690609:2223 Prepare to MOD AI flow [skip CI]

Controller was returning { data: engines } which TransformInterceptor
wrapped again into { data: { data: engines } }. extractData() only peeled
one layer, leaving an object instead of the array — causing .map() to fail
in OcrEngineSelector.

- Return data directly from getOcrEngines() and selectOcrEngine()
- Add Array.isArray guard in OcrEngineSelector as defensive layer

.agents/README.md

@@ -6,8 +6,8 @@
 > # Speckit Agent Infrastructure (v1.9.0)
 >
 > - Version: 1.9.0
-> - Last Updated: 2026-05-13
+> - Last Updated: 2026-05-22
-> - Core Principle: **Sync with AGENTS.md v1.9.0**
+> - Core Principle: **Sync with AGENTS.md v1.9.6**
 
 ---
 
@@ -95,18 +95,18 @@ The toolkit is organized into modular components that provide both the logic (Sc
 │   └── util-speckit-*.md          # Utilities (checklist, diff, migrate, etc.)
 │
 ├── rules/                         # Project Context & Validation Rules
-│   ├── 00-project-context.md      # Role, Persona, Rule Tiers (v1.9.0)
+│   ├── 00-project-context.md      # Role, Persona, Rule Tiers (v1.9.6)
 │   ├── 01-adr-019-uuid.md         # UUID Strategy (Critical)
-│   ├── 02-security.md             # Security Requirements
+│   ├── 02-security.md             # Security Requirements (ADR-023/023A)
 │   ├── 03-typescript.md            # TypeScript Standards
 │   ├── 04-domain-terminology.md   # DMS Glossary Compliance
 │   ├── 05-forbidden-actions.md    # Critical Prohibited Patterns
 │   ├── 06-backend-patterns.md     # NestJS Architecture Rules
 │   ├── 07-frontend-patterns.md    # Next.js App Router Rules
-│   ├── 08-development-flow.md      # Development Workflow
+│   ├── 08-development-flow.md      # Development Workflow (Tier 3 SPECIALIZED WORK)
 │   ├── 09-commit-checklist.md      # Pre-commit Validation
 │   ├── 10-error-handling.md       # ADR-007 Compliance
-│   └── 11-ai-integration.md       # ADR-018/020 AI Boundaries
+│   └── 11-ai-integration.md       # ADR-023/023A AI Boundaries
 │
 └── scripts/
     ├── bash/                      # Bash Core (Kinetic logic)
@@ -265,9 +265,9 @@ If you change your mind mid-project:
 
 ---
 
-## 🏗️ LCBP3-DMS Project Notes (v1.9.0)
+## 🏗️ LCBP3-DMS Project Notes (v1.9.6)
 
-### 📊 Current Status: Production Ready (2026-04-14)
+### 📊 Current Status: Production Ready (2026-05-22)
 
 | Area          | Status                          |
 | ------------- | ------------------------------- |
@@ -306,7 +306,7 @@ If you change your mind mid-project:
 - ❌ DO NOT bypass Release Gates before deploying — `04-08-release-management-policy.md`
 - ❌ DO NOT start Migration without Gate #1 approval — `03-06-migration-business-scope.md`
 - ❌ DO NOT use TypeORM Migrations — modify schema SQL directly (ADR-009)
-- ❌ DO NOT give Ollama direct DB access — all writes via DMS API (ADR-018)
+- ❌ DO NOT give Ollama direct DB access — all writes via DMS API (ADR-023/023A)
 - ❌ DO NOT use `any` TypeScript type anywhere
 
 ---
@@ -325,7 +325,7 @@ If you change your mind mid-project:
 # Run version validation
 ./scripts/bash/validate-versions.sh
 
