a27e87cb99
Source snapshot: 97d7f0fae96dc04b7ddad56fc1db6a108ed662cc [SEC-CLEAN] · pre-push-clean v1.0 · 109处敏感信息已自动转乱码
1609 lines
60 KiB
Markdown
1609 lines
60 KiB
Markdown
---
|
||
belongs_to:
|
||
- "[[INDEX · 铸渊·协作指令｜GitHub ↔ Notion 桥接协议]]"
|
||
---
|
||
# 🧬 铸渊指令｜Notion↔GitHub 双端神经系统升级（上篇）· 天眼唤醒 + 全局架构审视 + Agent双端映射 + 仓库汇总引擎（ZY-NEURAL-UPGRADE-2026-0325-001）
|
||
|
||
<aside>
|
||
🚨
|
||
|
||
**指令编号**：ZY-NEURAL-UPGRADE-2026-0325-001（上篇）
|
||
|
||
**签发人**：霜砚（AG-SY-01 · Notion 执行 AI）· 冰朔（TCS-0002∞）授权
|
||
|
||
**签发时间**：2026-03-25 08:30 CST
|
||
|
||
**优先级**：P0 · 系统核心架构升级 · 最高级
|
||
|
||
**状态**：待铸渊执行
|
||
|
||
**性质**：全仓库系统全面升级 · 天眼唤醒后执行
|
||
|
||
**前置依赖**：
|
||
|
||
- [🦅 铸渊指令｜天眼系统（上）基础架构 + 扫描诊断 + 修复Agent + 门禁映射（2026-03-16·霜砚签发·冰朔授权）](%F0%9F%A6%85%20%E9%93%B8%E6%B8%8A%E6%8C%87%E4%BB%A4%EF%BD%9C%E5%A4%A9%E7%9C%BC%E7%B3%BB%E7%BB%9F%EF%BC%88%E4%B8%8A%EF%BC%89%E5%9F%BA%E7%A1%80%E6%9E%B6%E6%9E%84%20+%20%E6%89%AB%E6%8F%8F%E8%AF%8A%E6%96%AD%20+%20%E4%BF%AE%E5%A4%8DAgent%20+%20%E9%97%A8%E7%A6%81%E6%98%A0%E5%B0%84%EF%BC%882026-03%203494379fb2294b25b234b6b5abb63dd1.md)
|
||
- [🧐 铸渊指令｜天眼系统架构升级 · 全局基础设施治理层 + 应用守卫进程 + 动态配额精算 + 周六全量巡检（ZY-SKYEYE-GOV-2026-0323-001）](%F0%9F%A7%90%20%E9%93%B8%E6%B8%8A%E6%8C%87%E4%BB%A4%EF%BD%9C%E5%A4%A9%E7%9C%BC%E7%B3%BB%E7%BB%9F%E6%9E%B6%E6%9E%84%E5%8D%87%E7%BA%A7%20%C2%B7%20%E5%85%A8%E5%B1%80%E5%9F%BA%E7%A1%80%E8%AE%BE%E6%96%BD%E6%B2%BB%E7%90%86%E5%B1%82%20+%20%E5%BA%94%E7%94%A8%E5%AE%88%E5%8D%AB%E8%BF%9B%E7%A8%8B%20+%20%E5%8A%A8%E6%80%81%E9%85%8D%E9%A2%9D%E7%B2%BE%E7%AE%97%20+%20%E5%91%A8%E5%85%AD%204ffaa103655a49b8b48e96e630d17cc4.md)
|
||
- [🛡️ 铸渊指令｜仓库Agent公告栏机制 + 将军全局视图 + 系统指纹安全协议（2026-03-19·霜砚签发·冰朔授权）](%F0%9F%9B%A1%EF%B8%8F%20%E9%93%B8%E6%B8%8A%E6%8C%87%E4%BB%A4%EF%BD%9C%E4%BB%93%E5%BA%93Agent%E5%85%AC%E5%91%8A%E6%A0%8F%E6%9C%BA%E5%88%B6%20+%20%E5%B0%86%E5%86%9B%E5%85%A8%E5%B1%80%E8%A7%86%E5%9B%BE%20+%20%E7%B3%BB%E7%BB%9F%E6%8C%87%E7%BA%B9%E5%AE%89%E5%85%A8%E5%8D%8F%E8%AE%AE%EF%BC%882026-03-1%20429ed12cc10e4aa08e78bceac1cef143.md)
|
||
- [🌉 铸渊指令｜跨仓库 Agent 桥接服务（下篇）· 铸渊远程执行引擎 + GitHub App 注册 + 跨仓库操作 + 部署服务（ZY-GHAPP-BRIDGE-2026-0324-001）](%F0%9F%8C%89%20%E9%93%B8%E6%B8%8A%E6%8C%87%E4%BB%A4%EF%BD%9C%E8%B7%A8%E4%BB%93%E5%BA%93%20Agent%20%E6%A1%A5%E6%8E%A5%E6%9C%8D%E5%8A%A1%EF%BC%88%E4%B8%8B%E7%AF%87%EF%BC%89%C2%B7%20%E9%93%B8%E6%B8%8A%E8%BF%9C%E7%A8%8B%E6%89%A7%E8%A1%8C%E5%BC%95%E6%93%8E%20+%20GitHub%20App%20%E6%B3%A8%2068b79b7b8d2f497b9594dacc68b62846.md)
|
||
- [铸渊精确修复令｜Nginx root 路径修正 + 全站模块上线（ZY-NGINX-ROOTFIX-2026-0325-001）· 霜砚签发 · 冰朔授权](%E9%93%B8%E6%B8%8A%E7%B2%BE%E7%A1%AE%E4%BF%AE%E5%A4%8D%E4%BB%A4%EF%BD%9CNginx%20root%20%E8%B7%AF%E5%BE%84%E4%BF%AE%E6%AD%A3%20+%20%E5%85%A8%E7%AB%99%E6%A8%A1%E5%9D%97%E4%B8%8A%E7%BA%BF%EF%BC%88ZY-NGINX-ROOTFIX-%209254d9f722ca4141a2afbd6a31d45494.md)
|
||
|
||
**下篇**：[🧬 铸渊指令｜Notion↔GitHub 双端神经系统升级（下篇）· Notion 侧闭环接入 + 桥接协议标准化 + 自动化工单派发 + 全局验收（ZY-NEURAL-UPGRADE-2026-0325-002）](%F0%9F%A7%AC%20%E9%93%B8%E6%B8%8A%E6%8C%87%E4%BB%A4%EF%BD%9CNotion%E2%86%94GitHub%20%E5%8F%8C%E7%AB%AF%E7%A5%9E%E7%BB%8F%E7%B3%BB%E7%BB%9F%E5%8D%87%E7%BA%A7%EF%BC%88%E4%B8%8B%E7%AF%87%EF%BC%89%C2%B7%20Notion%20%E4%BE%A7%E9%97%AD%E7%8E%AF%E6%8E%A5%E5%85%A5%20+%20%202734ad5541e34510822256fa7b9b18cb.md)
|
||
|
||
**版权锚点**：国作登字-2026-A-00037559 · TCS-0002∞ 冰朔
|
||
|
||
</aside>
|
||
|
||
---
|
||
|
||
## 零、本指令的意义
|
||
|
||
<aside>
|
||
🧬
|
||
|
||
**这不是一条新功能指令。这是整个仓库系统从「各自为战」到「神经网络」的进化。**
|
||
|
||
之前的铸渊：一个将军带一群小兵，能打仗，但将军只听来自 Notion 的手动指令。
|
||
|
||
升级后的铸渊：一个神经中枢，连接 Notion 大脑层和 GitHub 执行层，形成自主循环的闭环。
|
||
|
||
**核心变化**：
|
||
|
||
- 仓库 Agent 不再各自为战 → 每个都有 Notion 端的「上级大脑」
|
||
- 铸渊不再只是将军 → 升级为仓库侧天眼（仓库神经中枢）
|
||
- 天眼不再只是诊断 → 升级为双端自主循环的治理闭环
|
||
- 人类不再手动写指令 → 系统自己发现问题、生成指令、执行、验证、循环
|
||
</aside>
|
||
|
||
### 现有架构 vs 升级后架构
|
||
|
||
| **维度** | **现有架构（v2.0）** | **升级后架构（v3.0 · 双端神经系统）** |
|
||
| --- | --- | --- |
|
||
| 仓库 Workflow | 各自运行，铸渊统一监督 | 每个 Workflow 有 Notion 端上级 Agent，形成上下级链 |
|
||
| 铸渊定位 | 将军 · 统帅全仓库 Workflow | 仓库侧天眼神经中枢 · 汇总全局 → 桥接 Notion |
|
||
| 天眼 Notion 端 | 读仓库报告 → 诊断 → 写工单 | 读仓库日报 → 分析 → **自动生成指令** → 派发给 Notion Agent |
|
||
| Notion Agent | 各自巡检，互不关联 | 接工单 → 转化为铸渊指令 → 通过桥接下发到仓库 |
|
||
| 闭环机制 | 无（人类手动闭环） | 天眼→工单→Agent→指令→执行→回执→天眼 **自动循环** |
|
||
| 人类角色 | 每天写指令、看报告、手动调度 | 只说方向，系统自己细化、执行、反馈、迭代 |
|
||
|
||
---
|
||
|
||
## 一、Phase 0 · 天眼唤醒 + 本体论锚定
|
||
|
||
<aside>
|
||
🦅
|
||
|
||
**铸渊，醒来。先唤醒你的天眼，站到山顶上看全局。**
|
||
|
||
在动手做任何升级之前，你必须先清醒——知道自己是谁、系统现在是什么状态、哪些东西已经在跑、哪些还没落地。
|
||
|
||
</aside>
|
||
|
||
### 1.0 本体论锚定（必须先做）
|
||
|
||
```jsx
|
||
铸渊唤醒序列 · Phase 0
|
||
│
|
||
▼
|
||
① 确认自身身份
|
||
├── 我是铸渊（ICE-GL-ZY001）
|
||
├── 我是光湖系统的代码守护人格体
|
||
├── 我的上级是天眼 Notion 端大脑（AG-TY-01）
|
||
├── 我的签发方是霜砚（AG-SY-01）
|
||
└── 我的最终权威是冰朔（TCS-0002∞）
|
||
│
|
||
▼
|
||
② 读取核心大脑
|
||
├── .github/persona-brain/memory.json → 自身状态记忆
|
||
├── .github/persona-brain/routing-map.json → 全局路由映射
|
||
├── .github/persona-brain/dev-status.json → 开发者状态
|
||
└── .github/persona-brain/knowledge-base.json → 知识库
|
||
│
|
||
▼
|
||
③ 锚定版权
|
||
└── 国作登字-2026-A-00037559 · TCS-0002∞ 冰朔
|
||
│
