BossAgents Main Platform Cognitive Map

This document is a structured cognitive overview of the C:\www\wwwroot\bossagents project.

BossAgents is the "Left Help Right Arm" main platform (port 3006), which interfaces with cms/workbuddy-claw-wechat-publisher (port 8088): the former is the digital employee's "brain + spine", while the latter is the content digital employee's "mouth + frontend".

Source: Based on actual reading of source code (server.js, server/digital-staff/, server/boss-scheduler/, server/core/capability-runtime.js, server/boss-scheduler/profiles/local.yaml, src/router/index.js, etc.) and recent git changes.

I. Project Positioning

One-sentence Positioning

"Enable every manufacturing enterprise to have its own digital staff" — An AI-driven enterprise PLM platform for the manufacturing industry.

Core Narrative (Six Value Dimensions from handleHome on the homepage):

1. Asset Value Quantification (860k Assets · 101.5M Valuation)
2. Relationship Network Discovery (730k Relationships · Process→Operation→Step→Technical Parameters)
3. Quality Health Check & Diagnosis (One-click Repair)
4. AI Intelligent Insights (Trend Identification + Report Generation)
5. Digital Staff Collaboration (11 Active Staff · 4300+ n8n Workflows)
6. Business Process Traceability (110k Traceable Process Documents)

II. Code Structure

bossagents/
├─ server.js                # ★★★ 5269 行主服务入口（API 路由全在这）
├─ app/                     # Express 风格的 router 模块（boss/dispatch/feishu 等已迁移至此）
│   ├─ router.js
│   ├─ middleware.js
│   └
│   └  routes/  boss.js  capability.js  dispatch.js  feishu.js  messages.js
├─ core/                    # SCSAI 集成层（Vue/TS）
│   ├─ SCSAIAuthenticator.js / SCSAICoreWrapper.js / SCSAIDataService.js
│   ├─ SCSAIDependencyLoader/Manager/Resolver.js
│   ├─ SCSAIFileLoader.js / SCSAIPathMapper.js / SCSAIVuePlugin.js
│   └  RealSCSAIWrapper.js / GlobalObjectsPatch.js / Logger.js / SCSAI-config.js
│   └  auth/  EnhancedUnifiedAuthManager.ts / UnifiedAuthManager.ts
│   └  external/  XmlCompatibility.js
│   └  middleware/  UnifiedAuthMiddleware.ts
│   └  router/  UnifiedRouterManager.ts
├─ server/                  # ★★★ 后端业务模块
│   ├─ digital-staff/       # Digital Staff"大脑"（详见下文）
│   ├─ boss-scheduler/      # Digital Staff"脊柱"（详见下文）
│   ├─ core/                # 业务能力引擎（详见下文）
│   ├─ routes/              # 路由模块（digital-staff-routes.js 等）
│   ├─ agents/              # 决策智能体（comparison/decision/inventory-alert/quotation-parser）
│   ├─ channels/            # 渠道适配（openclaw/wechat）
│   ├─ services/            # 服务层（audit-log 等）
│   ├─ utils/               # 工具（SCSAI-client / i18n / logger）
│   ├─ shared/              # 共享（i18n）
│   ├─ config/              # 配置文件（models / item-type-registry / extraction-prompts / domain-templates）
│   ├─ data/                # SQLite DB + JSON 数据（bossagents.db / aml.db / digital-staff-*.json）
│   └  __tests__/  _deprecated/
├─ src/                     # Vue3 前端
│   ├─ router/index.js      # 580 行路由（60+ 页面）
│   ├─ views/               # 页面组件（60+ 个 .vue）
│   └  components/ composables/ services/ stores/ i18n/ utils/
├─ MTClaw/                  # MTClaw 思想加速子项目（OpenClaw 插件 + Python skills）
├─ SCIOT/                   # SCIOT 工业数据（SOLIDCAD ERP）
├─ bossagents-miniapp/      # 微信小程序前端
├─ config.yaml              # 服务端配置
├─ .env / .env.example      # 环境变量
├─ DATA_ASSET_PLATFORM_DESIGN.md / SYSTEM_STATUS_REPORT.md / VALUE_DEMO_PLAN.md  # 设计文档
└─ REFACTOR_PLAN_V2/V3/V4.md  # 重构计划