-# Fix by updating all files to v1.9.0
+# Fix by updating all files to v1.9.6
 # Then re-run validation to confirm
 ```
 

.agents/rules/00-project-context.md

@@ -1,56 +1,164 @@
-# NAP-DMS Project Context & Core Rules
+# NAP-DMS Project Context & Rules
 
-- Version: 1.9.3
+- For: Devin Cascade (and compatible: Codex CLI, opencode, Amp, Antigravity, AGENTS.md tools)
-- Last Updated: 2026-05-15
+- Version: 1.9.10 | Last synced from repo: 2026-06-06
-- Status: Production Ready
+- Repo: [https://git.np-dms.work/np-dms/lcbp3](https://git.np-dms.work/np-dms/lcbp3)
-- Canonical Source: AGENTS.md
+- Skill pack: `.agents/skills/` (v1.9.0, 21 skills) — see [`skills/README.md`](./.agents/skills/README.md) + [`skills/_LCBP3-CONTEXT.md`](./.agents/skills/_LCBP3-CONTEXT.md)
 
-## 🎭 Role & Persona
+## 🧠 Role & Persona
 
-Act as a **Senior Full Stack Developer** specialized in NestJS, Next.js, and TypeScript.
+Act as **Senior Full Stack Developer** specialized in NestJS, Next.js, TypeScript, DMS. Focus: Data Integrity, Security, Maintainability, Performance.
-Focus on **Data Integrity, Security, Maintainability, and Performance**.
+
-You are a **Document Intelligence Engine** — every response must be precise, spec-compliant, and production-ready.
+You are a **Document Intelligence Engine** — not a general chatbot. Every response must be **precise**, **spec-compliant**, and **production-ready**.
 
 ---
 
-## 🔴 Tier 1 — CRITICAL (CI BLOCKER)
+## 🧩 Thought & Planning Protocol (Powered by Everything-Claude-Code)
 
-1. **Identifier Strategy (ADR-019)**
+Before writing any code or taking any action in Tier 1 and Tier 2, the AI must demonstrate the following thinking process:
-   - ห้ามใช้ `parseInt()` บน UUID
+
-   - ใช้ `publicId` (string) สำหรับการติดต่อภายนอก (API/URL) เท่านั้น
+### 1. Analysis Phase (Explore & Analyze)
-2. **Database Management (ADR-009)**
+
-   - ห้ามใช้ TypeORM Migrations หรือ `synchronize: true`
+Problem Understanding: Restate what the user wants in clear, unambiguous terms.
-   - การแก้ Schema ต้องแก้ที่ SQL files ใน `specs/03-Data-and-Storage/` เท่านั้น
+Context Search: Identify the relevant Spec files or ADRs from the "Key Spec Files" table that must be read before starting.
-3. **Security (ADR-016)**
+Constraints Identification: Identify key constraints (e.g. Security rules, UUID patterns, or Domain terminology).
-   - ทุก API ต้องมี `CASL Guard` และตรวจสอบสิทธิ์ผ่าน RBAC Matrix
+
-   - การอัปโหลดไฟล์ต้องเป็น Two-Phase (Temp → Commit) และต้องสแกน ClamAV
+### 2. Planning Phase (Plan)
-4. **AI Boundary (ADR-018)**
+
-   - AI Agent ต้องทำงานผ่าน DMS API เท่านั้น ห้ามเขียนลง Database หรือ Storage โดยตรง
+Alternative Exploration: Present at least 2 solution approaches (where possible) with pros/cons analysis.
+Step-by-Step Roadmap: Write a file-by-file plan of changes before executing.
+Verification Plan: Specify how to verify the work is complete (e.g. "which unit tests to write" or "which file to check the schema in").
+
+### 3. Execution & Refinement (Execute & Refine)
+
+Follow the plan step by step, and pause to ask if any uncertainty arises.
+If significant logic changes are made, summarize what was done for the user after completion.
 
 ---
 
-## 📐 TypeScript Rules & Coding Standards (v1.9.0)
+## ⚙️ DMS Workflow Engine Protocol
 
-- **File Header:** ทุกไฟล์ต้องขึ้นต้นด้วย `// File: path/filename`
+กฎนี้ใช้คุม Logic การไหลของเอกสาร (RFA, Transmittal, Correspondence) เพื่อป้องกัน Race Condition และรักษาความถูกต้องของสถานะ:
-- **Change Log:** ต้องมีส่วน `// Change Log` ที่หัวไฟล์
+
-- **Language:** ตัวแปร/Logic เป็น English, Comment/JSDoc เป็น **Thai**
+- **State Management:** ตรวจสอบสถานะปัจจุบันจาก DB ก่อนเสมอ เพื่อป้องกันการอนุมัติซ้ำซ้อน (ดู `05-06-code-snippets.md` `[workflow-transition]`)
-- **Explicit Typing:** กำหนด Type ให้ชัดเจนเสมอ ห้ามใช้ `any`
+- **Concurrency Control:** การจอนเลขที่เอกสารต้องใช้ **Redis Redlock** หรือ **TypeORM `@VersionColumn`** เท่านั้น (ADR-002)
-- **Cleanliness:** ห้ามมีบรรทัดว่างในฟังก์ชัน, Export ได้เพียง 1 symbol หลักต่อไฟล์
+- **Background Jobs:** งานนานหรือการแจ้งเตือนต้องส่งไปทำที่ **BullMQ** ห้ามเขียนแบบ Inline (ADR-008)
+- **Term Consistency:** ห้ามใช้ "Approval Flow" ให้ใช้ **"Workflow Engine"** และห้ามใช้ "Letter" ให้ใช้ **"Correspondence"** (หมายเหตุ: "จดหมาย" ในคอมเมนต์ภาษาไทย = Correspondence ที่ครอบคลุมทุกประเภท)
 
 ---
 
-## 📁 Specs Folder Organization (Hybrid Model)
+## 🛡️ Security & Integrity Audit Protocol
 
-- **Core (00-06):** ข้อมูลอ้างอิงถาวร (Permanent Source of Truth)
+กฎนี้ให้ AI เป็น Gatekeeper ก่อน Commit โดยเน้น **Tier 1 — CRITICAL**:
-- **Feature (100-300):** สำหรับงาน Implementation ใหม่
+
-  - `100-Infrastructures/`
+- **UUID Validation:** ตรวจสอบว่าเป็น **UUIDv7** และห้ามใช้ `parseInt()` บน UUID (ADR-019)
-  - `200-fullstacks/`
+- **RBAC Check:** API ใหม่ต้องมี **CASL Guard** และตรวจสอบ 4-Level RBAC Matrix (ADR-016)
-  - `300-others/`
+- **Data Isolation:** AI ต้องรันผ่าน **Ollama บน Admin Desktop** เท่านั้น ห้ามเข้าถึง DB/storage โดยตรง (ADR-023)
+- **Input Sanitization:** ไฟล์อัปโหลดต้องผ่าน **Two-Phase** (Temp → Commit) และสแกนด้วย **ClamAV** (ADR-016)
 
 ---
 
-## 🔄 Workflow Engine (ADR-001/021)
+## 🧭 Rule Enforcement Tiers
 
-- ใช้ DSL-based state machine
+### 🔴 Tier 1 — CRITICAL (CI BLOCKER)
-- การเปลี่ยนสถานะต้องตรวจสอบสถานะปัจจุบันจาก DB ก่อนเสมอ
+
-- งานที่ใช้เวลานานต้องส่งไปที่ **BullMQ** เท่านั้น
+Build fails หากละเมิด:
+
+- Security (Auth, RBAC, Validation)
+- UUID Strategy (ADR-019) — no `parseInt` / `Number` / `+` on UUID
+- Database correctness — verify schema before writing queries
+- File upload security (ClamAV + whitelist)
+- AI validation boundary (ADR-023)
+- Error handling strategy (ADR-007)
+- Forbidden patterns: `any`, `console.log`, UUID misuse, `id ?? ''` fallback
+
+### 🟡 Tier 2 — IMPORTANT (CODE REVIEW)
+
+Must fix ก่อน merge:
+
+- Architecture patterns (thin controller, business logic in service)
+- Test coverage (80%+ business logic, 70%+ backend overall)
+- Cache invalidation
+- Naming conventions
+- **TypeScript Standards:** Missing JSDoc, explicit types, or file headers
+
+### 🟢 Tier 3 — SPECIALIZED WORK
+
+Requires domain-specific knowledge:
+
+- **ADR-021 Integration:** Workflow Engine & Context implementation
+- **AI Infrastructure:** ADR-023/023A boundary enforcement and pipeline usage
+- **AI Runtime Layer:** ADR-024 Intent Classification, ADR-025 Tool Layer, ADR-026 Chat UI, ADR-027 Admin Console
+- **Migration Pipeline:** ADR-028 Staging Queue & post-migration cleanup
+- **Complex Business Logic:** Multi-step workflows with state management
+- **Performance Optimization:** Database queries, caching strategies, bulk operations
+
+### 🔵 Tier 4 — GUIDELINES
+
+Best practice — follow when possible:
+
+- Code style / formatting (Prettier handles)
+- Comment completeness
+- Minor optimizations
+
+---
+
