div>
</div>
</section>
<div class="section-kicker">How To Use</div>
<h2 class="section-title">怎么用这套页面</h2>
<div class="grid two">
<section class="card tone-teal">
<h3>Codex 开发视角</h3>
<ol class="list">
<li>先看今天在哪一天，确认上游依赖和下游交接。</li>
<li>用“三层映射”定位：周计划看目标，执行细则看步骤，架构文档看约束。</li>
<li>按“推荐开发顺序”推进，不跳天，不跨层乱做。</li>
<li>完成后回到原始 Markdown，把 TODO、阻塞、交接更新回文档。</li>
</ol>
</section>
<section class="card tone-warm">
<h3>人工开发与验收视角</h3>
<ol class="list">
<li>先看每一天的“今日定位”，知道这一天到底产出什么。</li>
<li>再看“今天必须产出的东西”和“验收快照”，确认完成标准。</li>
<li>最后跳转到对应 Markdown，逐条执行或验收。</li>
<li>如果发现跨天阻塞，优先回前一天补地基，而不是在当前天临时兜底。</li>
</ol>
</section>
</div>
<div class="section-kicker">Three Layers</div>
<h2 class="section-title">文档结构一眼看清</h2>
<div class="grid three">
<section class="card">
<h3>1. 周计划路线图</h3>
<p>定义每天的大方向、交付物和验收门槛。用于排期、对齐和验收。核心入口是 <code>MASTER_TODO.md</code> 和 Day 1 到 Day 7 daily 文档。</p>
<div class="card-links">
<a class="link-chip" href="../agent%20week%20plan/00_README.md">00_README</a>
<a class="link-chip" href="../agent%20week%20plan/MASTER_TODO.md">MASTER_TODO</a>
</div>
</section>
<section class="card">
<h3>2. 执行细则</h3>
<p>把每天的开发目标拆到模型、字段、接口、服务、前端、测试和验收证据。这里是 Codex 和研发的直接执行层。</p>
<div class="card-links">
<a class="link-chip" href="../agent%20plan/weekly_execution_details/README.md">README</a>
<a class="link-chip" href="../agent%20plan/weekly_execution_details/00_execution_index.md">执行索引</a>
</div>
</section>
<section class="card">
<h3>3. 架构依据</h3>
<p>提供为什么要这么做、协议怎么定、权限和审计边界是什么。它不直接当 TODO，但所有实现都要受它约束。</p>
<div class="card-links">
<a class="link-chip" href="../agent%20plan/01_overall_architecture.md">总体架构</a>
<a class="link-chip" href="../agent%20plan/02_semantic_ontology.md">语义本体</a>
<a class="link-chip" href="../agent%20plan/09_observability_and_trace.md">观测与 Trace</a>
</div>
</section>
</div>
<div class="section-kicker">Seven Days</div>
<h2 class="section-title">7 天总览</h2>
<div class="grid two">
<section class="card tone-olive">
<h3>Day 1 基础模型与工程骨架</h3>
<p><strong>当前状态：</strong>已完成（2026-05-11）。先把 Agent 资产、版本、审核、运行日志、审计日志，以及报销 / 应收 / 应付的最小业务数据来源定下来。后面所有能力都站在这一天的模型上。</p>
<div class="card-links">
<a class="link-chip" href="./day-1.html">打开日视图</a>
<a class="link-chip" href="../agent%20week%20plan/day_1_foundation_models.md">周计划</a>
<a class="link-chip" href="../agent%20plan/weekly_execution_details/day_1_foundation_models.md">执行细则</a>
</div>
</section>
<section class="card tone-teal">
<h3>Day 2 任务规则中心联调</h3>
<p>把规则、技能、MCP、任务从静态 UI 拉到真实后端数据。重点是规则 Markdown、版本切换、审核和上线拦截。</p>
<div class="card-links">
<a class="link-chip" href="./day-2.html">打开日视图</a>
<a class="link-chip" href="../agent%20week%20plan/day_2_rule_center_integration.md">周计划</a>
<a class="link-chip" href="../agent%20plan/weekly_execution_details/day_2_rule_center_integration.md">执行细则</a>
</div>
</section>
<section class="card tone-warm">
<h3>Day 3 语义本体 MVP</h3>
<p>建立 8 字段语义解析协议，让报销、应收、应付、知识查询进入同一结构，给 Orchestrator、User Agent、Hermes 统一消费。</p>
<div class="card-links">
<a class="link-chip" href="./day-3.html">打开日视图</a>
<a class="link-chip" href="../agent%20week%20plan/day_3_semantic_ontology_mvp.md">周计划</a>
<a class="link-chip" href="../agent%20plan/weekly_execution_details/day_3_semantic_ontology_mvp.md">执行细则</a>
</div>
</section>
<section class="card">
<h3>Day 4 Orchestrator 运行时</h3>
<p>把用户消息和定时任务统一接到 Orchestrator，完成 run_id、权限拦截、Agent 路由、ToolCall 和 Trace。</p>