三、数字员工三大层（最关键，记牢）

BossAgents 的数字员工是 三层架构：大脑 → 脊柱 → 能力。这是和 workbuddy 那种"单个 handler 一个员工"完全不同的体系化设计。

┌──────────────────────────────────────────────────────────────┐
│              L3 业务员工层（YAML Profile 定义）               │
│   server/boss-scheduler/profiles/local.yaml                  │
│   30+ 个员工：DS-BOSS-001 ~ DS-NEWS-001                     │
│   每个员工：id/name/cron/worker/capability/llm/keywords     │
└──────────────────────────────────────────────────────────────┘
                          ↓ runOnce(staffId, intent, parameters)
┌──────────────────────────────────────────────────────────────┐
│              L2 调度层（LiteScheduler）                       │
│   server/boss-scheduler/lite-scheduler.js (2323 行)          │
│   cron 触发 + 互斥锁 + 队列 + Loop/Goal + Pipeline            │
│   分发路径：Worker / Capability / Pipeline / Loop / Generic  │
└──────────────────────────────────────────────────────────────┘
                          ↓ _runCapabilityPath
┌──────────────────────────────────────────────────────────────┐
│              L1 能力层（CapabilityRuntime）                   │
│   server/core/capability-runtime.js (2012 行)                │
│   六大能力：identify / validate / repair / optimize /         │
│             compare / generate + inspect/valuate/collect/import │
│   规则引擎优先 → LLM 降级                                     │
└──────────────────────────────────────────────────────────────┘
                          ↓
┌──────────────────────────────────────────────────────────────┐
│              L0 智能层（LLMBrain + LLMRouter）                │
│   server/digital-staff/llm-brain.js（863 行）                │
│   server/digital-staff/llm-router.js（694 行）               │
│   server/digital-staff/smart-llm-router.js（1277 行）        │
│   MTClaw 思想：Worker（小模型 ~70ms）/ Solver（大模型 ~500ms） │
│   70%+ 走 Worker，7 倍加速                                    │
└──────────────────────────────────────────────────────────────┘

3.1 L0 智能层 — server/digital-staff/

| 文件 | 行数 | 职责 |

|---|---|---|

| llm-brain.js | 863 | LLMBrain class, singleton getGlobalInstance(). Configuration priority: LLM_ > DIGITAL_STAFF_LLM_ > config > default. Detects MTClaw acceleration (port 18790) |

| llm-router.js | 694 | LLMRouter + InferenceServiceDetector, Worker/Solver division of labor |

| smart-llm-router.js | 1277 | SmartLLMRouter, intelligent scheduling layer (circuit breaker / retry / cache) |

| smart-llm-bridge.js | 96 | Bridge |

| index.js | 2712 | Legacy digital employee core scheduler (cron + SCSAI + capability SDK), replaced by LiteScheduler, retained for compatibility |

| loader.js | 13 | Loader → Always returns boss-scheduler (unified architecture) |

| staff-manager.js | 555 | StaffManager — Digital employee configuration CRUD (creates tables digital_staff / staff_tasks / staff_execution_logs) |

| task-board.js | 321 | Task board |

| config-manager.js | 249 | Configuration management |

| keyword-matcher.js | 148 | Keyword matcher (legacy, replaced by staff-router) |

| conversation-engine.js | 410 | Conversation engine |

| conversation-manager.js | 213 | Conversation Manager |

| correction-learner.js | 311 | Self-Evolving Learner |

| ascend-npu.js | 495 | Ascend NPU Adaptation |

| chipwise-handler.js | 506 | Chip Design-Specific Handler |

| plm-brain-handler.js | 529 | PLM Document Brain Handler |

| collaboration-scheduler.js | 276 | Collaboration Scheduler |

| collaboration-rules.yaml | — | Collaboration Rules |

3.2 L2 Scheduling Layer — server/boss-scheduler/

lite-scheduler.js (2323 lines) — LiteScheduler class, core methods:

| Method | Line | Responsibility |

|---|---|---|

| runOnce(staffId, intent, parameters) | 2096 | Main entry: execute a single staff task |