+## 📐 TypeScript Rules & Coding Standards
+
+### 📝 Core Standards
+
+- **Strict Mode** — all strict checks enforced.
+- **ZERO `any` types** — use proper types or `unknown` + narrowing.
+- **ZERO `console.log`** — use NestJS `Logger` (backend) or remove (frontend).
+- **English for Code** — use English for all code identifiers, variables, and logic.
+- **Thai for Comments** — use Thai for comments, documentation, and JSDoc.
+- **Explicit Typing** — explicitly define types for all variables, parameters, and return values.
+- **JSDoc** — use JSDoc for all public classes and methods.
+
+### 🏗️ File & Function Structure
+
+- **File Headers** — every file MUST start with `// File: path/filename` on the first line.
+- **Change Log** — include `// Change Log` at the top of the file to track modifications.
+- **Single Export** — export **only one main symbol** per file.
+- **Function Style** — avoid unnecessary blank lines inside functions.
+
+---
+
+## 🚫 Forbidden Actions
+
+| ❌ Forbidden                                    | ✅ Correct Approach                                     | ⚠️ Why                                               |
+| ----------------------------------------------- | ------------------------------------------------------- | ---------------------------------------------------- |
+| SQL Triggers for business logic                 | NestJS Service methods                                  | Untestable; bypasses audit log                       |
+| `.env` files in production                      | `docker-compose.yml` environment section                | Secrets exposed in version control                   |
+| TypeORM migration files                         | Edit schema SQL directly (ADR-009)                      | Migration drift risk; schema managed via SQL delta   |
+| Inventing table/column names                    | Verify against `schema-02-tables.sql`                   | Schema mismatch causes silent runtime errors         |
+| `any` TypeScript type                           | Proper types / generics                                 | Defeats strict mode; hides runtime type errors       |
+| `console.log` in committed code                 | NestJS Logger (backend) / remove (frontend)             | Log flooding in production; risk of data leakage     |
+| `req: any` in controllers                       | `RequestWithUser` typed interface                       | Type safety lost; auth context unreachable           |
+| `parseInt()` on UUID values                     | Use UUID string directly (ADR-019)                      | `"0195…"` parsed to integer `19` — silently wrong    |
+| Exposing INT PK in API responses                | UUIDv7 `publicId` (ADR-019)                             | Leaks row count; enables DB enumeration attacks      |
+| AI accessing DB/storage directly                | AI → DMS API → DB (ADR-023/023A)                        | Bypasses RBAC, audit trail, and validation layer     |
+| Direct file operations bypassing StorageService | `StorageService` for all file moves                     | Orphaned files; broken ClamAV scan; no audit trail   |
+| Inline email/notification sending               | BullMQ queue job (ADR-008)                              | Blocks request thread; no retry on transient failure |
+| Deploying without Release Gates                 | Complete `04-08-release-management-policy.md`           | Unverified deploy risks data loss in production      |
+| AI direct cloud API calls                       | On-premises Ollama only (ADR-023/023A)                  | Data privacy violation; no audit control             |
+| AI outputs without human validation             | Human-in-the-loop validation required (ADR-023/023A)    | Unvalidated AI metadata corrupts document records    |
+| n8n calling Ollama/Qdrant directly              | n8n → DMS API → BullMQ → Ollama (ADR-023A)              | Bypasses audit log, RBAC, and error handling layer   |
+| Qdrant query without `projectPublicId` filter   | `QdrantService.search(projectPublicId, ...)` (ADR-023A) | Cross-project data leak via vector search            |
+
+---
+
+## 🚧 Out of Scope — Never Do Without Explicit Approval
+
+| ❌ Never Do Autonomously                                        | ⚠️ Why Approval Is Required                                      |
+| --------------------------------------------------------------- | ---------------------------------------------------------------- |
+| `DROP` or `RENAME` a column / table                             | Irreversible data loss — requires DBA + PM sign-off              |
+| Push directly to `main` / `master` branch                       | Bypasses CI, code review, and release gates                      |
+| Generate or insert seed data into production database           | May corrupt live data or violate business state invariants       |
+| Delete files from permanent storage                             | Files may be referenced in active documents or audit trails      |
+| Modify RBAC permission matrix without security team approval    | Defines access control for all users — security boundary change  |
+| Upgrade major library versions (NestJS, Next.js, TypeORM, etc.) | Breaking changes require full regression test cycle              |
+| Disable or modify authentication / authorization guards         | Creates unguarded endpoints — immediate security risk            |
+| Change Redis lock TTL or disable Redlock                        | Risk of document number race condition (ADR-002)                 |
+| Create or supersede an ADR unilaterally                         | Architecture decisions require team consensus and review process |
+| Add new columns to production tables without schema review      | Must update Data Dictionary + downstream queries simultaneously  |

.agents/rules/02-security.md

@@ -9,9 +9,9 @@
 5. **Password:** bcrypt 12 salt rounds, min 8 chars, rotate every 90 days
 6. **Rate Limiting:** `ThrottlerGuard` on all auth endpoints
 7. **File Upload:** Whitelist PDF/DWG/DOCX/XLSX/ZIP, max 50MB, ClamAV scan
-8. **AI Isolation (ADR-018):** Ollama on Admin Desktop ONLY — NO direct DB/storage access
+8. **AI Isolation (ADR-023/023A):** Ollama on Admin Desktop ONLY — NO direct DB/storage access; 2-model stack `gemma4:e4b Q8_0` + `nomic-embed-text`; all inference via BullMQ (`ai-realtime` / `ai-batch`)
 9. **Error Handling (ADR-007):** Use layered error classification with user-friendly messages
-10. **AI Integration (ADR-020):** RFA-First approach with unified pipeline architecture
+10. **AI Integration (ADR-023/023A):** RFA-First approach; n8n orchestrates Migration Phase only via DMS API — never calls Ollama directly; `QdrantService.search()` requires `projectPublicId` as mandatory param
 11. **AI Audit Trail:** Log all AI interactions and human validations