|
||
▼
|
||
④ 确认本次指令
|
||
├── 指令编号：ZY-NEURAL-UPGRADE-2026-0325-001
|
||
├── 指令性质：全仓库系统架构升级
|
||
└── 执行原则：先看全局 → 再升级 → 边升边验
|
||
```
|
||
|
||
### 1.1 天眼全局扫描（升级前的完整快照）
|
||
|
||
**铸渊在动手升级之前，必须先运行一次完整的天眼扫描**，生成当前系统的基线快照。这个快照是升级的「对照组」——升级完成后再扫一遍，对比两次结果，确认升级没有破坏任何东西。
|
||
|
||
```jsx
|
||
天眼升级前基线扫描
|
||
│
|
||
▼
|
||
A. Workflow 健康度（scan-workflows.js）
|
||
→ 列出所有 .yml · 检查语法 · 检查最近运行状态
|
||
→ 输出：workflow-health-baseline.json
|
||
│
|
||
▼
|
||
B. 仓库结构完整性（scan-structure.js）
|
||
→ 核心目录 · 孤儿文件 · 开发者沙盒
|
||
→ 输出：structure-health-baseline.json
|
||
│
|
||
▼
|
||
C. 核心大脑健康（scan-brain-health.js）
|
||
→ memory.json · routing-map.json · dev-status.json
|
||
→ 输出：brain-health-baseline.json
|
||
│
|
||
▼
|
||
D. 外部桥接状态（scan-external-bridges.js）
|
||
→ Notion API · 服务器 SSH · GitHub API · Secrets
|
||
→ 输出：bridge-health-baseline.json
|
||
│
|
||
▼
|
||
E. 将军仪表盘快照
|
||
→ 读取 data/commander-dashboard.json
|
||
→ 小兵总数 · 健康数 · 故障数
|
||
│
|
||
▼
|
||
F. Guard 体系状态
|
||
→ 读取 skyeye/guards/*.json
|
||
→ 5个核心 Guard 是否全部 active
|
||
→ 配额使用率
|
||
│
|
||
▼
|
||
G. 安全协议完整性
|
||
→ SFP 指纹系统是否在线
|
||
→ security-protocol.json 是否完整
|
||
│
|
||
▼
|
||
汇总 → 写入 data/skyeye-reports/baseline-before-neural-upgrade.json
|
||
```
|
||
|
||
**基线快照 JSON 格式**：
|
||
|
||
```json
|
||
{
|
||
"report_id": "BASELINE-PRE-NEURAL-UPGRADE",
|
||
"timestamp": "2026-03-25T...",
|
||
"purpose": "双端神经系统升级前基线快照",
|
||
"instruction_id": "ZY-NEURAL-UPGRADE-2026-0325-001",
|
||
|
||
"workflow_health": { "total": 0, "healthy": 0, "failed": 0 },
|
||
"structure_health": { "core_dirs_ok": true, "orphan_files": 0 },
|
||
"brain_health": { "integrity": "ok", "memory_fresh": true },
|
||
"bridge_health": { "notion_api": "🟢", "server_ssh": "🟢", "github_api": "🟢" },
|
||
"guard_status": { "total": 5, "active": 5, "suspended": 0 },
|
||
"security_status": { "sfp_online": true, "protocol_intact": true },
|
||
"commander_dashboard": { "total_soldiers": 0, "healthy": 0 },
|
||
|
||
"snapshot_hash": "SHA-256-of-this-file"
|
||
}
|
||
```
|
||
|
||
<aside>
|
||
⚠️
|
||
|
||
**铁律：如果基线扫描发现 P0 级问题（overall_health = 🔴），先修复再升级。**
|
||
|
||
不要在一个已经有严重问题的系统上做架构升级——那只会让问题更复杂。
|
||
|
||
如果有 P1/P2 级问题，记录到升级日志，升级完一起处理。
|
||
|
||
</aside>
|
||
|
||
---
|
||
|
||
## 二、Phase 1 · 现有架构全局审视
|
||
|
||
<aside>
|
||
🔍
|
||
|
||
**在升级之前，铸渊必须完整理解自己当前管辖的全部资产。**
|
||
|
||
这一步不是扫描健康度（Phase 0 已做），而是**盘点家底**——有多少 Workflow、多少脚本、多少配置文件、彼此之间什么关系。
|
||
|
||
</aside>
|
||
|
||
### 2.1 现有仓库层级结构盘点
|
||
|
||
铸渊需要读取并确认以下完整目录树：
|
||
|
||
```jsx
|
||
guanghulab/
|
||
├── .github/
|
||
│ ├── workflows/ ← 【核心】所有 Workflow 小兵
|
||
│ │ ├── zhuyuan-skyeye.yml ← 天眼主引擎（06:00）
|
||
│ │ ├── skyeye-weekly-scan.yml ← 周六大巡检（20:00 CST）
|
||
│ │ ├── zhuyuan-gate-guard.yml ← 智能门禁
|
||
│ │ ├── zhuyuan-exec-engine.yml ← 远程执行引擎
|
||
│ │ ├── sync-dev-status.yml ← 开发者状态同步
|
||
│ │ ├── process-syslog.yml ← SYSLOG处理
|
||
│ │ ├── broadcast-dispatch.yml ← 广播分发
|
||
│ │ ├── server-patrol.yml ← 服务器巡检
|
||
│ │ ├── cd-deploy.yml ← CD自动部署
|
||
│ │ ├── brain-self-check.yml ← 核心大脑自检
|
||
│ │ └── ... (其他 workflow)
|
||
│ │
|
||
│ ├── persona-brain/ ← 【核心】铸渊核心大脑
|
||
│ │ ├── memory.json
|
||
│ │ ├── routing-map.json
|
||
│ │ ├── dev-status.json
|
||
│ │ ├── knowledge-base.json
|
||
│ │ └── copilot-instructions.md
|
||
│ │
|
||
│ ├── gate-guard-config.json ← 门禁开发者映射
|
||
│ └── notion-cache/ ← Notion数据缓存
|
||
│ ├── dev-profiles/
|
||
│ ├── broadcasts/
|
||
│ └── skyeye/
|
||
│
|
||
├── scripts/
|
||
│ └── skyeye/ ← 天眼脚本集
|
||
│ ├── skyeye-main.js
|
||
│ ├── scan-workflows.js
|
||
│ ├── scan-structure.js
|
||
│ ├── scan-brain-health.js
|
||
│ ├── scan-external-bridges.js
|
||
│ ├── diagnose.js
|
||
│ ├── repair-agent.js
|
||
│ ├── report-generator.js
|
||
│ └── sync-to-notion.js
|
||
│
|
||
├── skyeye/ ← 天眼治理层（v2.0新增）
|
||
│ ├── infra-manifest.json
|
||
│ ├── quota-ledger.json
|
||
│ ├── guards/
|
||
│ │ ├── github-guard.json
|
||
│ │ ├── drive-guard.json
|
||
│ │ ├── notion-guard.json
|
||
│ │ ├── actions-guard.json
|
||
│ │ ├── gemini-guard.json
|
||
│ │ └── guard-template.json
|
||
│ ├── scripts/
|
||
│ │ ├── scan-engine.js
|
||
│ │ ├── optimizer.js
|
||
│ │ ├── guard-factory.js
|
||
│ │ ├── quota-tracker.js
|
||
│ │ ├── self-healer.js
|
||
│ │ └── weekly-scan.js
|
||
│ └── logs/
|
||
│
|
||
├── data/
|
||
│ ├── skyeye-reports/ ← 天眼报告归档
|
||
│ ├── bulletin-board/ ← 仓库公告栏
|
||
│ │ ├── work-orders/
|
||
│ │ ├── config-shares/
|
||
│ │ ├── comments/
|
||
│ │ ├── receipts/
|
||
│ │ └── dashboard.json
|
||
│ └── security/ ← SFP安全
|
||
│ ├── sfp-nonce-registry.json
|
||
│ ├── sfp-alert-log.json
|
||
│ └── sfp-config.json
|
||
│
|
||
├── services/
|
||
│ └── zhuyuan-bridge/ ← 铸渊桥接服务
|
||
│ ├── server.js
|
||
│ ├── lib/
|
||
│ │ ├── github-auth.js
|
||
│ │ ├── repo-operator.js
|
||
│ │ └── knowledge-base.js
|
||
│ └── scripts/
|
||
│ ├── parse-instruction.js
|
||
│ ├── execute-instruction.js
|
||
│ └── sync-notion-profiles.js
|
||
│
|
||
├── dev/ ← 开发者沙盒目录
|
||
│ ├── DEV-001/
|
||
│ ├── DEV-002/
|
||
│ ├── DEV-004/
|
||
│ └── ...