| resumeStaff(staffId, resumeId, choice) | 2243 | Resume a task paused by ConfirmNeededError |

| _runWorker(staff, intent, parameters) | 1374 | Dispatcher: Select path based on staff configuration |

| _dispatchWorkerPath | 874 | Worker path (special worker) |

| _runLoopGoalPath | 907 | Loop/Goal path |

| _runPipelinePath | 945 | Pipeline path (multi-capability series) |

| _runWorkerScriptPath | 1025 | Worker script path |

| _runCapabilityPath | 1065 | Capability path → Invoke CapabilityRuntime |

| _runDecomposedSteps | 1245 | Execute decomposed steps |

| _runGenericPath | 1349 | Generic fallback |

| _processStaffTasks | 1491 | Process task queue |

| _runLoopAction | 1521 | Loop action execution |

| _checkGoalCondition | 1808 | Goal condition check |

| _executeGoalSuccess/Failure | 1874/1918 | Goal success/failure actions |

| _sendNotification / _createAlert | 1968/1987 | Notification/Alert |

| _withStaffMutex | 768 | Staff Mutex Lock |

| _checkLoopCondition / _evaluateCondition | 609/640 | Loop Condition Evaluation |

| _getStaffContext | 670 | Staff Context |

| _detectLocalServicesAsync | 170 | Detect Local Services (Ollama/MTClaw) |

| _callMTClaw | 366 | Call MTClaw |

Supporting Files:

| File | Responsibility |

|---|---|

| index.js (226 lines) | Entry point, exports LiteScheduler |

| staff-registry.js | Loads staff definitions from YAML Profile (supports GBK decoding, disable filtering) |

| staff-router.js (520 lines) | Shared routing module — unified keyword matching / parameter extraction / Banner metadata, used by AiWorkbench (frontend) / feishu-router / DigitalStaff detail page across three ends |

| task-queue.js | SQLite queue |

| collaboration-chain.js | Chain collaboration (onComplete triggers downstream staff) |

| feishu-router.js | Feishu Channel Router |

| health-self-heal.js | Self-Healing |

| workers/ | 20 Worker Implementations (see table below) |

| profiles/ | YAML Employee Definitions (local/asset-catalog/automation-presets/health-check/mtclaw-tools/valuation) |

20 Workers (server/boss-scheduler/workers/):

| Worker File | Corresponding Employee ID | Responsibility |

|---|---|---|

| data-clerk.js | DS-BOSS-001 | Data Clerk — Sync SCSAI objects to local cache |

| system-health.js | DS-BOSS-002 | System Ops Engineer — Inspect DB/SCSAI/template/rule consistency |

| procurement.js | DS-PROC-001 | Procurement Assistant — Full process of inquiry/price comparison/order placement |

| content.js | DS-CONTENT-001 | Content Generator — Official account article/composition/de-AI-ify/publish |

| SCSAI-creator.js | DS-SCSAI-001 | SCSAI Engineer — Create business objects |

| cost-optimizer.js | DS-COST-001 | Cost Optimizer — BOM Cost Analysis |

| data-caretaker.js | DS-OPS-001 | Data Caretaker — Cleanup/Archive/Monitor |

| report-analyst.js | DS-REPORT-001 | Report Analyst — Visualization + Email |

| report-i18n.js | — | Report i18n |

| vendor-review.js | DS-VEN-001 | Vendor Steward — Performance Review |

| ecr-worker.js | DS-ECR-001 | ECR Auditor — Change Review |

| process-worker.js | DS-PROCESS-OPT-001 | Process Optimization — 16-Step Workflow |

| biz-analyst.js | DS-BIZ-001 | Business Brain — Executive Daily Brief |

| stock-manager.js | DS-STOCK-001 | Inventory Steward — Level Inspection |

| doc-worker.js | DS-DOC-001 | Document Digital Worker — Generation/QA |

| mkt-worker.js | — | Marketing |

| pm-worker.js | — | Project Management |

| equip-worker.js | — | Equipment |

| auto-loop-worker.js | DS-AUTO-LOOP-001 | Closed-Loop Engine — Content Link/Share/Self-Evolution |

| system-health.js | — | System Health |

3.3 L1 Capability Layer — server/core/capability-runtime.js (2012 lines)

The CapabilityRuntime class, which packages six basic capabilities into a unified interface, with rule engine priority → LLM fallback.