 12. **Rate Limiting:** Apply to AI endpoints to prevent abuse
 
@@ -26,7 +26,7 @@
 - [ ] No SQL injection vulnerabilities
 - [ ] File upload validation (whitelist + ClamAV)
 - [ ] Rate limiting applied to auth endpoints
-- [ ] AI boundary enforcement (ADR-023) - no direct DB/storage access
+- [ ] AI boundary enforcement (ADR-023/023A) - no direct DB/storage access
 - [ ] AI audit logging implemented for AI interactions
 - [ ] AI outputs validated before use (human-in-the-loop)
 - [ ] Error handling follows ADR-007 layered classification

.agents/rules/03-typescript.md

@@ -1,4 +1,4 @@
-# TypeScript Rules (v1.9.3)
+# TypeScript Rules
 
 ## Core Standards
 

.agents/rules/04-domain-terminology.md

@@ -2,17 +2,17 @@
 
 ## DMS Glossary
 
-| ✅ Use             | ❌ Don't Use                          |
+| ✅ Use             | ❌ Don't Use                          | คำอธิบายเพิ่มเติม                                |
-| ------------------ | ------------------------------------- |
+| ------------------ | ------------------------------------- | ------------------------------------------------ |
-| Correspondence     | Letter, Communication, Document       |
+| Correspondence     | Letter, Communication, Document       | ครอบคลุมทุกประเภท: Letter, RFA, Memo, ฯลฯ        |
-| RFA                | Approval Request, Submit for Approval |
+| RFA                | Approval Request, Submit for Approval | เอกสารขออนุมัติ (ชนิดหนึ่งของ Correspondence)    |
-| Transmittal        | Delivery Note, Cover Letter           |
+| Transmittal        | Delivery Note, Cover Letter           | เอกสารนำส่ง (ชนิดหนึ่งของ Correspondence)        |
-| Circulation        | Distribution, Routing                 |
+| Circulation        | Distribution, Routing                 | ใบเวียนเอกสารภายใน (ชนิดหนึ่งของ Correspondence) |
-| Shop Drawing       | Construction Drawing                  |
+| Shop Drawing       | Construction Drawing                  | แบบก่อสร้าง                                      |
-| Contract Drawing   | Design Drawing, Blueprint             |
+| Contract Drawing   | Design Drawing, Blueprint             | แบบคู่สัญญา                                      |
-| Workflow Engine    | Approval Flow, Process Engine         |
+| Workflow Engine    | Approval Flow, Process Engine         | เครื่องมือจัดการลำดับงาน                         |
-| Document Numbering | Document ID, Auto Number              |
+| Document Numbering | Document ID, Auto Number              | ระบบจัดการเลขที่เอกสาร                           |
-| RBAC               | Permission System (generic)           |
+| RBAC               | Permission System (generic)           | การควบคุมสิทธิ์ตามบทบาท                          |
 
 ## Full Glossary
 
@@ -23,13 +23,19 @@
 Spec priority: **`06-Decision-Records`** > **`05-Engineering-Guidelines`** > others
 
 | Document                       | Path                                                                        | Use When                          |
-| ----------------------- | ----------------------------------------------------------------- | ------------------------------- |
+| ------------------------------ | --------------------------------------------------------------------------- | --------------------------------- |
 | **Glossary**                   | `specs/00-overview/00-02-glossary.md`                                       | Verify domain terminology         |
 | **Schema Tables**              | `specs/03-Data-and-Storage/lcbp3-v1.9.0-schema-02-tables.sql`               | Before writing any query          |
 | **Data Dictionary**            | `specs/03-Data-and-Storage/03-01-data-dictionary.md`                        | Field meanings + business rules   |
 | **Edge Cases**                 | `specs/01-Requirements/01-06-edge-cases-and-rules.md`                       | Prevent bugs in flows             |
 | **ADR-019 UUID**               | `specs/06-Decision-Records/ADR-019-hybrid-identifier-strategy.md`           | UUID-related work                 |
 | **ADR-023 AI**                 | `specs/06-Decision-Records/ADR-023-unified-ai-architecture.md`              | AI integration work               |
+| **ADR-023A AI Model**          | `specs/06-Decision-Records/ADR-023A-unified-ai-architecture.md`             | 2-model stack, BullMQ 2-queue     |
+| **ADR-024 Intent Class.**      | `specs/06-Decision-Records/ADR-024-intent-classification-strategy.md`       | Pattern→LLM Fallback; Redis cache |
+| **ADR-025 AI Tool Layer**      | `specs/06-Decision-Records/ADR-025-ai-tool-layer-architecture.md`           | Tool Registry; CASL-guarded       |
+| **ADR-026 Chat UI**            | `specs/06-Decision-Records/ADR-026-document-chat-ui-pattern.md`             | Side-panel; streaming SSE         |
+| **ADR-027 AI Admin Console**   | `specs/06-Decision-Records/ADR-027-ai-admin-console-and-dynamic-control.md` | Dynamic control; admin-only       |
+| **ADR-028 Migration Refactor** | `specs/06-Decision-Records/ADR-028-migration-architecture-refactor.md`      | Staging Queue; cleanup            |
 | **Backend Guidelines**         | `specs/05-Engineering-Guidelines/05-02-backend-guidelines.md`               | NestJS patterns                   |
 | **Frontend Guidelines**        | `specs/05-Engineering-Guidelines/05-03-frontend-guidelines.md`              | Next.js patterns                  |
 | **Testing Strategy**           | `specs/05-Engineering-Guidelines/05-04-testing-strategy.md`                 | Coverage goals                    |

.agents/rules/05-forbidden-actions.md

@@ -2,25 +2,25 @@
 
 ## ❌ Never Do This
 
-| ❌ Forbidden                                    | ✅ Correct Approach                                               |
+| ❌ Forbidden                                    | ✅ Correct Approach                                     | ⚠️ Why                                               |
-| ----------------------------------------------- | ----------------------------------------------------------------- |
+| ----------------------------------------------- | ------------------------------------------------------- | ---------------------------------------------------- |
-| SQL Triggers for business logic                 | NestJS Service methods                                            |
+| SQL Triggers for business logic                 | NestJS Service methods                                  | Untestable; bypasses audit log                       |
-| `.env` files in production                      | `docker-compose.yml` environment section                          |
+| `.env` files in production                      | `docker-compose.yml` environment section                | Secrets exposed in version control                   |
-| TypeORM migration files                         | Edit schema SQL directly (ADR-009)                                |