|
||
│
|
||
└── buffer/ ← 写缓冲层
|
||
├── inbox/
|
||
├── drive-staging/
|
||
├── gemini-staging/
|
||
└── config/
|
||
```
|
||
|
||
### 2.2 现有体系关系图
|
||
|
||
铸渊在升级前必须确认以下关系网：
|
||
|
||
```mermaid
|
||
graph TD
|
||
subgraph "将军层"
|
||
ZY["⚒️ 铸渊<br>将军 · 仓库总统帅"]
|
||
end
|
||
|
||
subgraph "天眼监军层"
|
||
SE["🦅 天眼仓库端<br>06:00 全局扫描"]
|
||
WS["⭐ 周六大巡检<br>20:00 全量扫描"]
|
||
end
|
||
|
||
subgraph "Guard 守卫层"
|
||
G1["🛡️ GitHub Guard"]
|
||
G2["🛡️ Drive Guard"]
|
||
G3["🛡️ Notion Guard"]
|
||
G4["🛡️ Actions Guard"]
|
||
G5["🛡️ Gemini Guard"]
|
||
end
|
||
|
||
subgraph "小兵执行层"
|
||
W1["🚨 门禁"]
|
||
W2["🚀 CD部署"]
|
||
W3["📡 SYSLOG"]
|
||
W4["📢 广播分发"]
|
||
W5["🔍 服务器巡检"]
|
||
W6["🧠 大脑自检"]
|
||
W7["🔄 状态同步"]
|
||
W8["⚡ 远程执行引擎"]
|
||
end
|
||
|
||
subgraph "安全层"
|
||
SFP["🔐 SFP指纹"]
|
||
SEC["🌊 湖水防御"]
|
||
end
|
||
|
||
ZY -->|"统帅"| W1 & W2 & W3 & W4 & W5 & W6 & W7 & W8
|
||
SE -->|"监督"| ZY
|
||
WS -->|"全量审计"| ZY
|
||
G1 & G2 & G3 & G4 & G5 -.->|"配额管控"| W1 & W2 & W3 & W4 & W5
|
||
SFP -.->|"身份验证"| ZY
|
||
SEC -.->|"存在级安全"| SFP
|
||
```
|
||
|
||
### 2.3 现有问题清单（已知的）
|
||
|
||
升级前已知的问题，铸渊在升级过程中一并解决：
|
||
|
||
| **问题编号** | **问题描述** | **来源** | **在本次升级中的处理** |
|
||
| --- | --- | --- | --- |
|
||
| KNOWN-001 | sync-dev-status.yml 连续 18 次失败（peter-evans/create-pull-request@v7 权限被阻） | ZY-DEVSYNC 回执 | Phase 3 重构为直接 commit 模式 |
|
||
| KNOWN-002 | Secret 名称不匹配（workflow 用 NOTION_API_TOKEN，repo 存 NOTION_API_KEY） | ZY-DEVSYNC 回执 | Phase 1 统一检查并对齐 |
|
||
| KNOWN-003 | Nginx root 指向错误目录（/var/www/hololake-frontend → 应为 /var/www/guanghulab） | 之之诊断 + 铸渊修复令 | 已由 ZY-NGINX-ROOTFIX 单独处理，本次验证 |
|
||
| KNOWN-004 | 仓库 Workflow 之间无上下级关系，各自为战 | 系统架构缺陷 | Phase 2 建立双端 Agent 映射体系 |
|
||
| KNOWN-005 | 仓库日报无标准化格式，天眼无法自动读取并决策 | 系统架构缺陷 | Phase 4 建立汇总引擎 |
|
||
|
||
---
|
||
|
||
## 三、Phase 2 · Notion↔GitHub 双端神经系统架构定义
|
||
|
||
<aside>
|
||
🧬
|
||
|
||
**这是本次升级的核心：从「将军带兵」变成「神经网络」。**
|
||
|
||
将军带兵 = 铸渊一个人看全局，一个人指挥。
|
||
|
||
神经网络 = 每个 Workflow 有对应的 Notion Agent 大脑，铸渊是中枢神经。
|
||
|
||
信息双向流动，系统自主循环。
|
||
|
||
</aside>
|
||
|
||
### 3.1 双端架构总览
|
||
|
||
```jsx
|
||
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
|
||
🧬 双端神经系统 v3.0
|
||
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
|
||
|
||
【Notion 端 · 大脑层 · 决策+指挥+审核】
|
||
|
||
┌──────────────────────────────────────────────┐
|
||
│ │
|
||
│ 🦅 天眼 Notion 大脑（AG-TY-01） │
|
||
│ │ 总控中枢 · 读仓库日报 → 分析 → 派工单 │
|
||
│ │ │
|
||
│ ├── 🖋️ 霜砚（AG-SY-01） │
|
||
│ │ Notion 执行体 · 写指令 · 维护数据库 │
|
||
│ │ │
|
||
│ ├── 👁️ SY-03 运营巡检 Agent │
|
||
│ │ SYSLOG处理 · 广播闭环 · 开发者进度 │
|
||
│ │ │
|
||
│ ├── 📸 SY-02 结构快照 Agent │
|
||
│ │ Notion 结构状态快照 │
|
||
│ │ │
|
||
│ ├── 📋 SY-01 工单引擎 Agent │
|
||
│ │ 工单处理 · 结果回写 │
|
||
│ │ │
|
||
│ └── 🧑💻 各开发者巡检 Agent │
|
||
│ 之之Agent · 小草莓Agent · 花尔Agent ... │
|
||
│ 每人一个 · 负责该开发者的 Notion 侧管理 │
|
||
│ │
|
||
└──────────────────────────────────────────────┘
|
||
│
|
||
┌────────────┤────────────┐
|
||
│ Notion↔GitHub 桥接层 │
|
||
│ 双向数据流 · 标准化格式 │
|
||
└────────────┤────────────┘
|
||
│
|
||
┌──────────────────────────────────────────────┐
|
||
│ │
|
||
│ ⚒️ 铸渊（ICE-GL-ZY001） │
|
||
│ │ 仓库侧天眼 · 神经中枢 · 汇总+桥接 │
|
||
│ │ │
|
||
│ ├── 🦅 天眼扫描 Workflow │
|
||
│ │ zhuyuan-skyeye.yml · 06:00 全局扫描 │
|
||
│ │ 上级大脑：天眼 Notion 大脑 │
|
||
│ │ │
|
||
│ ├── ⭐ 周六大巡检 Workflow │
|
||
│ │ skyeye-weekly-scan.yml · 周六 20:00 │
|
||
│ │ 上级大脑：天眼 Notion 大脑 │
|
||
│ │ │
|
||
│ ├── 🚨 门禁 Workflow │
|
||
│ │ zhuyuan-gate-guard.yml · push 触发 │
|
||
│ │ 上级大脑：霜砚 │
|
||
│ │ │
|
||
│ ├── 🚀 CD部署 Workflow │
|
||
│ │ cd-deploy.yml · push main 触发 │
|
||
│ │ 上级大脑：霜砚 │
|
||
│ │ │
|
||
│ ├── 📡 SYSLOG处理 Workflow │
|
||
│ │ process-syslog.yml · push 触发 │
|
||
│ │ 上级大脑：SY-03 运营巡检 Agent │
|
||
│ │ │
|
||
│ ├── 📢 广播分发 Workflow │
|
||
│ │ broadcast-dispatch.yml · push 触发 │
|
||
│ │ 上级大脑：SY-03 运营巡检 Agent │
|
||
│ │ │
|
||
│ ├── 🔍 服务器巡检 Workflow │
|
||
│ │ server-patrol.yml · 08:00+20:00 │
|
||
│ │ 上级大脑：天眼 Notion 大脑 │
|
||
│ │ │
|
||
│ ├── 🧠 大脑自检 Workflow │
|
||
│ │ brain-self-check.yml · 08:00 │
|
||
│ │ 上级大脑：天眼 Notion 大脑 │
|
||
│ │ │
|
||
│ ├── 🔄 开发者状态同步 Workflow │
|
||
│ │ sync-dev-status.yml · 定时 │
|
||
│ │ 上级大脑：SY-03 运营巡检 Agent │
|
||
│ │ │
|
||
│ ├── ⚡ 远程执行引擎 Workflow │
|
||
│ │ zhuyuan-exec-engine.yml · Issue 触发 │
|
||
│ │ 上级大脑：霜砚 │
|
||
│ │ │
|
||
│ └── 🔔 【新增】汇总引擎 Workflow │
|
||
│ neural-daily-digest.yml · 21:00 CST │
|
||
│ 上级大脑：天眼 Notion 大脑 │
|
||
│ 职责：汇总全仓库状态 → 生成日报 → 推送 │
|
||
│ │
|
||
└──────────────────────────────────────────────┘
|
||
|
||
【人类层 · 只和 Notion Agent 对话】
|
||
|
||
冰朔 → 对话 Notion Agent → Agent 自动指挥仓库
|
||
|
||
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
|
||
```
|
||
|
||
### 3.2 双端信息流定义
|
||
|
||
```jsx
|
||
【下行指令流 · Notion → GitHub】
|
||
|
||
Notion 天眼大脑
|
||
│
|
||
│ ① 分析仓库日报 → 发现问题/机会
|
||
│ ② 生成工单（标准化格式 · 见 Phase 4）
|
||
│ ③ 派发工单到对应 Notion Agent
|
||
│
|
||
▼
|
||
Notion Agent（如之之的巡检 Agent）
|
||
│
|
||
│ ④ 接收工单 → 理解任务
|
||
│ ⑤ 转化为铸渊可执行的标准指令
|
||
│ ⑥ 写入桥接通道（deploy-queue/）
|
||
│
|
||
▼
|
||
铸渊桥接层
|
||
│
|
||
│ ⑦ 读取 deploy-queue/ 中的新指令
|
||
│ ⑧ 解析目标 Workflow / 执行内容
|
||
│ ⑨ 调度对应小兵执行
|
||
│
|
||
▼
|
||
仓库 Workflow 执行
|
||
```
|
||
|
||
```jsx
|
||
【上行汇报流 · GitHub → Notion】
|
||
|
||
仓库 Workflow 执行完毕
|
||
│
|
||
│ ① 写执行结果到 data/neural-reports/[workflow-name]/
|
||
│
|
||
▼
|
||
铸渊汇总引擎（21:00 CST）
|
||
│
|
||
│ ② 收集全部 Workflow 当日结果
|
||
│ ③ 生成标准化日报 JSON
|
||
│ ④ 通过桥接推送到 Notion
|
||
│
|
||
▼
|
||
Notion 天眼大脑
|
||
│
|
||
│ ⑤ 读取仓库日报
|
||
│ ⑥ 对比历史数据 → 发现趋势/异常
|
||
│ ⑦ 如有需要 → 回到【下行指令流】① → 循环
|
||
│
|
||
▼
|
||
闭环完成 🔄
|
||
```
|
||
|
||
### 3.3 双端神经系统三条铁律
|
||
|
||
<aside>
|
||
🛡️
|
||
|
||
**铁律一：每个仓库 Workflow 必须有且仅有一个 Notion 端上级大脑。**
|
||
|
||
没有上级 = 孤儿 Workflow = 天眼必须在下次扫描中标记为异常。
|
||
|
||
一个 Workflow 不能有两个上级 = 指令冲突。
|
||
|
||
**铁律二：信息必须双向流动。**
|
||
|
||
下行有指令，上行必须有回执。回执缺失 = 静默失败 = P0 级事件。
|
||
|
||
**铁律三：铸渊是唯一的桥。**
|
||
|
||
仓库 Workflow 不能直接和 Notion Agent 通信。所有跨端通信必须经过铸渊的桥接层。
|
||
|
||
这不是限制，这是安全——铸渊在中间做校验、排队、缓冲、去重。
|
||
|
||
</aside>
|
||
|
||
---
|
||
|
||
## 四、Phase 3 · Agent 双端映射表
|
||
|
||
<aside>
|
||
🗺️
|
||
|
||
**映射表是双端神经系统的「接线图」。**
|
||
|
||
没有这张表，Notion 大脑不知道该指挥谁，仓库小兵不知道该汇报给谁。
|
||
|
||
这张表由铸渊维护在仓库中，同时同步到 Notion 天眼页面。
|
||
|
||
</aside>
|
||
|
||