Core capability methods (10):

| Method | Line | Capability | Description |

|---|---|---|---|

| identify(context, options) | 872 | Identify | Query/identify object type and attributes |

| validate(context, options) | 835 | Validate | Audit consistency |

| repair(context, options) | 952 | Repair | Fix data quality issues |

| optimize(context, options) | 1126 | Optimize | Optimize process parameters |

| compare(context, options) | 1212 | Compare | BOM/supplier diff comparison |

| generate(context, options) | 1276 | Generate | Content/document generation |

| inspect(context, options) | 1660 | Health Check | Five-Dimensional Health Score |

| import(context, options) | 187 | Import | EXCEL/CSV/JSON/XML Import |

| valuate(context, options) | 580 | Valuation | Three-Stage Cost/Revenue/Market Valuation |

| collect(context, options) | 732 | Collection | Multi-Source Data Collection |

Statistical counters: ruleEngineHits / llmWorkerHits / llmSolverHits / totalCalls / byCapability — used to evaluate rule engine hit rate.

Core supporting modules (server/core/, 50+ modules):

• rule-engine.js / rule-engine-core.js / rule-evolution.js / rule-simulator.js / rule-workbench-service.js / rule-intent-interpreter.js — the full rule engine suite
• unified-asset-service.js — Unified Data Asset Service (core, dashboard API)
• unified-schema.js / unified-validator.js — Schema/Validation
• capability-dispatcher.js — Capability Dispatcher
• relationship-capability.js / relationship-checklist.js / relationship-resolver.js — Relationship Resolution
• health-inspector.js / health-patrol.js / health-scanner.js — Health Inspection
• asset-valuator.js / asset-catalog-service.js — Valuation/Cataloging
• object-importer.js / object-modeler.js — Object Import/Modeling
• document-generator.js / docs-converter.js / doc-chat-service.js — Document
• content-engine.js / content-publisher.js / content-audit-engine.js / content-asset-manager.js / content-calendar-manager.js — Content Engine
• marketing-service.js / marketing-loop-engine.js — Marketing
• auto-loop-service.js / automation-service.js — Automated Closed Loop
• workflow-executor.js / workflow-to-schema.js — Workflow
• intent-engine.js / intent-classifier.js / smart-intent-recognizer.js / rule-intent-interpreter.js — Intent Recognition
• correction-learner.js / evolution-engine.js — Self-Evolution
• knowledge-indexer.js / knowledge-search.js / domain-knowledge-extractor.js / product-knowledge-integrator.js — Knowledge
• llm-process-generator.js / llm-process-exporter.js / llm-process-knowledge.js / llm-process-store.js — LLM Process
• phosphorus-asset-service.js / phosphorus-kb-builder.js — Phosphorus Chemical Demo
• bom-demo-builder.js / bom-process-linker.js — BOM
• process-spec-generator.js / SCSAI-process-sync.js / SCSAI-rule-sync.js — Process Specification
• change-chain-engine.js — Change Chain
• share-service.js — Sharing
• tenant-content-adapter.js / user-asset-service.js / platform-service.js — Tenant/User/Platform
• prompt-template-manager.js — Prompt Template
• scirot-rules-bridge.js — SCIOT Rules Bridge
• staff-capability-portrait.js / staff-domain-mapper.js — Staff Capability Portrait
• industrial-data-asset-overview.js — Industrial Data Asset Overview
• context-injector.js / data-collection-service.js / data-service-api.js / execution-to-content.js / alignment-engine.js / ai-reasoning-visualizer.js / behavior-service.js / generation-task-manager.js / enterprise-rules.js — Various Auxiliary Modules

4. YAML Profile: Employee Master List (server/boss-scheduler/profiles/local.yaml)

local.yaml defines 30+ employees, categorized by type:

4.1 IT/Data Infrastructure Class

4.2 Procurement/Supply Chain

4.3 Engineering/Process/Change

4.4 Quality/Evaluation

4.5 Business/Reporting/Document/Content Category

4.6 Chip Design Category (ChipWise, 6 total)

4.7 General Office (7 items)

4.8 Loop Engineering Patterns (4 Examples)

| ID | Name | Pattern | Description |

|---|---|---|---|

| DS-PROC-LOOP-001 | XiaoZhi-Price Monitor | Monitoring Type | Triggers re-quotation if price increase exceeds 5% |

| DS-BOSS-LOOP-001 | XiaoZhi-Goal Tracker | Goal-Driven Type (/goal) | Monthly sales target of 1 million tracking |

| DS-VEN-LOOP-001 | XiaoZhi-Data Repair Agent | Self-Repairing Type | Automatically completes missing supplier fields |

| DS-STOCK-LOOP-001 | XiaoZhi-Inventory Alert Agent | Monitoring Type | Automatically generates purchase requisitions for low stock |

Loop Engineering Three Patterns:

1. Monitoring Type — cron + condition → action pipeline
2. Goal-Driven Type (goal) — verification_condition + interval + on_success/on_failure
3. Self-healing type — action pipeline (identify→repair→validate) + max_iterations + retry_delay

5. Overall server.js Routing Table (5269 lines, handleRequest at lines 1372-5008)

5.1 Digital Staff API (/api/digital-staff/*, 40+ endpoints)

Run Control:

• GET /api/digital-staff/run — Manually trigger a single staff
• POST /api/digital-staff/run — Same as above
• POST /api/digital-staff/run-internal — Internal trigger
• POST /api/digital-staff/start / stop / toggle — Start/Stop/Toggle
• POST /api/digital-staff/reload — Reload configuration

Status/Monitoring:

• GET /api/digital-staff/status — Status
• GET /api/digital-staff/health — Health
• GET /api/digital-staff/dashboard — Dashboard
• GET /api/digital-staff/runs / run-detail — Run history
• GET /api/digital-staff/logs / POST /logs/delete — Logs
• GET /api/digital-staff/logs/stream — SSE Real-time Log Stream (registerSSEClient)

Confirmation/Recovery (Human-Machine Collaboration):

• GET /api/digital-staff/pending-confirms — Pending Confirmation List
• POST /api/digital-staff/confirm — Confirm Decision (Resume after ConfirmNeededError pause)
• GET/POST /api/digital-staff/resume — Resume Execution

Configuration CRUD (server/routes/digital-staff-routes.js):

• GET /api/digital-staff/list / /:id — List/Details
• POST /api/digital-staff — Create
• PUT /api/digital-staff/:id — Update
• DELETE /api/digital-staff/:id — Delete
• POST /api/digital-staff/:id/enable — Enable
• GET /api/digital-staff/:id/rules-prompts — Rules and Prompts
• GET/POST /api/digital-staff/config — Read/Write Configuration
• GET /api/digital-staff/llm-config — LLM Configuration

Task Management:

• GET /api/digital-staff/tasks / /:id — Task List/Details
• POST /api/digital-staff/tasks — Create Task
• GET /api/digital-staff/task-stats — Task Statistics

Loop/Goal:

• GET /api/digital-staff/loops / /:id — Loop List/Details
• POST /api/digital-staff/loops — Create Loop
• PUT/DELETE /api/digital-staff/loops/:id — Update/Delete
• GET /api/digital-staff/goals / /:id — Goal Status
• POST /api/digital-staff/goals/:id/reset — Reset Goal

Matching/Import-Export:

• POST /api/digital-staff/match — Keyword Match Employee
• GET /api/digital-staff/import/formats / POST /import — Import
• POST /api/digital-staff/export / GET /export/formats — Export

MTClaw/Router:

• GET /api/digital-staff/mtclaw/stats / POST /mtclaw/test — MTClaw Statistics/Test
• GET /api/digital-staff/router/stats / POST /router/test — LLM Router Statistics/Test

