BossAgents User Guide

BossAgents User Guide

System Status

Service Running — Port 3006, connected to SCSAI PLM (ylxt.chat)

Digital Employees: 70 (consolidated from 87, with zero functionality loss)

Demo Mode: Click the '? Demo Mode' button in the chat header to select a scenario for one-click experience.


1. Quick Start

Open a browser and visit https://ylxt.chat. Enter natural language instructions directly in the chat input box at the bottom.

⚠️ Do NOT use @ to mention a specific digital employee. Just type directly. The system will automatically decide whether to assign the task to a specialist or have the AI assistant answer directly.

Demo Scenarios (No need to memorize, just type directly)

| What You Say (without @) | What the System Will Do |

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

| Query the impact scope of P-1001 or Help me analyze what will be affected by the change to P-1001 or Query the change impact of P-1001 | L1 Rule-based Direct Impact Analysis Tool → Reverse Where-Used Lookup → Find 3 parent assemblies (ASSY-A001/A002/A003) → Generate Impact Analysis Report (4 parts, 3 BOMs, Severity: Medium) |

Check Part data quality or Perform a BOM data health check or Data quality inspection | L1 rule-based direct data scan → Scanned 1193 Parts → Returned approximately 5500 issues (missing weight, cost, classification, etc.), categorized by severity level |

View recent governance records or View recent operation history or Display audit logs | L1 rule-based direct audit log → Displays the last 20 operation records (inspection logs, creation events, system start/stop, etc.) |

Approve ECR-2026-0005 or Approve change order ECR-2026-0005 or Pass approval for ECR-2026-0001 | L1 rule-based direct approval tool → Executes local ECR approval workflow: submit→pending_approval→approved, automatically creates corresponding ECO (Engineering Change Order) upon approval |

Submit ECR-2026-0005 or Reject ECR-2026-0002 | Submit→enters pending approval; Reject→status changes to rejected (follows local state machine) |

Release P-1001 or release part P-1001 as Released | L1 rule direct approval tool → promoteItem → call SCSAI PE_ManualRelease method (as Sciot PLM identity) → Part becomes Released (tested successfully). Attempting to release an already released part will prompt "Must be in Preliminary/Manual Change status".

✅ All three scenarios are deterministic L1 direct passes (auto_execute), not relying on LLM tool calls, demonstrating repeatable and stable operation.

✅ Approval/Release also uses L1 direct pass approve_item (bi_approve_item rule). Approval objects follow the local state machine for a closed loop (reliable); Part release uses SCSAI PE_ManualRelease for actual release (requires Sciot PLM identity demo, fixed on 2026-08-02, previous "controlled" conclusion was a permission issue).

? If you @mention a specific digital employee (e.g., @Xiaozhi-Document Validator to check Part data quality), the system first validates if that employee supports the business type in the query (Part/Document/BOM…). If not supported, the message is automatically routed to the AI assistant for handling, preventing erroneous assignment to document-type employees.


II. Backend API Overview

2.1 Impact Analysis

curl.exe -s -X POST https://ylxt.chat/api/change/analyze -H "Content-Type: application/json" -d '{\"description\":\"查询 P-1001 的影响范围\"}'

返回示例:

{
  "success": true,
  "change_type": "part",
  "target_id": "284C6CFC12B34FACA25A416B3D2CD897",
  "affectedParts": 4,
  "affectedBoms": 3,
  "severity": "Medium"
}

2.2 数据巡检

curl.exe -s -X POST https://ylxt.chat/api/ai-inspector/inspect -H "Content-Type: application/json" -d '{\"scope\":\"instance\",\"itemType\":\"Part\",\"autoFix\":false}'

2.3 审计日志

curl.exe -s https://ylxt.chat/api/audit-logs

2.4 Where-Used 反查

curl.exe -s -X POST https://ylxt.chat/api/change/where-used -H "Content-Type: application/json" -d '{\"item_id\":\"284C6CFC12B34FACA25A416B3D2CD897\",\"item_type\":\"Part\"}'

2.5 直接执行 AML 查询

curl.exe -s -X POST https://ylxt.chat/api/unified/apply-aml -H "Content-Type: application/json" -d '{\"aml\":\"<AML><Item type=\\\"Part\\\" action=\\\"get\\\" select=\\\"name,item_number\\\" maxRecords=\\\"5\\\"><id>284C6CFC12B34FACA25A416B3D2CD897</id></Item></AML>\"}'