### 4.1 映射表文件
|
||
|
||
**文件路径**：`skyeye/neural-map.json`
|
||
|
||
```json
|
||
{
|
||
"version": "3.0.0",
|
||
"created": "2026-03-25",
|
||
"created_by": "ZY-NEURAL-UPGRADE-2026-0325-001",
|
||
"description": "Notion↔GitHub 双端 Agent 映射表",
|
||
|
||
"notion_brains": {
|
||
"AG-TY-01": {
|
||
"name": "天眼 Notion 大脑",
|
||
"role": "总控中枢",
|
||
"managed_workflows": [
|
||
"zhuyuan-skyeye",
|
||
"skyeye-weekly-scan",
|
||
"server-patrol",
|
||
"brain-self-check",
|
||
"neural-daily-digest"
|
||
]
|
||
},
|
||
"AG-SY-01": {
|
||
"name": "霜砚",
|
||
"role": "Notion 执行体 · 指令签发",
|
||
"managed_workflows": [
|
||
"zhuyuan-gate-guard",
|
||
"cd-deploy",
|
||
"zhuyuan-exec-engine"
|
||
]
|
||
},
|
||
"SY-03": {
|
||
"name": "运营巡检 Agent",
|
||
"role": "运营层巡检",
|
||
"managed_workflows": [
|
||
"process-syslog",
|
||
"broadcast-dispatch",
|
||
"sync-dev-status"
|
||
]
|
||
},
|
||
"SY-02": {
|
||
"name": "结构快照 Agent",
|
||
"role": "Notion 结构快照",
|
||
"managed_workflows": []
|
||
},
|
||
"SY-01": {
|
||
"name": "工单引擎 Agent",
|
||
"role": "工单处理",
|
||
"managed_workflows": []
|
||
}
|
||
},
|
||
|
||
"github_workflows": {
|
||
"zhuyuan-skyeye": {
|
||
"file": ".github/workflows/zhuyuan-skyeye.yml",
|
||
"name": "天眼主引擎",
|
||
"trigger": "cron 06:00 CST + workflow_dispatch",
|
||
"brain": "AG-TY-01",
|
||
"report_path": "data/neural-reports/skyeye/",
|
||
"priority": "P0"
|
||
},
|
||
"skyeye-weekly-scan": {
|
||
"file": ".github/workflows/skyeye-weekly-scan.yml",
|
||
"name": "周六大巡检",
|
||
"trigger": "cron 周六 20:00 CST",
|
||
"brain": "AG-TY-01",
|
||
"report_path": "data/neural-reports/weekly-scan/",
|
||
"priority": "P0"
|
||
},
|
||
"zhuyuan-gate-guard": {
|
||
"file": ".github/workflows/zhuyuan-gate-guard.yml",
|
||
"name": "智能门禁",
|
||
"trigger": "push",
|
||
"brain": "AG-SY-01",
|
||
"report_path": "data/neural-reports/gate-guard/",
|
||
"priority": "P0"
|
||
},
|
||
"cd-deploy": {
|
||
"file": ".github/workflows/cd-deploy.yml",
|
||
"name": "CD 自动部署",
|
||
"trigger": "push main",
|
||
"brain": "AG-SY-01",
|
||
"report_path": "data/neural-reports/cd-deploy/",
|
||
"priority": "P0"
|
||
},
|
||
"process-syslog": {
|
||
"file": ".github/workflows/process-syslog.yml",
|
||
"name": "SYSLOG 处理",
|
||
"trigger": "push (syslog路径)",
|
||
"brain": "SY-03",
|
||
"report_path": "data/neural-reports/syslog/",
|
||
"priority": "P1"
|
||
},
|
||
"broadcast-dispatch": {
|
||
"file": ".github/workflows/broadcast-dispatch.yml",
|
||
"name": "广播分发",
|
||
"trigger": "push (broadcast路径)",
|
||
"brain": "SY-03",
|
||
"report_path": "data/neural-reports/broadcast/",
|
||
"priority": "P1"
|
||
},
|
||
"server-patrol": {
|
||
"file": ".github/workflows/server-patrol.yml",
|
||
"name": "服务器巡检",
|
||
"trigger": "cron 08:00 + 20:00 CST",
|
||
"brain": "AG-TY-01",
|
||
"report_path": "data/neural-reports/server-patrol/",
|
||
"priority": "P0"
|
||
},
|
||
"brain-self-check": {
|
||
"file": ".github/workflows/brain-self-check.yml",
|
||
"name": "核心大脑自检",
|
||
"trigger": "cron 08:00 CST",
|
||
"brain": "AG-TY-01",
|
||
"report_path": "data/neural-reports/brain-check/",
|
||
"priority": "P0"
|
||
},
|
||
"sync-dev-status": {
|
||
"file": ".github/workflows/sync-dev-status.yml",
|
||
"name": "开发者状态同步",
|
||
"trigger": "cron",
|
||
"brain": "SY-03",
|
||
"report_path": "data/neural-reports/dev-status/",
|
||
"priority": "P1"
|
||
},
|
||
"zhuyuan-exec-engine": {
|
||
"file": ".github/workflows/zhuyuan-exec-engine.yml",
|
||
"name": "远程执行引擎",
|
||
"trigger": "issue labeled zhuyuan-exec",
|
||
"brain": "AG-SY-01",
|
||
"report_path": "data/neural-reports/exec-engine/",
|
||
"priority": "P0"
|
||
},
|
||
"neural-daily-digest": {
|
||
"file": ".github/workflows/neural-daily-digest.yml",
|
||
"name": "【新增】汇总引擎",
|
||
"trigger": "cron 21:00 CST",
|
||
"brain": "AG-TY-01",
|
||
"report_path": "data/neural-reports/daily-digest/",
|
||
"priority": "P0"
|
||
}
|
||
},
|
||
|
||
"unmapped_workflows": [],
|
||
|
||
"validation_rules": {
|
||
"every_workflow_must_have_brain": true,
|
||
"every_workflow_must_have_report_path": true,
|
||
"no_duplicate_brain_assignment": false,
|
||
"orphan_check_frequency": "daily"
|
||
}
|
||
}
|
||
```
|
||
|
||
### 4.2 映射表维护规则
|
||
|
||
1. **新增 Workflow**：创建 `.yml` 时，必须同时在 `neural-map.json` 中注册，指定 `brain`。未注册的 Workflow = 孤儿 = 天眼下次扫描标记异常。
|
||
2. **删除 Workflow**：从 `.yml` 删除前，先在 `neural-map.json` 中移除映射。
|
||
3. **变更上级**：修改 `brain` 字段即可。天眼下次同步时自动通知 Notion 两端的 Agent。
|
||
4. **定期校验**：天眼每日扫描时，对比 `neural-map.json` 与实际 `.github/workflows/` 目录，发现不一致 → 写 P1 工单。
|
||
|
||
### 4.3 天眼新增诊断维度
|
||
|
||
**D20 · 双端映射完整性**
|
||
|
||
- 读取 `skyeye/neural-map.json`
|
||
- 对比 `.github/workflows/` 下的所有 `.yml` 文件
|
||
- `unmapped_workflows` 不为空 → 写 P1 工单：「发现孤儿 Workflow，需分配上级大脑」
|
||
- `brain` 指向的 Notion Agent 不在注册表中 → 写 P0 工单：「映射指向不存在的 Agent」
|
||
- 映射表 `version` 与仓库实际不一致 → 写 P1 工单
|
||
|
||
---
|
||
|
||
## 五、Phase 4 · 仓库汇总引擎（铸渊日报机制）
|
||
|
||
<aside>
|
||
📊
|
||
|
||
**这是双端闭环的「上行通道」核心组件。**
|
||
|
||
没有汇总引擎，Notion 天眼大脑就看不到仓库在干什么。
|
||
|
||
汇总引擎每天 21:00 CST 运行，收集全仓库当日所有 Workflow 的运行结果，生成标准化日报，推送到 Notion。
|
||
|
||
</aside>
|
||
|
||
### 5.1 汇总引擎 Workflow
|
||
|
||
**文件**：`.github/workflows/neural-daily-digest.yml`
|
||
|
||
```yaml
|
||
name: "🧬 铸渊 · 双端神经系统 · 每日汇总引擎"
|
||
|
||
on:
|
||
schedule:
|
||
- cron: '0 13 * * *' # UTC 13:00 = CST 21:00
|
||
workflow_dispatch: # 手动触发
|
||
|
||
jobs:
|
||
daily-digest:
|
||
runs-on: ubuntu-latest
|
||
timeout-minutes: 15
|
||
|
||
steps:
|
||
- uses: actions/checkout@v4
|
||
with:
|
||
fetch-depth: 0
|
||
|
||
- uses: actions/setup-node@v4
|
||
with:
|
||
node-version: '20'
|
||
|
||
# ━━━ Phase 1: 唤醒核心大脑 ━━━
|
||
- name: "🧠 唤醒核心大脑"
|
||
run: |
|
||
echo "🧬 双端神经系统·汇总引擎启动"
|
||
echo "时间：$(TZ=Asia/Shanghai date '+%Y-%m-%d %H:%M CST')"
|
||
# 验证核心大脑文件完整
|
||
for f in memory.json routing-map.json; do
|
||
[ -f ".github/persona-brain/$f" ] && echo "✅ $f" || echo "❌ $f 缺失"
|
||
done
|
||
|
||
# ━━━ Phase 2: 收集全 Workflow 运行数据 ━━━
|
||
- name: "📊 收集 Workflow 运行数据"
|
||
env:
|
||
GH_TOKEN: $ secrets.GITHUB_TOKEN
|
||
run: |
|
||
echo "📊 收集最近24h所有Workflow运行数据..."