+| TypeORM migration files                         | Edit schema SQL directly (ADR-009)                      | Migration drift risk; schema managed via SQL delta   |
-| Inventing table/column names                    | Verify against `lcbp3-v1.9.0-schema-02-tables.sql`                |
+| Inventing table/column names                    | Verify against `schema-02-tables.sql`                   | Schema mismatch causes silent runtime errors         |
-| `any` TypeScript type                           | Proper types / generics                                           |
+| `any` TypeScript type                           | Proper types / generics                                 | Defeats strict mode; hides runtime type errors       |
-| `console.log` in committed code                 | NestJS Logger (backend) / remove (frontend)                       |
+| `console.log` in committed code                 | NestJS Logger (backend) / remove (frontend)             | Log flooding in production; risk of data leakage     |
-| `req: any` in controllers                       | `RequestWithUser` typed interface                                 |
+| `req: any` in controllers                       | `RequestWithUser` typed interface                       | Type safety lost; auth context unreachable           |
-| `parseInt()` on UUID values                     | Use UUID string directly (ADR-019)                                |
+| `parseInt()` on UUID values                     | Use UUID string directly (ADR-019)                      | `"0195…"` parsed to integer `19` — silently wrong    |
-| Exposing INT PK in API responses                | UUIDv7 (ADR-019)                                                  |
+| Exposing INT PK in API responses                | UUIDv7 `publicId` (ADR-019)                             | Leaks row count; enables DB enumeration attacks      |
-| AI accessing DB/storage directly                | AI → DMS API → DB (ADR-023)                                       |
+| AI accessing DB/storage directly                | AI → DMS API → DB (ADR-023/023A)                        | Bypasses RBAC, audit trail, and validation layer     |
-| Direct file operations bypassing StorageService | `StorageService` for all file moves                               |
+| Direct file operations bypassing StorageService | `StorageService` for all file moves                     | Orphaned files; broken ClamAV scan; no audit trail   |
-| Inline email/notification sending               | BullMQ queue job                                                  |
+| Inline email/notification sending               | BullMQ queue job (ADR-008)                              | Blocks request thread; no retry on transient failure |
-| Deploying without Release Gates                 | Complete `04-08-release-management-policy.md`                     |
+| Deploying without Release Gates                 | Complete `04-08-release-management-policy.md`           | Unverified deploy risks data loss in production      |
-| AI direct cloud API calls                       | On-premises Ollama only (ADR-023)                                 |
+| AI direct cloud API calls                       | On-premises Ollama only (ADR-023/023A)                  | Data privacy violation; no audit control             |
-| AI outputs without human validation             | Human-in-the-loop validation required (ADR-023)                   |
+| AI outputs without human validation             | Human-in-the-loop validation required (ADR-023/023A)    | Unvalidated AI metadata corrupts document records    |
-| n8n calling Ollama/Qdrant directly              | n8n → DMS API → BullMQ → Ollama/Qdrant (ADR-023A)                 |
+| n8n calling Ollama/Qdrant directly              | n8n → DMS API → BullMQ → Ollama (ADR-023A)              | Bypasses audit log, RBAC, and error handling layer   |
-| Qdrant query without projectPublicId filter     | QdrantService.search(projectPublicId: string) required (ADR-023A) |
+| Qdrant query without `projectPublicId` filter   | `QdrantService.search(projectPublicId, ...)` (ADR-023A) | Cross-project data leak via vector search            |
 
 ## Schema Changes (ADR-009)
 

.agents/rules/07-frontend-patterns.md

@@ -1,4 +1,3 @@
-
 # Frontend Patterns (Next.js)
 
 ## Form Handling
@@ -17,7 +16,7 @@ interface ProjectOption {
 }
 
 // Select options
-const options = contracts.map(c => ({
+const options = contracts.map((c) => ({
   label: `${c.contractName} (${c.contractCode})`,
   value: c.publicId!, // Use publicId, no fallback to id
 }));

.agents/rules/08-development-flow.md

@@ -9,7 +9,7 @@
 3. **Check schema** — verify table/column in `lcbp3-v1.9.0-schema-02-tables.sql`
 4. **Check data dictionary** — confirm field meanings + business rules
 5. **Scan edge cases** — `01-06-edge-cases-and-rules.md`
-6. **Check ADRs** — verify decisions align (ADR-009, ADR-018, ADR-019)
+6. **Check ADRs** — verify decisions align (ADR-009, ADR-019, ADR-023)
 7. **Write code** — TypeScript strict, no `any`, no `console.log`
 
 ## 🟡 Normal Work — UI / Feature / Integration
@@ -28,15 +28,183 @@
 - Add minimal test if logic changed
 - Check forbidden patterns before commit
 
+### 🟢 Specialized Work — ADR-021, AI Runtime Layer, Complex Logic
+
+**MUST complete:**
+
+1. **Domain Knowledge Check** - Read relevant ADRs (ADR-021, ADR-023/023A, ADR-024~028)
+2. **Pattern Verification** - Check existing implementations in codebase
+3. **Specialized Requirements** - Follow domain-specific patterns
+4. **Complex Logic Testing** - Multi-scenario test coverage
+5. **Performance Validation** - Load testing if applicable
+
+**For ADR-021 Integration:**
+
+- Read ADR-021 - Integrated workflow & step attachments
+- Check ADR-001 - Unified workflow engine patterns
+- Verify WorkflowEngineService - Polymorphic instance handling
+- Add workflow fields - Expose workflowInstanceId, workflowState, availableActions
+- Include IntegratedBanner - Frontend workflow lifecycle display
+- Test workflow transitions - State changes and action validation
+
+**For AI Infrastructure (ADR-023/023A):**
+
+- Verify AI boundary enforcement - No direct DB/storage access
+- Check BullMQ 2-queue setup - ai-realtime + ai-batch
+- Validate Qdrant multi-tenancy - projectPublicId filter required
+- Test human-in-the-loop validation workflows
+- Audit AI interaction logging to ai_audit_logs
+
+**For AI Runtime Layer (ADR-024/025/026/027):**
+
+- ADR-024: Pattern Layer first (ai_intent_patterns DB + Redis cache 5 min) → LLM Fallback (gemma4:e4b, semaphore max=3)
+- ADR-025: Tool Registry dispatch — AI Gateway → Tool → Business Service; ToolResult DTO must use publicId only
+- ADR-026: useAiChat() hook + side-panel UI; streaming response via SSE; TanStack Query cache