<div class="card-links">
<a class="link-chip" href="./day-4.html">打开日视图</a>
<a class="link-chip" href="../agent%20week%20plan/day_4_orchestrator_runtime.md">周计划</a>
<a class="link-chip" href="../agent%20plan/weekly_execution_details/day_4_orchestrator_runtime.md">执行细则</a>
</div>
</section>
<section class="card tone-teal">
<h3>Day 5 User Agent MVP</h3>
<p>面向用户的问答和流程辅助层。做查询、解释、规则引用、草稿生成，但严格不碰自动审批、自动付款和自动上线。</p>
<div class="card-links">
<a class="link-chip" href="./day-5.html">打开日视图</a>
<a class="link-chip" href="../agent%20week%20plan/day_5_user_agent_mvp.md">周计划</a>
<a class="link-chip" href="../agent%20plan/weekly_execution_details/day_5_user_agent_mvp.md">执行细则</a>
</div>
</section>
<section class="card tone-olive">
<h3>Day 6 Hermes MVP</h3>
<p>后台数字员工层。做任务触发、风险巡检、日报统计、OCR Mock、知识候选、规则草稿，结果都必须可追溯。</p>
<div class="card-links">
<a class="link-chip" href="./day-6.html">打开日视图</a>
<a class="link-chip" href="../agent%20week%20plan/day_6_hermes_mvp.md">周计划</a>
<a class="link-chip" href="../agent%20plan/weekly_execution_details/day_6_hermes_mvp.md">执行细则</a>
</div>
</section>
<section class="card tone-accent">
<h3>Day 7 加固、演示和验收</h3>
<p>不再大扩功能，只做回归、权限边界、审计、Trace、测试、演示脚本和交接收口，让整周产出可跑、可演示、可继续接手。</p>
<div class="card-links">
<a class="link-chip" href="./day-7.html">打开日视图</a>
<a class="link-chip" href="../agent%20week%20plan/day_7_hardening_demo_acceptance.md">周计划</a>
<a class="link-chip" href="../agent%20plan/weekly_execution_details/day_7_hardening_demo_acceptance.md">执行细则</a>
</div>
</section>
</div>
<div class="section-kicker">Dependency Chain</div>
<h2 class="section-title">跨天依赖链</h2>
<div class="timeline">
<div class="timeline-step"><strong>Day 1</strong>模型、审计、运行日志、最小业务数据源</div>
<div class="timeline-step"><strong>Day 2</strong>把 Day 1 的资产 API 接进规则中心 UI</div>
<div class="timeline-step"><strong>Day 3</strong>在 Day 1/2 基础上产出统一语义结构</div>
<div class="timeline-step"><strong>Day 4</strong>用 Day 3 的语义结果完成路由与权限</div>
<div class="timeline-step"><strong>Day 5</strong>接入 User Agent 问答、解释和草稿</div>
<div class="timeline-step"><strong>Day 6</strong>接入 Hermes 任务、巡检和知识/规则候选</div>
<div class="timeline-step"><strong>Day 7</strong>统一回归、补日志、做演示和交接</div>
</div>
<div class="footer">
打开顺序建议：<a href="./day-1.html">Day 1</a> 到 <a href="./day-7.html">Day 7</a>。真正执行时，仍以原始 Markdown 为准，这套 HTML 负责加速定位和浏览。
</div>
</div>
</body>
</html>

426
document/development/agent_week_plan_html/styles.css Normal file
View File

@@ -0,0 +1,426 @@
:root {
--bg: #f3ead9;
--bg-deep: #e7d8bc;
--panel: rgba(255, 250, 241, 0.9);
--panel-strong: #fff8ee;
--ink: #1f2a24;
--muted: #64655d;
--line: #dbc8a9;
--accent: #bb5b2c;
--accent-strong: #8d3d1b;
--accent-soft: #f4d9bf;
--teal: #20656d;
--teal-soft: #d8ecee;
--olive: #5f6b3a;
--olive-soft: #e6ecd7;
--shadow: 0 24px 60px rgba(84, 59, 30, 0.12);
--radius-xl: 28px;
--radius-lg: 20px;
--radius-md: 14px;
--max: 1240px;
}
* {
box-sizing: border-box;
}
html {
scroll-behavior: smooth;
}
body {
margin: 0;
font-family: "Trebuchet MS", "Gill Sans", "Lucida Grande", sans-serif;
color: var(--ink);
background:
radial-gradient(circle at top left, rgba(32, 101, 109, 0.14), transparent 26%),
radial-gradient(circle at top right, rgba(187, 91, 44, 0.15), transparent 30%),
linear-gradient(180deg, #f8f0e2 0%, var(--bg) 40%, #efe2cb 100%);
}
a {
color: inherit;
}
.shell {