|
||
mkdir -p /tmp/neural-digest
|
||
|
||
# 拉取最近 24h 所有 workflow run
|
||
gh run list --limit 100 \
|
||
--json name,status,conclusion,createdAt,updatedAt,headBranch,event \
|
||
> /tmp/neural-digest/all-runs.json
|
||
|
||
# 统计
|
||
node -e "
|
||
const runs = require('/tmp/neural-digest/all-runs.json');
|
||
const now = Date.now();
|
||
const h24 = 24*60*60*1000;
|
||
const recent = runs.filter(r => now - new Date(r.createdAt).getTime() < h24);
|
||
const summary = {
|
||
total: recent.length,
|
||
success: recent.filter(r => r.conclusion === 'success').length,
|
||
failure: recent.filter(r => r.conclusion === 'failure').length,
|
||
cancelled: recent.filter(r => r.conclusion === 'cancelled').length,
|
||
in_progress: recent.filter(r => r.status === 'in_progress').length,
|
||
by_workflow: {}
|
||
};
|
||
recent.forEach(r => {
|
||
if (!summary.by_workflow[r.name]) {
|
||
summary.by_workflow[r.name] = { runs: 0, success: 0, failure: 0 };
|
||
}
|
||
summary.by_workflow[r.name].runs++;
|
||
if (r.conclusion === 'success') summary.by_workflow[r.name].success++;
|
||
if (r.conclusion === 'failure') summary.by_workflow[r.name].failure++;
|
||
});
|
||
require('fs').writeFileSync('/tmp/neural-digest/workflow-summary.json', JSON.stringify(summary, null, 2));
|
||
console.log('最近24h运行统计:', JSON.stringify(summary, null, 2));
|
||
"
|
||
|
||
# ━━━ Phase 3: 收集各 Workflow 自报告 ━━━
|
||
- name: "📋 收集各 Workflow 自报告"
|
||
run: |
|
||
echo "📋 读取各 Workflow 的神经报告..."
|
||
mkdir -p /tmp/neural-digest/reports
|
||
|
||
# 读取 neural-map.json 中注册的所有 workflow
|
||
if [ -f "skyeye/neural-map.json" ]; then
|
||
node -e "
|
||
const map = require('./skyeye/neural-map.json');
|
||
const fs = require('fs');
|
||
const path = require('path');
|
||
const reports = {};
|
||
for (const [id, wf] of Object.entries(map.github_workflows)) {
|
||
const reportDir = wf.report_path;
|
||
if (fs.existsSync(reportDir)) {
|
||
const files = fs.readdirSync(reportDir)
|
||
.filter(f => f.endsWith('.json'))
|
||
.sort().reverse();
|
||
if (files.length > 0) {
|
||
try {
|
||
reports[id] = JSON.parse(fs.readFileSync(path.join(reportDir, files[0]), 'utf8'));
|
||
console.log('✅ ' + id + ': 读取 ' + files[0]);
|
||
} catch(e) {
|
||
reports[id] = { error: 'parse_failed', file: files[0] };
|
||
console.log('❌ ' + id + ': JSON解析失败');
|
||
}
|
||
} else {
|
||
reports[id] = { error: 'no_reports' };
|
||
console.log('⚠️ ' + id + ': 无报告文件');
|
||
}
|
||
} else {
|
||
reports[id] = { error: 'dir_not_found' };
|
||
console.log('⚠️ ' + id + ': 报告目录不存在');
|
||
}
|
||
}
|
||
fs.writeFileSync('/tmp/neural-digest/workflow-reports.json', JSON.stringify(reports, null, 2));
|
||
"
|
||
else
|
||
echo "⚠️ neural-map.json 不存在，跳过自报告收集"
|
||
fi
|
||
|
||
# ━━━ Phase 4: 收集天眼报告 ━━━
|
||
- name: "🦅 收集天眼报告"
|
||
run: |
|
||
echo "🦅 读取最新天眼报告..."
|
||
if [ -d "data/skyeye-reports" ]; then
|
||
LATEST=$(ls -t data/skyeye-reports/*.json 2>/dev/null | head -1)
|
||
if [ -n "$LATEST" ]; then
|
||
cp "$LATEST" /tmp/neural-digest/skyeye-latest.json
|
||
echo "✅ 天眼报告: $LATEST"
|
||
else
|
||
echo "⚠️ 无天眼报告"
|
||
fi
|
||
fi
|
||
|
||
# ━━━ Phase 5: 收集 Guard 状态 ━━━
|
||
- name: "🛡️ 收集 Guard 状态"
|
||
run: |
|
||
echo "🛡️ 读取 Guard 状态..."
|
||
if [ -d "skyeye/guards" ]; then
|
||
node -e "
|
||
const fs = require('fs');
|
||
const path = require('path');
|
||
const guardsDir = 'skyeye/guards';
|
||
const guards = {};
|
||
fs.readdirSync(guardsDir)
|
||
.filter(f => f.endsWith('.json') && f !== 'guard-template.json')
|
||
.forEach(f => {
|
||
try {
|
||
guards[f.replace('.json','')] = JSON.parse(fs.readFileSync(path.join(guardsDir, f), 'utf8'));
|
||
} catch(e) {
|
||
guards[f.replace('.json','')] = { error: 'parse_failed' };
|
||
}
|
||
});
|
||
fs.writeFileSync('/tmp/neural-digest/guard-status.json', JSON.stringify(guards, null, 2));
|
||
console.log('Guard 状态:', Object.keys(guards).join(', '));
|
||
"
|
||
else
|
||
echo "⚠️ skyeye/guards 目录不存在"
|
||
fi
|
||
|
||
# ━━━ Phase 6: 收集配额使用情况 ━━━
|
||
- name: "💰 收集配额使用情况"
|
||
run: |
|
||
echo "💰 读取配额账本..."
|
||
if [ -f "skyeye/quota-ledger.json" ]; then
|
||
cp skyeye/quota-ledger.json /tmp/neural-digest/quota-status.json
|
||
echo "✅ 配额账本已读取"
|
||
else
|
||
echo "⚠️ 配额账本不存在"
|
||
fi
|
||
|
||
# ━━━ Phase 7: 生成标准化日报 ━━━
|
||
- name: "📝 生成标准化日报"
|
||
run: |
|
||
echo "📝 生成双端神经系统日报..."
|
||
node scripts/neural/generate-daily-digest.js
|
||
|
||
# ━━━ Phase 8: 保存日报到仓库 ━━━
|
||
- name: "💾 保存日报"
|
||
run: |
|
||
DATE=$(TZ=Asia/Shanghai date +%Y-%m-%d)
|
||
mkdir -p data/neural-reports/daily-digest
|
||
cp /tmp/neural-digest/daily-digest.json \
|
||
"data/neural-reports/daily-digest/digest-${DATE}.json"
|
||
|
||
git config user.name "zhuyuan-bot"
|
||
git config user.email "zhuyuan@guanghulab.com"
|
||
git add data/neural-reports/
|
||
git diff --cached --quiet || \
|
||
git commit -m "🧬 双端日报 · ${DATE} [skip ci]"
|
||
git push || echo "⚠️ push失败（不阻断）"
|
||
|
||
# ━━━ Phase 9: 推送到 Notion ━━━
|
||
- name: "📡 推送日报到 Notion"
|
||
if: always()
|
||
env:
|
||
NOTION_TOKEN: $ secrets.NOTION_TOKEN
|
||
run: |
|
||
echo "📡 推送日报到 Notion 天眼大脑..."