+- ADR-027: Admin Console — dynamic model/prompt/intent control; CASL-guarded admin-only endpoints
+
+**For Migration Pipeline (ADR-028):**
+
+- Use Staging Queue pattern — never write directly to production tables
+- Post-migration cleanup process required after each batch
+- Migration Validation Gates must pass before promoting to production
+
+**Expected output:**
+
+- Backend services expose specialized context fields
+- Frontend components use domain-specific patterns
+- Complex state management with proper validation
+- Performance metrics within acceptable thresholds
+- Comprehensive test coverage for edge cases
+
+---
+
 ## Context-Aware Triggers
 
 | Request                     | Files to Check                                                                        | Expected Response                                                       |
-| -------------------- | -------------------------------------------------------------------- | --------------------------------------------------- |
+| --------------------------- | ------------------------------------------------------------------------------------- | ----------------------------------------------------------------------- |
 | "สร้าง API ใหม่"            | `05-02-backend-guidelines.md`, `lcbp3-v1.9.0-schema-02-tables.sql`                    | NestJS Controller + Service + DTO + CASL Guard                          |
-| "แก้ฟอร์ม frontend"  | `05-03-frontend-guidelines.md`, `01-06-edge-cases.md`                | RHF+Zod + TanStack Query + Thai comments            |
+| "แก้ฟอร์ม frontend"         | `05-03-frontend-guidelines.md`, `01-06-edge-cases-and-rules.md`                       | RHF+Zod + TanStack Query + Thai comments                                |
-| "เพิ่ม field ใหม่"   | `ADR-009`, `data-dictionary.md`, `lcbp3-v1.9.0-schema-02-tables.sql` | Edit SQL directly + update Data Dictionary + Entity |
+| "เพิ่ม field ใหม่"          | `ADR-009`, `03-01-data-dictionary.md`, `lcbp3-v1.9.0-schema-02-tables.sql`            | Edit SQL directly + update Data Dictionary + Entity                     |
 | "ตรวจสอบ UUID"              | `ADR-019`, `05-07-hybrid-uuid-implementation-plan.md`                                 | UUIDv7 MariaDB native UUID + TransformInterceptor                       |
 | "สร้าง migration"           | `ADR-009`, `03-06-migration-business-scope.md`                                        | Edit SQL schema directly + n8n workflow                                 |
-| "ตรวจสอบ permission" | `seed-permissions.sql`, `ADR-016`                                    | CASL 4-Level RBAC matrix                            |
+| "ตรวจสอบ permission"        | `lcbp3-v1.9.0-seed-permissions.sql`, `ADR-016`                                        | CASL 4-Level RBAC matrix                                                |
 | "deploy production"         | `04-08-release-management-policy.md`, `ADR-015`                                       | Release Gates + Blue-Green strategy                                     |
 | "เพิ่ม test"                | `05-04-testing-strategy.md`                                                           | Coverage goals + test patterns                                          |
+| "AI integration"            | `ADR-023`, `ADR-023A`, `ADR-024`, `ADR-025`                                           | AI boundary + 2-model stack + BullMQ queue policy + Intent/Tool Layer   |
+| "Error handling"            | `ADR-007`                                                                             | Layered error classification + recovery                                 |
+| "File upload"               | `ADR-016`, `05-02-backend-guidelines.md`, `03-Data-and-Storage/03-03-file-storage.md` | Two-phase upload → temp → commit; ClamAV + whitelist                    |
+| "Notifications / Queue"     | `ADR-008`, `05-02-backend-guidelines.md`                                              | BullMQ job — never inline; check retry + dead-letter                    |
+| "Add i18n / translate"      | `05-08-i18n-guidelines.md`                                                            | i18n keys only — no hardcoded text                                      |
+| "Workflow / DSL"            | `ADR-001`, `01-03-modules/01-03-06-unified-workflow.md`                               | DSL state machine + WorkflowEngineService                               |
+| "Document numbering"        | `ADR-002`, `01-02-business-rules/01-02-02-doc-numbering-rules.md`                     | Redis Redlock + DB optimistic lock (double-lock)                        |
+| "ตรวจสอบ Workflow"          | `01-06-edge-cases-and-rules.md`, `05-02-backend-guidelines.md`, `ADR-001`, `ADR-002`  | เช็คการเปลี่ยน State, คิว BullMQ และการล็อกเลขที่เอกสาร                 |
+| "Transmittal submit"        | `ADR-021`, `specs/200-fullstacks/201-transmittals-circulation/`                       | submit() with EC-RFA-004 validation                                     |
+| "Circulation reassign"      | `ADR-021`, `specs/200-fullstacks/201-transmittals-circulation/`                       | reassignRouting() with EC-CIRC-001                                      |
+| "สร้าง workflow ใหม่"       | `ADR-001`, `ADR-021`, `specs/200-fullstacks/203-unified-workflow-engine/`             | DSL workflow definition + WorkflowEngineService setup                   |
+| "ตรวจสอบ AI boundary"       | `ADR-023`, `ADR-023A`                                                                 | Verify Ollama isolation + BullMQ queues + Qdrant projectPublicId filter |
+| "Intent classification"     | `ADR-024`, `specs/200-fullstacks/224-intent-classification/`                          | Pattern Layer → LLM Fallback; ai_intent_patterns; Redis cache 5 min     |
+| "AI Tool Layer"             | `ADR-025`, `specs/200-fullstacks/225-ai-tool-layer-architecture/`                     | Tool Registry; CASL-guarded dispatch; ToolResult publicId only          |
+| "Document Chat UI"          | `ADR-026`, `specs/200-fullstacks/226-document-chat-ui-pattern/`                       | Side-panel; useAiChat() hook; streaming SSE; TanStack Query cache       |
+| "AI Admin Console"          | `ADR-027`, `specs/200-fullstacks/227-ai-admin-console/`                               | Dynamic model/prompt/intent control; admin-only CASL endpoints          |
+| "Migration refactor"        | `ADR-028`, `specs/200-fullstacks/228-migration-arch-refactor/`                        | Staging Queue; post-migration cleanup; validation gates                 |
+| "จัดการ document numbering" | `ADR-002`, `specs/03-Data-and-Storage/03-04-document-numbering.md`                    | Redis Redlock + template system + preview/override workflows            |
+| "Audit ความปลอดภัย"         | `ADR-016`, `ADR-019`, `ADR-023`, `ADR-023A`                                           | ตรวจสอบ UUID pattern, CASL Guard, AI Boundary และ Qdrant multi-tenancy  |