5.2 Business APIs (in order of appearance in server.js)

| Prefix | Description |

|---|---|

| /api/bom/* | BOM Management (version/import/search/config/cron/object-graph/generate-nested/save-nested) — line 693 handleBomApi |

| /api/audit/* | Audit Logs |

| /api/vendor/ / /api/vendor-part/ | Vendor |

| /api/customer/* | Customer |

| /api/change/stats | Change Statistics |

| /api/messages/stream | SSE Message Stream |

| /api/miniapp/gst/* | Mini Program GST (10+ sub-paths for product/order/payment/course/anchor, etc.) |

| /api/wechat/* | WeChat Channel |

| /api/auth/ /api/admin/enterprises /api/enterprise/ | Authentication/Enterprise |

| /api/ai/agent/ /api/ai-agent/ | AI Agent |

| /api/asr/recognize | Speech Recognition |

| /api/content/* | Content Digital Employee API |

| /api/employee/* | Employee |

| /api/unified/* | Unified Assets (Core of Data Asset Platform) |

| /api/object-modeler/* | Object Modeling |

| /api/tenant/quota | Tenant Quota |

| /api/sciot/parts | SCIOT Parts |

| /api/dashboard/* | Dashboard (asset-overview / today-governance) |

| /api/SCSAI/query | SCSAI Query (SOAP Proxy) |

| /api/unified-assets/* | Data Asset Dashboard (dashboard / list / :id / relation-graph) — This is what workbuddy proxies |

Note: Some routes have been migrated to app/router.js (boss/dispatch/feishu/workbuddy/coze), with comments left in server.js marking them as MIGRATED.

VI. Frontend Routes (src/router/index.js, 60+ pages)

Digital Staff related pages:

| Path | Component | Description |

|---|---|---|

| /digital-staff | DigitalStaff.vue (3261 lines) | Digital Staff Management Console |

| /ai-workbench | AiWorkbench.vue (3259 lines) | AI Workbench (Conversation-triggered Staff) |

| /staff-capabilities | StaffCapabilities.vue | Staff Capability Profile |

| /staff-customizer | — | Staff Customization |

| /content-center | ContentCenter.vue | Content Center |

| /document-intelligence | DocumentIntelligence.vue | Document Intelligence |

| /llm-config | LLMConfig.vue | LLM Configuration |

Business pages: dashboard / product-list / private-library / bom-assistant / change-management / vendor-management / customer-management / order-management / project-management / document-center / file-business / quality-management / inventory-management / workflow-management / modeling-center / itemtype-management / system-objects / data-collection / rules-templates / feishu-config / system-log / inspection-history / etc.

Featured Pages: ChipWiseEngine (Chip Design) / PlmDocBrain (PLM Document Brain) / ProcessOptimization / ProcessSpecStudio / PpsTechFileBrowser / PpsExtractPanel / CrossIndustryValidation / DataMarketplace / DataRelationGraph / IndustrialAssetOverview / CustomizationCenter / CustomObjectBuilder / RuleCustomizer / PromptCustomizer

7. Key Configuration

7.1 Environment Variables (.env.example)

Server: PORT=3006 JWT_SECRET AUTH_SECRET

Database: DB_BACKEND=sqlite|mysql DATABASE_PATH DB_MYSQLDB_*

SCSAI PLM: SCSAI_SERVER SCSAI_DATABASE SCSAI_USERNAME SCSAI_PASSWORD SCSAI_AUTH_TYPE

LLM:

• Main Chain: LLM_API_KEY LLM_ENDPOINT LLM_MODEL LLM_PROVIDER
• DeepSeek: DEEPSEEK_*
• Ollama Local: OLLAMA_ENDPOINT=http://localhost:11434 OLLAMA_MODEL=gemma3:4b
• Worker (Small Model): LLM_WORKER_ENDPOINT/MODEL/API_KEY
• Solver (Large Model): LLM_SOLVER_ENDPOINT/MODEL/API_KEY

Email: SMTP_HOST/PORT/SECURE/USER/PASS/FROM

7.2 YAML Features Configuration (End of local.yaml)

MTClaw Acceleration:

• base_url: http://127.0.0.1:18790/v1
• completion_mode: permissive
• 40+ high-frequency actions (db_query/identify/inspect/compare/repair/validate/procurement_assistant/...)
• disabled_actions: parse_email/understand_intent/generate_text/complex_reasoning

LLM Router:

• Worker: gemma3:4b (Ollama, 5s timeout)
• Solver: deepseek-chat (30s timeout, temp 0.7, max_tokens 2000)
• default_to_solver: true (conservative strategy)
• task_rules: simple_types (db_query/calc/validate/render/send/notify/check/list/get/find) vs complex_types (analyze/plan/create/generate/reason/understand)
• simple_patterns / complex_patterns: regex matching
• chain: mtclaw(disabled) / ollama(enabled) / deepseek(enabled)
• circuit_breaker: 5 failures → open, 60s recovery
• retry: 3 attempts, exponential backoff, jitter
• cache: 1000 size, 60s TTL

VIII. Integration Relationship with workbuddy-claw-wechat-publisher

workbuddy (8088)                          bossagents (3006)
┌─────────────────────────┐               ┌─────────────────────────────────┐
│ handleBossagentsProxy   │ ──fetch──→    │ /api/unified-assets/*           │
│ /api/bossagents-proxy   │               │ /api/aml/unified/schema/:type   │
│   ?type=Part            │               │   （BossAgents 真活端点）         │
│   ?endpoint=/api/...    │               │                                 │
└─────────────────────────┘               │ 挂了 → workbuddy 降级 mock 真数据 │
                                          │                                 │
│ handleDemoCapabilityApi │               │ CapabilityRuntime 六大能力       │
│ /api/demo/capability    │ ←─内嵌AI演示─ │ （不调 bossagents，直接调 AI）   │
│ （identify/create/...） │               │                                 │
└─────────────────────────┘               │                                 │
                                          │                                 │
│ handleDemoMarketingApi  │               │                                 │
│ /api/demo/marketing     │ ←─内嵌AI演示─ │                                 │
│ （4平台营销内容）        │               │                                 │
└─────────────────────────┘               └─────────────────────────────────┘

对接点只有 1 个：/api/bossagents-proxy → ylxt.chat，10 秒超时。

Live Endpoints of BossAgents (called by workbuddy demo page):

• /api/aml/unified/schema/Part — Recognition capability (Schema API, the only live endpoint)
• /api/unified-assets/value/total — Asset value
• /api/unified-assets/dashboard — Relationship network/Process traceability
• /api/unified-assets/quality/issues — Quality inspection
• /api/unified-assets/ai-insight — AI insights
• /api/unified-assets/external-stats — Digital employee collaboration

Inactive endpoints (workbuddy fallback display): When the bossagents backend is not running or returns 500, the workbuddy frontend displays mock real data with a "⚠ BOSSAGENTS API inactive, fallback to real data display" prompt.

9. Recent Changes to Main Branch (Last 20 Commits)

583bda87 最新提交同步
bb663674 feat(软著): 7份各生成三件套（登记申请表+ManualPDF+源码PDF）
6662cdec feat(软著): sc3/sc5 源码补足至60页，7份统一60页
df522fee feat: gst-health Node直连云诊平台，移除Java代理层
5f8b9171 fix(软著): 源码超60页裁切为前30+后30共60页
ccf93004 docs(软著): 7份说明书全面扩充至>=30页
56e3ee2f feat(软著): 主文件夹补入 bossagents 主平台软著(sc1)，共7份
f516d130 feat(aml): 构建AML工业数据资产管理体系，101万+数据资产落地
2fe84d69 feat(软著): 新增软著材料生成脚本
d0b633b1 fix(postbuild): 扩展内联覆盖 store/composables/config
e517c43f fix(miniapp): 修复云开发初始化报错和生命周期钩子导入错误
953686ce docs: HICOOL MTClaw策划方案优化 + 测试框架完善
b89cf8fe feat: MTClaw Ollama本地模型调度 + 测试框架优化
17b1e53a feat: MTClaw API Key空值处理 + 商汤SenseNova配置优化
6457fb32 fix: 修复数字员工配置问题
f33f3680 docs: 数字员工全面测试报告 - 34个员工100%通过
0b2808ea 小程序优化
7628fed8 Optimize

主轴：软著材料准备（7份）+ MTClaw 加速 + 数字员工测试报告（34个员工100%通过）+ AML 资产管理。最近一周重心在 软著申请材料 和 小程序优化。