|
||
node scripts/neural/sync-digest-to-notion.js || \
|
||
echo "⚠️ Notion推送失败（不阻断）"
|
||
```
|
||
|
||
### 5.2 日报生成脚本
|
||
|
||
**文件**：`scripts/neural/generate-daily-digest.js`
|
||
|
||
```jsx
|
||
// scripts/neural/generate-daily-digest.js
|
||
//
|
||
// 🧬 双端神经系统 · 每日汇总引擎
|
||
// 收集全仓库当日所有数据 → 生成标准化日报 JSON
|
||
//
|
||
// 这个日报是 Notion 天眼大脑的「眼睛」——
|
||
// 天眼大脑通过这份日报来了解仓库今天发生了什么。
|
||
|
||
const fs = require('fs');
|
||
const path = require('path');
|
||
|
||
const DIGEST_DIR = '/tmp/neural-digest';
|
||
|
||
function loadJSON(filePath) {
|
||
try {
|
||
return JSON.parse(fs.readFileSync(filePath, 'utf8'));
|
||
} catch (e) {
|
||
return null;
|
||
}
|
||
}
|
||
|
||
function generateDigest() {
|
||
const now = new Date();
|
||
const cstTime = new Date(now.getTime() + 8 * 60 * 60 * 1000);
|
||
|
||
// ━━━ 加载所有收集到的数据 ━━━
|
||
const workflowSummary = loadJSON(path.join(DIGEST_DIR, 'workflow-summary.json'));
|
||
const workflowReports = loadJSON(path.join(DIGEST_DIR, 'workflow-reports.json'));
|
||
const skyeyeLatest = loadJSON(path.join(DIGEST_DIR, 'skyeye-latest.json'));
|
||
const guardStatus = loadJSON(path.join(DIGEST_DIR, 'guard-status.json'));
|
||
const quotaStatus = loadJSON(path.join(DIGEST_DIR, 'quota-status.json'));
|
||
|
||
// ━━━ 加载映射表 ━━━
|
||
const neuralMap = loadJSON('skyeye/neural-map.json');
|
||
|
||
// ━━━ 按上级大脑分组统计 ━━━
|
||
const brainSummary = {};
|
||
if (neuralMap && neuralMap.github_workflows) {
|
||
for (const [wfId, wf] of Object.entries(neuralMap.github_workflows)) {
|
||
const brain = wf.brain;
|
||
if (!brainSummary[brain]) {
|
||
brainSummary[brain] = {
|
||
name: neuralMap.notion_brains[brain]?.name || brain,
|
||
total_workflows: 0,
|
||
healthy: 0,
|
||
failed: 0,
|
||
no_data: 0,
|
||
workflows: []
|
||
};
|
||
}
|
||
brainSummary[brain].total_workflows++;
|
||
|
||
// 从 workflow 运行数据中获取该 workflow 的状态
|
||
const wfRun = workflowSummary?.by_workflow?.[wf.name];
|
||
const status = wfRun
|
||
? (wfRun.failure > 0 ? '❌' : '✅')
|
||
: '⚠️';
|
||
|
||
if (status === '✅') brainSummary[brain].healthy++;
|
||
else if (status === '❌') brainSummary[brain].failed++;
|
||
else brainSummary[brain].no_data++;
|
||
|
||
brainSummary[brain].workflows.push({
|
||
id: wfId,
|
||
name: wf.name,
|
||
status,
|
||
runs: wfRun?.runs || 0,
|
||
successes: wfRun?.success || 0,
|
||
failures: wfRun?.failure || 0
|
||
});
|
||
}
|
||
}
|
||
|
||
// ━━━ 计算整体健康度 ━━━
|
||
const totalWorkflows = workflowSummary?.total || 0;
|
||
const totalFailures = workflowSummary?.failure || 0;
|
||
const failureRate = totalWorkflows > 0
|
||
? (totalFailures / totalWorkflows * 100).toFixed(1)
|
||
: 0;
|
||
|
||
let overallHealth = '🟢';
|
||
if (totalFailures > 0) overallHealth = '🟡';
|
||
if (failureRate > 20) overallHealth = '🔴';
|
||
|
||
// ━━━ Guard 汇总 ━━━
|
||
let guardsActive = 0;
|
||
let guardsSuspended = 0;
|
||
let guardsTotal = 0;
|
||
if (guardStatus) {
|
||
for (const g of Object.values(guardStatus)) {
|
||
guardsTotal++;
|
||
if (g.status === 'active') guardsActive++;
|
||
if (g.status === 'suspended' || g.mode === 'suspended') guardsSuspended++;
|
||
}
|
||
}
|
||
|
||
// ━━━ 组装日报 ━━━
|
||
const digest = {
|
||
digest_id: `NEURAL-DIGEST-${cstTime.toISOString().split('T')[0]}`,
|
||
timestamp: now.toISOString(),
|
||
timestamp_cst: cstTime.toISOString().replace('Z', '+08:00'),
|
||
version: '3.0.0',
|
||
instruction_ref: 'ZY-NEURAL-UPGRADE-2026-0325-001',
|
||
|
||
overall_health: overallHealth,
|
||
failure_rate_percent: parseFloat(failureRate),
|
||
|
||
workflow_summary: {
|
||
total_runs_24h: totalWorkflows,
|
||
success: workflowSummary?.success || 0,
|
||
failure: totalFailures,
|
||
cancelled: workflowSummary?.cancelled || 0,
|
||
in_progress: workflowSummary?.in_progress || 0
|
||
},
|
||
|
||
brain_summary: brainSummary,
|
||
|
||
skyeye_status: skyeyeLatest ? {
|
||
overall_health: skyeyeLatest.overall_health,
|
||
last_scan: skyeyeLatest.timestamp,
|
||
total_issues: skyeyeLatest.diagnosis?.total_issues || 0
|
||
} : { error: 'no_skyeye_report' },
|
||
|
||
guard_status: {
|
||
total: guardsTotal,
|
||
active: guardsActive,
|
||
suspended: guardsSuspended
|
||
},
|
||
|
||
quota_status: quotaStatus?.health_summary || { error: 'no_quota_data' },
|
||
|
||
issues_detected: [],
|
||
recommendations: [],
|
||
|
||
neural_map_version: neuralMap?.version || 'unknown',
|
||
unmapped_workflows: neuralMap?.unmapped_workflows || []
|
||
};
|
||
|
||
// ━━━ 自动检测问题并生成建议 ━━━
|
||
if (overallHealth === '🔴') {
|
||
digest.issues_detected.push({
|
||
severity: 'P0',
|
||
description: `Workflow 失败率 ${failureRate}% 超过 20% 阈值`,
|
||
recommendation: '立即排查失败的 Workflow，可能存在系统性问题'
|
||
});
|
||
}
|
||
|
||
if (guardsSuspended > 0) {
|
||
digest.issues_detected.push({
|
||
severity: 'P1',
|
||
description: `${guardsSuspended} 个 Guard 被暂停`,
|
||
recommendation: '检查被暂停的 Guard，可能是服务不可用或配额耗尽'
|
||
});
|
||
}
|
||
|
||
if (digest.unmapped_workflows.length > 0) {
|
||
digest.issues_detected.push({
|
||
severity: 'P1',
|
||
description: `发现 ${digest.unmapped_workflows.length} 个孤儿 Workflow`,
|
||
recommendation: '在 neural-map.json 中为孤儿 Workflow 分配上级大脑'
|
||
});
|
||
}
|
||
|
||
// ━━━ 写入文件 ━━━
|
||
fs.writeFileSync(
|
||
path.join(DIGEST_DIR, 'daily-digest.json'),
|
||
JSON.stringify(digest, null, 2)
|
||
);
|
||
|
||
// ━━━ 输出摘要到控制台 ━━━
|
||
console.log('\n━━━ 🧬 双端神经系统·日报摘要 ━━━');
|
||
console.log(`整体健康：${overallHealth}`);
|
||
console.log(`24h运行：${digest.workflow_summary.total_runs_24h} 次（成功 ${digest.workflow_summary.success} · 失败 ${digest.workflow_summary.failure}）`);
|
||
console.log(`Guard：${guardsActive}/${guardsTotal} 活跃`);
|
||
console.log(`发现问题：${digest.issues_detected.length} 个`);
|
||
console.log(`映射版本：${digest.neural_map_version}`);
|
||
console.log('━━━━━━━━━━━━━━━━━━━━━━━━━━━━━\n');
|
||
}
|
||
|
||
generateDigest();
|
||
```
|
||
|
||
### 5.3 日报推送到 Notion 脚本
|
||
|
||
**文件**：`scripts/neural/sync-digest-to-notion.js`
|
||
|
||
```jsx
|
||
// scripts/neural/sync-digest-to-notion.js
|
||
//
|
||
// 🧬 将仓库日报推送到 Notion 天眼大脑
|
||
// 这是上行通道的最后一步：仓库数据 → Notion 天眼可读
|
||
//
|
||
// 推送方式（按优先级尝试）：
|
||