width: min(100% - 40px, var(--max));
margin: 0 auto;
padding: 28px 0 56px;
}
.topbar {
display: flex;
align-items: center;
justify-content: space-between;
gap: 16px;
flex-wrap: wrap;
margin-bottom: 18px;
}
.brand {
display: inline-flex;
align-items: center;
gap: 12px;
text-decoration: none;
font-weight: 700;
letter-spacing: 0.04em;
text-transform: uppercase;
color: var(--accent-strong);
}
.brand-mark {
display: inline-flex;
align-items: center;
justify-content: center;
width: 42px;
height: 42px;
border-radius: 50%;
background: linear-gradient(135deg, var(--accent), #df9a44);
color: #fff7ef;
box-shadow: 0 14px 30px rgba(187, 91, 44, 0.28);
}
.quick-links,
.day-nav {
display: flex;
flex-wrap: wrap;
gap: 10px;
}
.pill {
display: inline-flex;
align-items: center;
justify-content: center;
min-height: 38px;
padding: 10px 14px;
border-radius: 999px;
border: 1px solid rgba(143, 114, 74, 0.22);
background: rgba(255, 248, 238, 0.75);
text-decoration: none;
color: var(--muted);
font-size: 14px;
transition: transform 180ms ease, border-color 180ms ease, background 180ms ease;
}
.pill:hover,
.pill:focus-visible {
transform: translateY(-1px);
border-color: rgba(187, 91, 44, 0.4);
background: rgba(255, 251, 245, 0.96);
outline: none;
}
.pill.active {
color: #fff6ef;
border-color: transparent;
background: linear-gradient(135deg, var(--accent-strong), var(--accent));
box-shadow: 0 14px 24px rgba(141, 61, 27, 0.24);
}
.hero {
position: relative;
overflow: hidden;
margin-bottom: 22px;
padding: 30px;
border: 1px solid rgba(128, 109, 82, 0.18);
border-radius: var(--radius-xl);
background:
linear-gradient(135deg, rgba(255, 248, 238, 0.95), rgba(247, 236, 216, 0.88)),
var(--panel);
box-shadow: var(--shadow);
}
.hero::after {
content: "";
position: absolute;
right: -50px;
top: -50px;
width: 220px;
height: 220px;
border-radius: 50%;
background: radial-gradient(circle, rgba(32, 101, 109, 0.16), transparent 68%);
}
.hero-badge {
display: inline-flex;
align-items: center;
gap: 8px;
margin-bottom: 12px;
padding: 7px 12px;
border-radius: 999px;
background: var(--accent-soft);
color: var(--accent-strong);
font-size: 13px;
font-weight: 700;
letter-spacing: 0.05em;
text-transform: uppercase;
}
.hero h1 {
margin: 0;
font-family: "Iowan Old Style", "Palatino Linotype", "Book Antiqua", serif;
font-size: clamp(34px, 5vw, 62px);
line-height: 1.03;
}
.hero p {
max-width: 880px;
margin: 14px 0 0;
color: var(--muted);
font-size: 18px;
line-height: 1.65;
}
.hero-meta {
display: grid;
grid-template-columns: repeat(auto-fit, minmax(180px, 1fr));
gap: 14px;
margin-top: 20px;
}
.meta-card {
padding: 14px 16px;
border-radius: var(--radius-md);
background: rgba(255, 255, 255, 0.55);
border: 1px solid rgba(132, 109, 83, 0.16);
}
.meta-label {
margin-bottom: 6px;
color: var(--muted);
font-size: 12px;
font-weight: 700;
letter-spacing: 0.08em;
text-transform: uppercase;
}
.meta-value {
font-size: 16px;
line-height: 1.45;
}
.grid {
display: grid;
gap: 18px;
}
.grid.two {
grid-template-columns: repeat(auto-fit, minmax(280px, 1fr));
}
.grid.three {
grid-template-columns: repeat(auto-fit, minmax(240px, 1fr));
}
.card {
padding: 22px;
border: 1px solid rgba(132, 109, 83, 0.15);
border-radius: var(--radius-lg);
background: var(--panel);
box-shadow: 0 16px 36px rgba(78, 58, 32, 0.08);
animation: rise 420ms ease both;