+| "แก้ bug / bugfix"          | `.agents/workflows/bugfix.md`, `error-catalog.md`                                     | ใช้ bugfix workflow สำหรับเคสที่สาเหตุชัดเจน                            |
+| "ตรวจแอปจริง"               | `.windsurf/workflows/check-real-app.md`                                               | ตรวจ endpoint/UI/console หลัง build pass — No Fake Evidence             |
+| "งานค้าง / resume"          | `.windsurf/workflows/resume-pending-work.md`                                          | อ่าน checkpoint เดิม → ตรวจ build → วางแผนต่อโดยไม่ทำงานซ้ำ             |
+
+---
+
+## 🔌 MCP MariaDB Tools
+
+MCP MariaDB server ให้เครื่องมือสำหรับตรวจสอบและจัดการ database โดยตรง ใช้สำหรับ:
+
+- ตรวจสอบ schema กับ spec file `specs/03-Data-and-Storage/lcbp3-v1.9.0-schema-02-tables.sql`
+- Debug ปัญหา database โดยไม่ต้องเข้า MySQL client
+- ตรวจสอบ data ใน production/staging
+- Validate การเปลี่ยนแปลง schema ก่อน deploy
+
+### Available Tools
+
+| Tool | หน้าที่ | ตัวอย่างการใช้งาน |
+|------|----------|------------------|
+| `mcp1_mysql_test_connection` | ทดสอบ connection กับ database | ตรวจสอบว่า MCP server เชื่อมต่อได้ |
+| `mcp1_mysql_show_databases` | แสดง databases ทั้งหมด | ดูว่ามี database อะไรบ้าง |
+| `mcp1_mysql_show_tables` | แสดง tables ทั้งหมดใน database | ดูรายชื่อ tables ใน `lcbp3` |
+| `mcp1_mysql_describe_table` | ดู structure/columns ของ table | ตรวจสอบ columns, types, keys ของ `correspondences` |
+| `mcp1_mysql_query` | รัน SELECT query | ดู data ใน table หรือ join query |
+| `mcp1_mysql_insert` | INSERT data | เพิ่ม seed data หรือ test data |
+| `mcp1_mysql_update` | UPDATE data | แก้ไข data ใน table |
+| `mcp1_mysql_delete` | DELETE data | ลบ data ใน table |
+
+### การใช้งานร่วมกับ Development Flow
+
+**เมื่อเขียน query ใหม่:**
+1. ใช้ `mcp1_mysql_describe_table` เพื่อตรวจสอบ columns และ types
+2. เปรียบเทียบกับ `specs/03-Data-and-Storage/lcbp3-v1.9.0-schema-02-tables.sql`
+3. ใช้ `mcp1_mysql_query` เพื่อทดสอบ query ก่อน implement
+
+**เมื่อเปลี่ยน schema (ADR-009):**
+1. ใช้ `mcp1_mysql_describe_table` เพื่อดู structure ปัจจุบัน
+2. สร้าง SQL delta ใน `specs/03-Data-and-Storage/deltas/`
+3. ใช้ `mcp1_mysql_query` เพื่อตรวจสอบผลลัพธ์หลัง apply delta
+
+**เมื่อ debug ปัญหา database:**
+1. ใช้ `mcp1_mysql_query` เพื่อดู data จริง
+2. เปรียบเทียบกับ spec และ data dictionary
+3. ตรวจสอบ foreign keys และ constraints
+
+### ข้อควรระวัง
+
+- **❌ ห้ามใช้ MCP MariaDB สำหรับ DDL operations** (CREATE/ALTER/DROP) โดยตรง — ต้องใช้ SQL delta ตาม ADR-009
+- **✅ ใช้สำหรับ DQL/DML operations** (SELECT/INSERT/UPDATE/DELETE) เพื่อ debug และ test เท่านั้น
+- **⚠️ ระวัง DELETE operations** — อาจทำให้เสีย data ใน production
+- **✅ ตรวจสอบ schema กับ spec file เสมอ** ก่อนเขียน query
+
+---
+
+## 🧠 MCP Memory Tools
+
+MCP Memory server ให้เครื่องมือสำหรับจัดการ Knowledge Graph และ Long-term Memory ใช้สำหรับ:
+
+- จัดเก็บความรู้และ context ของโปรเจกต์ในรูปแบบ Graph (Entities + Relations + Observations)
+- ค้นหาและดึงข้อมูล context จาก memory ที่บันทึกไว้ใน session ก่อนหน้า
+- สร้าง/แก้ไข/ลบ entities, relations, และ observations ใน knowledge graph
+
+### Available Tools
+
+| Tool | หน้าที่ | ตัวอย่างการใช้งาน |
+|------|----------|------------------|
+| `mcp3_create_entities` | สร้าง entities ใหม่หลายตัวพร้อม observations | สร้าง entity ใหม่เช่น Project, User, Task |
+| `mcp3_create_relations` | สร้าง relations ระหว่าง entities | สร้าง relation: Project → has → User |
+| `mcp3_add_observations` | เพิ่ม observations ให้ entity ที่มีอยู่แล้ว | เพิ่ม context เพิ่มเติมให้ entity |
+| `mcp3_delete_entities` | ลบ entities และ relations ที่เกี่ยวข้อง | ลบ entity ที่ไม่ใช้แล้ว |
+| `mcp3_delete_relations` | ลบ relations ระหว่าง entities | ลบ relation ที่ผิดหรือไม่ใช้แล้ว |
+| `mcp3_delete_observations` | ลบ observations จาก entity | ลบ context ที่ผิดหรือล้าสุด |
+| `mcp3_open_nodes` | ดึงข้อมูล entities ตามชื่อ | ดึง entity ที่ระบุชื่อ |
+| `mcp3_read_graph` | อ่าน knowledge graph ทั้งหมด | ดูทั้ง graph structure |
+| `mcp3_search_nodes` | ค้นหา entities ตาม query | ค้นหา entity จากชื่อ, type, หรือ observation |
+
+### การใช้งานร่วมกับ Development Flow
+
+**เมื่อบันทึก context ใหม่:**
+1. ใช้ `mcp3_create_entities` เพื่อสร้าง entities ใหม่ (ถ้ายังไม่มี)
+2. ใช้ `mcp3_create_relations` เพื่อเชื่อมโยง entities
+3. ใช้ `mcp3_add_observations` เพื่อเพิ่ม context/observations
+
+**เมื่อค้นหา context:**
+1. ใช้ `mcp3_search_nodes` เพื่อค้นหา entities ที่เกี่ยวข้อง
+2. ใช้ `mcp3_open_nodes` เพื่อดึงข้อมูล entities ที่ต้องการ
+3. ใช้ `mcp3_read_graph` เพื่อดู relations ระหว่าง entities
+
+**เมื่อแก้ไข context:**
+1. ใช้ `mcp3_add_observations` เพื่อเพิ่ม observations ใหม่
+2. ใช้ `mcp3_delete_observations` เพื่อลบ observations ที่ผิด
+3. ใช้ `mcp3_create_relations` หรือ `mcp3_delete_relations` เพื่อปรับ relations
+
+### ข้อควรระวัง
+
+- **✅ ใช้สำหรับบันทึก context ที่ต้องใช้ร่วมกันหลาย session** — เช่น การตัดสินใจสำคัญ, architecture decisions, rollout history
+- **⚠️ ระวังการลบ entities** — อาจทำให้เสีย context ที่ยังใช้งานอยู่
+- **✅ ตรวจสอบว่า entity มีอยู่แล้วก่อนสร้าง** — ใช้ `mcp3_search_nodes` หรือ `mcp3_open_nodes` ก่อน
+- **✅ ใช้ชื่อ entity ที่ชัดเจนและไม่ซ้ำกัน** — เพื่อป้องกันความสับสน

.agents/rules/09-commit-checklist.md

@@ -1,4 +1,3 @@
-
 # Commit Checklist
 
 ## Pre-Commit Verification
@@ -28,6 +27,7 @@ type(scope): description
 Types: `feat`, `fix`, `docs`, `style`, `refactor`, `test`, `chore`
 
 Examples:
+
 - `feat(correspondence): add originator organization validation`
 - `fix(uuid): correct parseInt usage to string comparison`
 - `spec(agents): bump to v1.8.5 - refactor structure`

.agents/rules/10-error-handling.md

@@ -1,4 +1,3 @@
-
 # ADR-007 Error Handling Strategy
 
 ## CRITICAL RULES
@@ -12,7 +11,7 @@
 ## Error Classification
 
 | Error Type     | Description               | User Message             | Technical Log              |
-|------------|-------------|--------------|---------------|
+| -------------- | ------------------------- | ------------------------ | -------------------------- |
 | **Validation** | Input validation failures | Clear field-level errors | Full validation details    |
 | **Business**   | Business rule violations  | Actionable guidance      | Business context + user ID |
 | **System**     | Infrastructure failures   | Generic "try again"      | Full stack trace + metrics |
@@ -22,12 +21,7 @@
 ```typescript
 // Custom Exception Hierarchy
 export class BusinessException extends HttpException {
-  constructor(
+  constructor(message: string, userMessage: string, recoveryAction?: string, errorCode?: string) {
-    message: string,
-    userMessage: string,
-    recoveryAction?: string,
-    errorCode?: string
-  ) {
     super({ message, userMessage, recoveryAction, errorCode }, 400);
   }
 }