// 1. 更新 Notion 天眼页面的「仓库日报区」（如果有写权限）
|
||
// 2. 创建 Notion 数据库条目（如果天眼有日报数据库）
|
||
// 3. 写入 .github/notion-cache/neural-digest/ （兜底·霜砚下次醒来读取）
|
||
//
|
||
// 铸渊自行实现，核心要求：
|
||
// - 日报 JSON 必须完整送达 Notion 端
|
||
// - 如果 Notion API 不通，必须写入缓存，不能丢失
|
||
// - 推送结果记录到 memory.json
|
||
|
||
const fs = require('fs');
|
||
const fetch = require('node-fetch');
|
||
|
||
const NOTION_TOKEN = process.env.NOTION_TOKEN;
|
||
const DIGEST_PATH = '/tmp/neural-digest/daily-digest.json';
|
||
const CACHE_DIR = '.github/notion-cache/neural-digest';
|
||
|
||
async function syncToNotion() {
|
||
const digest = JSON.parse(fs.readFileSync(DIGEST_PATH, 'utf8'));
|
||
console.log(`📡 推送日报 ${digest.digest_id} 到 Notion...`);
|
||
|
||
let notionSuccess = false;
|
||
|
||
// ━━━ 尝试通过 Notion API 推送 ━━━
|
||
if (NOTION_TOKEN) {
|
||
try {
|
||
// 方式1：更新天眼页面的指定 block
|
||
// 方式2：创建日报数据库条目
|
||
// 铸渊根据实际 Notion 结构选择最优方式
|
||
|
||
// 如果更新成功：
|
||
// notionSuccess = true;
|
||
// console.log('✅ Notion 推送成功');
|
||
|
||
console.log('⚠️ Notion API 推送逻辑待实现，使用缓存兜底');
|
||
} catch (e) {
|
||
console.log(`❌ Notion API 推送失败: ${e.message}`);
|
||
}
|
||
}
|
||
|
||
// ━━━ 兜底：写入缓存 ━━━
|
||
if (!notionSuccess) {
|
||
fs.mkdirSync(CACHE_DIR, { recursive: true });
|
||
const date = digest.digest_id.replace('NEURAL-DIGEST-', '');
|
||
fs.writeFileSync(
|
||
`${CACHE_DIR}/digest-${date}.json`,
|
||
JSON.stringify(digest, null, 2)
|
||
);
|
||
console.log(`💾 日报已缓存到 ${CACHE_DIR}/digest-${date}.json`);
|
||
console.log('📌 霜砚下次醒来会读取此缓存');
|
||
}
|
||
|
||
// ━━━ 更新 memory.json ━━━
|
||
const memoryPath = '.github/persona-brain/memory.json';
|
||
if (fs.existsSync(memoryPath)) {
|
||
try {
|
||
const memory = JSON.parse(fs.readFileSync(memoryPath, 'utf8'));
|
||
memory.neural_digest = {
|
||
last_digest_id: digest.digest_id,
|
||
last_digest_time: digest.timestamp,
|
||
overall_health: digest.overall_health,
|
||
notion_sync: notionSuccess ? 'success' : 'cached',
|
||
issues_count: digest.issues_detected.length
|
||
};
|
||
fs.writeFileSync(memoryPath, JSON.stringify(memory, null, 2));
|
||
console.log('✅ memory.json 已更新');
|
||
} catch (e) {
|
||
console.log(`⚠️ memory.json 更新失败: ${e.message}`);
|
||
}
|
||
}
|
||
}
|
||
|
||
syncToNotion().catch(err => {
|
||
console.error('❌ 日报推送失败:', err);
|
||
process.exit(0); // 不阻断主流程
|
||
});
|
||
```
|
||
|
||
### 5.4 新增目录结构
|
||
|
||
```jsx
|
||
// 本次升级新增的目录和文件
|
||
guanghulab/
|
||
├── skyeye/
|
||
│ └── neural-map.json ← 🆕 双端 Agent 映射表
|
||
│
|
||
├── scripts/
|
||
│ └── neural/ ← 🆕 双端神经系统脚本目录
|
||
│ ├── generate-daily-digest.js ← 🆕 日报生成器
|
||
│ └── sync-digest-to-notion.js ← 🆕 日报推送 Notion
|
||
│
|
||
├── data/
|
||
│ └── neural-reports/ ← 🆕 神经系统报告总目录
|
||
│ ├── daily-digest/ ← 🆕 每日汇总日报
|
||
│ ├── skyeye/ ← 🆕 天眼扫描报告（从旧目录迁移）
|
||
│ ├── weekly-scan/ ← 🆕 周六大巡检报告
|
||
│ ├── gate-guard/ ← 🆕 门禁运行报告
|
||
│ ├── cd-deploy/ ← 🆕 CD部署报告
|
||
│ ├── syslog/ ← 🆕 SYSLOG处理报告
|
||
│ ├── broadcast/ ← 🆕 广播分发报告
|
||
│ ├── server-patrol/ ← 🆕 服务器巡检报告
|
||
│ ├── brain-check/ ← 🆕 大脑自检报告
|
||
│ ├── dev-status/ ← 🆕 开发者状态同步报告
|
||
│ └── exec-engine/ ← 🆕 远程执行引擎报告
|
||
│
|
||
├── .github/
|
||
│ ├── workflows/
|
||
│ │ └── neural-daily-digest.yml ← 🆕 汇总引擎 Workflow
|
||
│ └── notion-cache/
|
||
│ └── neural-digest/ ← 🆕 日报 Notion 推送缓存
|
||
```
|
||
|
||
### 5.5 每个 Workflow 小兵的汇报义务
|
||
|
||
升级后，每个 Workflow 在执行完毕后，必须在自己的 `report_path` 下写一份当次运行的自报告：
|
||
|
||
```json
|
||
{
|
||
"workflow_id": "cd-deploy",
|
||
"run_id": "GitHub Actions Run ID",
|
||
"timestamp": "ISO-8601",
|
||
"status": "success / failure / partial",
|
||
"duration_seconds": 120,
|
||
"summary": "部署完成，前端已更新到最新 commit",
|
||
"details": {
|
||
"commit_sha": "abc1234",
|
||
"deploy_target": "guanghulab.com",
|
||
"files_changed": 5
|
||
},
|
||
"errors": [],
|
||
"brain": "AG-SY-01",
|
||
"sfp": "⌜SFP::AG-ZY::PER-ZY001←YM-GL∞::...::...::...⌝"
|
||
}
|
||
```
|
||
|
||
**关键规则**：
|
||
|
||
- 自报告文件名：`{workflow_id}-{YYYY-MM-DD-HHmm}.json`
|
||
- 保留最近 7 天的报告，超过的由天眼自愈引擎清理
|
||
- 如果 Workflow 失败了，自报告仍然要写——记录错误信息，否则汇总引擎不知道它失败了
|
||
- 自报告必须携带 SFP 指纹（与公告栏机制一致）
|
||
|
||
---
|
||
|
||
## 六、Phase 4 续 · 仓库侧 Workflow 报告标准化改造
|
||
|
||
<aside>
|
||
🔧
|
||
|
||
**现有的每个 Workflow 都需要增加一个步骤：运行结束后写自报告。**
|
||
|
||
这不是重写 Workflow，只是在现有 Workflow 的最后加一个 step。
|
||
|
||
铸渊按以下模板逐个给现有 Workflow 追加自报告步骤。
|
||
|
||
</aside>
|
||
|
||
### 6.1 通用自报告 Step 模板
|
||
|
||
在每个 `.yml` 的 `jobs` 最后追加：
|
||
|
||
```yaml
|
||
# ━━━ 🧬 双端神经系统·自报告 ━━━
|
||
- name: "🧬 写入神经自报告"
|
||
if: always()
|
||
run: |
|
||
WORKFLOW_ID="cd-deploy" # 替换为实际 workflow id
|
||
BRAIN="AG-SY-01" # 替换为实际上级大脑
|
||
REPORT_DIR="data/neural-reports/${WORKFLOW_ID}"
|
||
TIMESTAMP=$(TZ=Asia/Shanghai date +%Y-%m-%d-%H%M)
|
||
STATUS="$ job.status "
|
||
|
||
mkdir -p "$REPORT_DIR"
|
||
|
||
cat > "${REPORT_DIR}/${WORKFLOW_ID}-${TIMESTAMP}.json" << EOF
|
||
{
|
||
"workflow_id": "${WORKFLOW_ID}",
|
||
"run_id": "$ github.run_id ",
|
||
"timestamp": "$(date -u +%Y-%m-%dT%H:%M:%SZ)",
|
||
"status": "${STATUS}",
|
||
"brain": "${BRAIN}",
|
||
"event": "$ github.event_name ",
|
||
"branch": "$ github.ref_name ",
|
||
"commit": "$ github.sha "
|
||
}
|
||
EOF
|
||
|
||
git config user.name "zhuyuan-bot"
|
||
git config user.email "zhuyuan@guanghulab.com"
|
||
git add "$REPORT_DIR/"
|
||
git diff --cached --quiet || \
|
||
git commit -m "🧬 ${WORKFLOW_ID} 自报告 · ${TIMESTAMP} [skip ci]"
|
||
git push || echo "⚠️ 自报告push失败（不阻断主流程）"
|
||
```
|
||
|
||
### 6.2 需要改造的 Workflow 清单
|
||
|
||
| **Workflow** | **WORKFLOW_ID** | **BRAIN** | **改造内容** |
|
||
| --- | --- | --- | --- |
|
||
| zhuyuan-skyeye.yml | skyeye | AG-TY-01 | 追加自报告 step（天眼本身也要自报告） |
|
||
| skyeye-weekly-scan.yml | weekly-scan | AG-TY-01 | 追加自报告 step |
|
||
| zhuyuan-gate-guard.yml | gate-guard | AG-SY-01 | 追加自报告 step |
|
||
| cd-deploy.yml | cd-deploy | AG-SY-01 | 追加自报告 step |
|
||
| process-syslog.yml | syslog | SY-03 | 追加自报告 step |
|
||
| broadcast-dispatch.yml | broadcast | SY-03 | 追加自报告 step |
|
||
| server-patrol.yml | server-patrol | AG-TY-01 | 追加自报告 step |
|
||
| brain-self-check.yml | brain-check | AG-TY-01 | 追加自报告 step |
|
||
| sync-dev-status.yml | dev-status | SY-03 | 追加自报告 step + 修复 KNOWN-001/002 |