2.6 审批 / 发布(approve_item)

走聊天界面最方便(自动识别类型/编号/动作)。也可直接调接口:

# 审批一个 ECR(自动创建对应 ECO)
curl.exe -s -X POST https://ylxt.chat/api/ai-agent/execute -H "Content-Type: application/json" -d '{\"toolName\":\"approve_item\",\"params\":{\"item_type\":\"ECR\",\"item_id\":\"ECR-2026-0005\",\"action\":\"approve\",\"comment\":\"审批通过\"}}'

approve_item 工具支持的参数:

| Parameter | Required | Description |

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

| item_type | Yes | Part / ECR / ECN / ECO / PR / Document / Vendor |

| item_id | Yes | Object number (ECR-2026-0005, P-1001) or SCSAI GUID |

| action | No | approve / reject / submit / promote / release (for ECN) / implement |

| target_state | No | Target state name (for promote/release, default is Released) |

| comment | No | Approval comment |

Behavior of various objects:

| Type | Behavior | Programmatic flow possible |

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

| ECR / ECN / ECO / PR | Follow local state machine loop (data/ecrs.json, etc.), submit → approval → automatically create ECO | ✅ Reliable |

| Part and other SCSAI objects | Follow promoteItem → call SCSAI PE_ManualRelease (as Sciot PLM identity), truly released as Released | ✅ Can be released (fixed 2026-08-02) |

Common phrases:

Submit ECR-2026-0005
审批 ECR-2026-0005
驳回 ECR-2026-0002
发布 P-1001        ← 发布为 Released(已发布的零件会提示需在 Preliminary/Manual Change 状态)

三、创建新对象

在聊天输入框直接输入:

帮我创建一个 Part,名称为"测试零件-001",单位为 EA,重量 1.5kg

系统流程:

  1. LLM parses natural language into structured JSON
  2. Call SCSAI to create Part and populate fields
  3. If LLM output contains nested relationships or item_properties, the system automatically recursively creates sub-objects (up to 5 levels deep)
  4. Automatically process required fields such as classification and numbering

4. Frequently Asked Questions

PhenomenonCauseSolution
Entered "Impact Scope/Change Impact" but triggered a regular conversationLegacy version relied on LLM, occasionally failing to invoke the toolChanged to L1 deterministic passthrough; directly input "Query the impact scope of P-1001"
"@Xiaozhi-Document Reviewer Check Part data quality" went to a document employeeMentioned an employee who does not support the Part typeDo not @ employees; directly input "Check Part data quality" (the system will automatically route to the inspection tool)
Inspection returns 0 issuesData has been auto-populatedManually clear a certain field and try again
"Release P-1001" reports "Must be in Preliminary or Manual Change status"The Part is already released (Released)Change to release other unreleased parts, such as releasing PART-1780803072387 or a newly created part
"Release" reports ItemIsLocked / State transition failedThe Part is an old version (is_current=0) or is lockedUse the current version of the part for release (create a new one and release)
"Approve ECR-2026-0005" reports "Current state is not pending approval"The ECR is already approved or not in pending_approvalBefore demo, use Submit ECR-2026-0005 to put it into pending_approval; or approve another ECR
ECO not seen after approvalECO is in data/ecos.jsonCheck View Recent Governance Records or directly check data/ecos.json
"??????" displayedConsole encoding issueData is correct, only a display issue

V. Technical Configuration

ItemValue
Server Port3006
SCSAI Serverhttps://ylxt.chat/scplm
LLM Modeldeepseek-v4-flash
Timeout Setting120s (Frontend + Server)
Nested Creation DepthMax 5 levels
Inspection Rules78 built-in + 10,687 total rules
Approval Toolapprove_item (ECR/ECN/ECO/PR use local state machine; Part uses SCSAI real validation)
Intent Rulesbi_change_impact / bi_data_inspect / bi_gov_history / bi_approve_item (All L1 direct pass)
← 返回案例列表

Try 左帮右臂 Industrial Agent — your digital workforce

Available as a WeChat mini program, web app (ylxt.chat) and desktop client. Start from whichever fits your workflow.

🌐 Web App 💻 Desktop Download 📞 Call 189-8606-6876
WeChat mini program QR
Scan for the WeChat mini program
Share:
🤖 Try Now →
🤖
🎁