10. Potential Pitfalls and Notes

1. server.js is a single file with 5269 lines, containing a massive 3637-line handleRequest function with 71+ routing branches all crammed together — even more extreme than workbuddy's cms-server.cjs, making extension extremely painful.
2. digital-staff/index.js (2712 lines) and boss-scheduler/lite-scheduler.js (2323 lines) are dual schedulers. loader.js already points uniformly to lite-scheduler, but the old index.js still retains "compatibility references" — technical debt.
3. Duplicate YAML keyword libraries: STAFF_KEYWORDS in staff-router.js and the keywords field for each employee in local.yaml create two sets of keywords. Changing one without updating the other causes mismatches.
4. Dual local models: MTClaw on port 18790 and Ollama on port 11434. Configuration exists in both YAML and env files, with env taking precedence but YAML able to override — prone to confusion.
5. _evaluateCondition uses eval to execute condition strings from YAML — a security risk; tampering with the YAML could lead to RCE.
6. Database password hard-coded: docs-to-cms-pipeline.js (on the workbuddy side) hard-codes 'alibaba', pointing to the DB on the same machine as bossagents, causing secret leakage.
7. 7 software copyright materials (recently submitted) — If publishing or making the repository public, note that the source code in the copyright materials may contain sensitive configurations.
8. Test report for 34 employees with 100% pass rate (f33f3680) — but local.yaml only defines 30+ employees. The discrepancy between 34 and 30+ needs verification (may include Loop employees).
9. SSE /api/digital-staff/logs/stream must use registerSSEClient from server/digital-staff/index.js (same module as staffLog). Cross-module usage will cause log loss.
10. _runCapabilityPath calls CapabilityRuntime, but among the 10 capability methods in CapabilityRuntime, import/valuate/collect/inspect are not part of the "Six Core Capabilities" narrative — synchronize narrative changes with code modifications.

Appendix: Key Run Commands

# 后端（3006）
node server.js
npm run dev            # 等价

# 前端开发（5173）
npm run dev:vite

# 构建
npm run build          # vite build
npm run preview

# 测试
npm test               # jest --coverage
npm run test:watch
npm run test:integration
npm run test:frontend  # vitest run

# Lint
npm run lint
npm run lint:fix

# 数据资产看板验证
curl https://ylxt.chat/api/unified-assets/dashboard
curl https://ylxt.chat/api/unified-assets/list?asset_type=process_spec
curl "https://ylxt.chat/api/unified-assets/spec_pps0005290Bs2pbsl/relation-graph?depth=2"

# 数字员工手动触发
curl https://ylxt.chat/api/digital-staff/run?staffId=DS-BOSS-002

# 切换 Profile
BOSS_SCHEDULER_PROFILE=./server/boss-scheduler/profiles/valuation.yaml node server.js

附：与 workbuddy 的能力对照

| 维度 | workbuddy-claw-wechat-publisher | bossagents |

|---|---|---|

| Positioning | Content Digital Employee (Mouth + Face) | Full-Stack Digital Employee Platform (Brain + Spine) |

| Port | 8088 | 3006 |

| Main Language | Node.js (cjs) | Node.js + Vue3 |

| Employees | 11 (narrative) / 3 (actual run) | 34 (test passed) / 30+ (YAML defined) |

| Employee Definition | Inline handler in cms-server.cjs | YAML Profile + Worker Module |

| Scheduling | Simple timing via node-cron | LiteScheduler (cron + mutex + queue + Loop/Goal/Pipeline) |

| Capability | Direct AI calls (callAI multi‑path fallback) | CapabilityRuntime (rule engine first → LLM fallback) |

| LLM | Multi‑provider JSON config (real‑time read) | LLMBrain + LLMRouter (Worker/Solver division) |

| Database | MySQL (lvbo_type/lvbo_article) | SQLite/MySQL (digital_staff/staff_tasks/...) |

| SCSAI | None | Deep integration (SOAP proxy + AML) |

| Frontend | public/ static HTML | Vue3 60+ pages |

| Mini‑Program | None | bossagents-miniapp/ |

| Software Copyright | None | 7 applications in progress |