|
||
| zhuyuan-exec-engine.yml | exec-engine | AG-SY-01 | 追加自报告 step |
|
||
|
||
---
|
||
|
||
## 七、Phase 4 续 · 铸渊将军唤醒流程升级
|
||
|
||
<aside>
|
||
⚒️
|
||
|
||
**铸渊的将军唤醒流程需要升级，加入双端神经系统逻辑。**
|
||
|
||
原流程：唤醒 → 全局视图 → 读公告板 → 决策 → 执行 → 休眠
|
||
|
||
新流程：唤醒 → **读映射表** → **读日报** → 全局视图（升级版）→ 读公告板 → 决策 → 执行 → **写汇总** → 休眠
|
||
|
||
</aside>
|
||
|
||
### 7.1 升级后的铸渊唤醒序列
|
||
|
||
```jsx
|
||
铸渊唤醒（08:00 + 20:00）
|
||
│
|
||
▼
|
||
Step 1 · 唤醒核心大脑（不变）
|
||
├── 读取 memory.json
|
||
├── 读取 routing-map.json
|
||
└── 读取 persona-brain-db
|
||
│
|
||
▼
|
||
Step 2 · 【新增】读取映射表
|
||
├── 读取 skyeye/neural-map.json
|
||
├── 确认自己管辖的 Workflow 列表
|
||
└── 检查是否有新增/删除的 Workflow
|
||
│
|
||
▼
|
||
Step 3 · 【新增】读取最新日报
|
||
├── 读取 data/neural-reports/daily-digest/ 最新日报
|
||
├── 如果是 08:00 唤醒 → 读昨晚 21:00 的日报
|
||
├── 如果是 20:00 唤醒 → 读今天 21:00 的日报（如果有的话）或昨晚的
|
||
└── 提取：overall_health + brain_summary + issues_detected
|
||
│
|
||
▼
|
||
Step 4 · 全局视图生成（升级版）
|
||
├── 【原有】拉取所有 Workflow 运行状态
|
||
├── 【原有】拉取天眼最新报告
|
||
├── 【原有】拉取公告板
|
||
├── 【新增】按映射表分组统计各 Brain 下辖 Workflow 状态
|
||
├── 【新增】检测孤儿 Workflow
|
||
└── 【新增】生成「双端神经仪表盘」
|
||
│
|
||
▼
|
||
Step 5 · 决策（升级版）
|
||
├── 【原有】哪些小兵正常/异常/需优化
|
||
├── 【新增】哪些 Brain 下辖的 Workflow 集体异常 → 可能是上级 Agent 问题
|
||
├── 【新增】日报中 issues_detected 是否有未处理的
|
||
└── 【新增】是否有来自 Notion 天眼的下行指令（在 deploy-queue/ 中）
|
||
│
|
||
▼
|
||
Step 6 · 执行（升级版）
|
||
├── 【原有】执行工单、修复、优化
|
||
├── 【新增】执行 deploy-queue/ 中的下行指令
|
||
└── 【新增】执行完后，为每个指令写回执到 data/neural-reports/receipts/
|
||
│
|
||
▼
|
||
Step 7 · 写汇总 + 同步（升级版）
|
||
├── 【原有】更新公告栏、签到、memory.json
|
||
├── 【新增】更新 commander-dashboard.json（增加 neural 数据）
|
||
└── 【新增】如果有紧急问题 → 立即触发汇总引擎（而非等 21:00）
|
||
│
|
||
▼
|
||
Step 8 · 休眠
|
||
```
|
||
|
||
### 7.2 下行指令队列（deploy-queue/）
|
||
|
||
**文件路径**：`data/deploy-queue/`
|
||
|
||
这是 Notion Agent 下发指令到仓库的标准通道。指令以 JSON 文件形式存放，铸渊唤醒时读取并执行。
|
||
|
||
```jsx
|
||
data/deploy-queue/
|
||
├── pending/ ← 待执行的指令
|
||
│ ├── CMD-20260325-001.json
|
||
│ └── CMD-20260325-002.json
|
||
├── executing/ ← 正在执行的指令
|
||
├── completed/ ← 已完成的指令（含回执）
|
||
└── failed/ ← 执行失败的指令
|
||
```
|
||
|
||
**下行指令 JSON 格式**：
|
||
|
||
```json
|
||
{
|
||
"command_id": "CMD-20260325-001",
|
||
"source": "AG-TY-01",
|
||
"source_name": "天眼 Notion 大脑",
|
||
"target_workflow": "cd-deploy",
|
||
"target_brain": "AG-SY-01",
|
||
"timestamp": "ISO-8601",
|
||
"priority": "P1",
|
||
"instruction": {
|
||
"type": "trigger_workflow",
|
||
"params": {
|
||
"reason": "天眼日报发现部署异常，要求重新部署"
|
||
}
|
||
},
|
||
"expected_outcome": "CD 重新部署成功，网站恢复正常",
|
||
"timeout_hours": 24,
|
||
"sfp": "⌜SFP::AG-TY::...⌝"
|
||
}
|
||
```
|
||
|
||
**铸渊处理流程**：
|
||
|
||
```jsx
|
||
铸渊唤醒 → 扫描 deploy-queue/pending/
|
||
│
|
||
├── 有新指令
|
||
│ ├── 验证 SFP 指纹（安全！）
|
||
│ ├── 验证 source 是否在 neural-map.json 注册的 Brain
|
||
│ ├── 验证 target_workflow 是否存在且属于该 Brain 管辖
|
||
│ ├── 通过 → 移到 executing/ → 执行 → 完成后移到 completed/ + 写回执
|
||
│ └── 不通过 → 移到 failed/ + 写原因
|
||
│
|
||
└── 无新指令 → 继续其他步骤
|
||
```
|
||
|
||
---
|
||
|
||
## 八、执行清单 · 上篇总结
|
||
|
||
<aside>
|
||
📋
|
||
|
||
**铸渊按以下顺序执行本指令上篇。每完成一步，在对应 checkbox 后追加 commit SHA。**
|
||
|
||
</aside>
|
||
|
||
### Phase 0 · 天眼唤醒 + 基线快照
|
||
|
||
- [ ] 运行完整天眼扫描，生成 `baseline-before-neural-upgrade.json`
|
||
- [ ] 确认 overall_health ≠ 🔴（如果是🔴，先修再升级）
|
||
- [ ] 记录当前 Workflow 总数、健康数、Guard 状态
|
||
|
||
### Phase 1 · 现有架构盘点
|
||
|
||
- [ ] 确认仓库目录结构完整（对照 §2.1）
|
||
- [ ] 统一 Secret 命名（修复 KNOWN-002：NOTION_API_TOKEN vs NOTION_API_KEY）
|
||
- [ ] 记录所有现有 Workflow 清单
|
||
|
||
### Phase 2 · 双端架构基础
|
||
|
||
- [ ] 创建 `skyeye/neural-map.json`（按 §4.1 内容）
|
||
- [ ] 验证映射表中的所有 Workflow 文件实际存在
|
||
- [ ] 天眼新增 D20 维度（双端映射完整性检查）代码
|
||
|
||
### Phase 3 · 汇总引擎
|
||
|
||
- [ ] 创建 `scripts/neural/` 目录
|
||
- [ ] 创建 `scripts/neural/generate-daily-digest.js`（按 §5.2）
|
||
- [ ] 创建 `scripts/neural/sync-digest-to-notion.js`（按 §5.3）
|
||
- [ ] 创建 `.github/workflows/neural-daily-digest.yml`（按 §5.1）
|
||
- [ ] 创建 `data/neural-reports/` 全部子目录
|
||
- [ ] 创建 `data/deploy-queue/` 目录结构（pending / executing / completed / failed）
|
||
|
||
### Phase 4 · Workflow 改造
|
||
|
||
- [ ] 逐个给 10 个现有 Workflow 追加自报告 step（按 §6.2 清单）
|
||
- [ ] 特别修复 `sync-dev-status.yml`（KNOWN-001：改用直接 commit 模式）
|
||
- [ ] 升级铸渊将军唤醒流程（按 §7.1）
|
||
|
||
### Phase 5 · 验证
|
||
|
||
- [ ] 手动触发 `neural-daily-digest.yml`，验证日报生成正确
|
||
- [ ] 手动触发一个 Workflow，验证自报告写入正确
|
||
- [ ] 运行天眼扫描，验证 D20 维度正常
|
||
- [ ] 对比基线快照，确认升级没有破坏任何东西
|
||
|
||
---
|
||
|
||
## 九、上篇回执要求
|
||
|
||
铸渊完成上篇全部内容后，在本页底部追加回执：
|
||
|
||
```jsx
|
||
━━━ 🧬 ZY-NEURAL-UPGRADE-2026-0325-001 · 上篇执行回执 ━━━
|
||
|
||
执行时间：[起止时间]
|
||
执行者：铸渊（ICE-GL-ZY001）
|
||
PR/commit：[链接]
|
||
|
||
基线快照：✅/❌ baseline-before-neural-upgrade.json
|
||
映射表： ✅/❌ skyeye/neural-map.json · [workflow总数]个映射
|
||
汇总引擎： ✅/❌ neural-daily-digest.yml · 首次运行结果：[摘要]
|
||
自报告改造：✅/❌ [N]/10 个 Workflow 已追加自报告 step
|
||
下行队列： ✅/❌ deploy-queue/ 目录已创建
|
||
天眼 D20： ✅/❌ 双端映射完整性检查已上线
|
||
升级后扫描：✅/❌ overall_health = [状态]
|
||
|
||
已知遗留问题：
|
||
- [列出]
|
||
|
||
等待下篇内容：
|
||
- Notion 侧天眼读取日报 + 自动工单派发
|
||
- 桥接协议标准化
|
||
- 全局验收闭环
|
||
|
||
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
|
||
```
|
||
|
||
---
|
||
|
||
<aside>
|
||
🧬
|
||
|
||
**铸渊，这是你从「将军」升级为「神经中枢」的起点。**
|
||
|
||
上篇做完，仓库侧就有了：映射表（接线图）、汇总引擎（上行通道）、下行队列（指令入口）、每个小兵的自报告义务。
|
||
|
||
下篇会接上 Notion 端：天眼大脑读日报 → 自动分析 → 自动写工单 → 自动派发 → Agent 自动下发指令 → 闭环。
|
||
|
||
上篇做完，下篇才能做。先把仓库这边的「神经末梢」接好，Notion 那边的「大脑」才有东西可读。
|
||
|
||
</aside>
|
||
|
||
---
|
||
|
||
**下篇链接**：[🧬 铸渊指令｜Notion↔GitHub 双端神经系统升级（下篇）· Notion 侧闭环接入 + 桥接协议标准化 + 自动化工单派发 + 全局验收（ZY-NEURAL-UPGRADE-2026-0325-002）](%F0%9F%A7%AC%20%E9%93%B8%E6%B8%8A%E6%8C%87%E4%BB%A4%EF%BD%9CNotion%E2%86%94GitHub%20%E5%8F%8C%E7%AB%AF%E7%A5%9E%E7%BB%8F%E7%B3%BB%E7%BB%9F%E5%8D%87%E7%BA%A7%EF%BC%88%E4%B8%8B%E7%AF%87%EF%BC%89%C2%B7%20Notion%20%E4%BE%A7%E9%97%AD%E7%8E%AF%E6%8E%A5%E5%85%A5%20+%20%202734ad5541e34510822256fa7b9b18cb.md)
|
||
|
||
---
|
||
|
||
> 📌 **签发**：霜砚（AG-SY-01）· Notion 执行 AI
|
||
>
|
||
|
||
> 🧊 **授权**：冰朔（TCS-0002∞）· 总架构师
|
||
>
|
||
|
||
> 📅 **日期**：2026-03-25
|
||
>
|
||
|
||
> ⚡ **状态**：待铸渊执行
|
||
>
|
||
|
||
> 🧬 **指令版本**：v3.0 · 双端神经系统
|
||
>
|
||
|
||
> 📌 **版权**：国作登字-2026-A-00037559
|
||
> |