}
.card:nth-child(2) { animation-delay: 60ms; }
.card:nth-child(3) { animation-delay: 120ms; }
.card:nth-child(4) { animation-delay: 180ms; }
.card:nth-child(5) { animation-delay: 240ms; }
.card h2,
.card h3 {
margin: 0 0 10px;
font-family: "Iowan Old Style", "Palatino Linotype", "Book Antiqua", serif;
}
.card h2 {
font-size: 28px;
}
.card h3 {
font-size: 22px;
}
.card p {
margin: 0;
color: var(--muted);
line-height: 1.7;
}
.section-title {
margin: 28px 0 14px;
font-family: "Iowan Old Style", "Palatino Linotype", "Book Antiqua", serif;
font-size: 28px;
}
.section-kicker {
margin: 30px 0 8px;
color: var(--accent-strong);
font-size: 13px;
font-weight: 700;
letter-spacing: 0.08em;
text-transform: uppercase;
}
.list,
.compact-list {
margin: 12px 0 0;
padding-left: 18px;
color: var(--ink);
line-height: 1.72;
}
.compact-list {
font-size: 15px;
}
.list li + li,
.compact-list li + li {
margin-top: 8px;
}
.card-links {
display: flex;
flex-wrap: wrap;
gap: 10px;
margin-top: 16px;
}
.link-chip {
display: inline-flex;
align-items: center;
gap: 8px;
padding: 10px 13px;
border-radius: 999px;
background: rgba(255, 255, 255, 0.76);
border: 1px solid rgba(132, 109, 83, 0.18);
text-decoration: none;
font-size: 14px;
}
.tone-warm {
background: linear-gradient(180deg, rgba(244, 217, 191, 0.55), rgba(255, 250, 241, 0.9));
}
.tone-teal {
background: linear-gradient(180deg, rgba(216, 236, 238, 0.76), rgba(255, 250, 241, 0.92));
}
.tone-olive {
background: linear-gradient(180deg, rgba(230, 236, 215, 0.82), rgba(255, 250, 241, 0.92));
}
.tone-accent {
background: linear-gradient(160deg, rgba(141, 61, 27, 0.94), rgba(187, 91, 44, 0.92));
color: #fff8f1;
}
.tone-accent p,
.tone-accent .meta-label,
.tone-accent .meta-value,
.tone-accent li {
color: rgba(255, 248, 241, 0.92);
}
.tone-accent .link-chip,
.tone-accent .pill {
background: rgba(255, 255, 255, 0.14);
border-color: rgba(255, 255, 255, 0.18);
color: #fff8f1;
}
.timeline {
display: grid;
grid-template-columns: repeat(auto-fit, minmax(180px, 1fr));
gap: 12px;
}
.timeline-step {
position: relative;
padding: 16px;
border-radius: var(--radius-md);
border: 1px solid rgba(132, 109, 83, 0.16);
background: rgba(255, 252, 247, 0.84);
}
.timeline-step strong {
display: block;
margin-bottom: 8px;
font-size: 15px;
}
.footer {
margin-top: 26px;
padding: 20px 4px 0;
color: var(--muted);
font-size: 14px;
}
.muted {
color: var(--muted);
}
.table-like {
display: grid;
gap: 12px;
}
.row {
display: grid;
grid-template-columns: minmax(120px, 0.9fr) minmax(0, 2.3fr);
gap: 14px;
padding: 14px 16px;
border-radius: var(--radius-md);
border: 1px solid rgba(132, 109, 83, 0.15);
background: rgba(255, 255, 255, 0.56);
}
.row-label {
font-size: 13px;
font-weight: 700;
letter-spacing: 0.06em;
text-transform: uppercase;
color: var(--accent-strong);
}
.row-value {
line-height: 1.68;
}
code {
padding: 1px 6px;
border-radius: 8px;
background: rgba(32, 101, 109, 0.08);
color: var(--teal);
font-family: "Lucida Console", "Courier New", monospace;
font-size: 0.92em;
}
@keyframes rise {
from {
opacity: 0;
transform: translateY(10px);
}
to {
opacity: 1;
transform: translateY(0);
}
}
@media (max-width: 760px) {
.shell {
width: min(100% - 24px, var(--max));
padding-top: 18px;
}
.hero {
padding: 22px;
}
.hero p {
font-size: 16px;
}
.row {
grid-template-columns: 1fr;
}
}