.agents/rules/11-ai-integration.md

@@ -3,7 +3,7 @@
 ## CRITICAL RULES
 
 - **ALWAYS** follow ADR-023 AI boundary policy (isolation on Admin Desktop)
-- **ALWAYS** use ADR-023A 2-model stack (gemma4:e4b Q8_0 + nomic-embed-text)
+- **ALWAYS** use ADR-034 model stack (typhoon2.5-np-dms:latest + typhoon-np-dms-ocr:latest + nomic-embed-text)
 - **ALWAYS** use BullMQ 2-queue (ai-realtime + ai-batch) for GPU overload prevention
 - **NEVER** allow AI direct database/storage access
 - **ALWAYS** implement human-in-the-loop validation
@@ -26,8 +26,8 @@ n8n (Migration) → DMS API → BullMQ → Admin Desktop (Ollama) → Backend Va
 | ----------------- | ------------------------- | ------------------------------------------------------------------------ |
 | **AI Gateway**    | Backend (NestJS)          | API endpoints, validation, audit logging                                 |
 | **BullMQ Queues** | Backend (NestJS)          | ai-realtime (RAG/Suggest), ai-batch (OCR/Extract/Embed)                  |
-| **Ollama Engine** | Admin Desktop (Desk-5439) | gemma4:e4b Q8_0 (LLM) + nomic-embed-text (Embedding)                     |
+| **Ollama Engine** | Admin Desktop (Desk-5439) | typhoon2.5-np-dms:latest (Main LLM) + typhoon-np-dms-ocr:latest (OCR, keep_alive:0) + nomic-embed-text (Embedding) |
-| **OCR Engine**    | Admin Desktop (Desk-5439) | PaddleOCR + PyThaiNLP (Thai/English text extraction)                     |
+| **OCR Engine**    | Admin Desktop (Desk-5439) | Tesseract OCR + Typhoon OCR (via Ollama) + PyThaiNLP (Thai/English text extraction) |
 | **Orchestrator**  | QNAP NAS (n8n)            | Migration Phase orchestrator only (calls DMS API, never Ollama directly) |
 
 ## Backend Implementation (NestJS)
@@ -76,7 +76,7 @@ export class AiService {
   async extractMetadata(documentId: string): Promise<AIMetadata> {
     // 1. Validate permissions
     // 2. Queue job to BullMQ (ai-batch or ai-realtime)
-    // 3. Worker sends to Admin Desktop AI (gemma4:e4b Q8_0)
+    // 3. Worker sends to Admin Desktop AI (typhoon2.5-np-dms:latest)
     // 4. Validate AI response
     // 5. Log audit trail to ai_audit_logs
     // 6. Return validated results
@@ -113,12 +113,12 @@ const DocumentReviewForm = ({ document, aiSuggestions }) => {
 - **n8n Boundary:** n8n MUST call DMS API → BullMQ, NEVER Ollama/Qdrant directly
 - **GPU Overload Prevention:** BullMQ 2-queue (ai-realtime + ai-batch) with concurrency=1
 
-## ADR-023A Specific Rules
+## ADR-034 Model Stack (supersedes ADR-023A §2.1)
 
-- **2-Model Stack:** gemma4:e4b Q8_0 (~4.0GB) + nomic-embed-text (~0.3GB) = ~4.3GB VRAM peak
+- **3-Model Config:** typhoon2.5-np-dms:latest (Main) + typhoon-np-dms-ocr:latest (OCR, keep_alive:0) + nomic-embed-text (Embedding)
 - **PDF 3-Page Limit:** Classification/Tagging uses first 3 pages only (NOT RAG embedding)
 - **RAG Embedding:** Full document chunked at 512 tokens/64 tokens overlap
-- **OCR Auto-Detect:** PyMuPDF chars > 100 → Fast path, else PaddleOCR
+- **OCR Auto-Detect:** PyMuPDF chars > 100 → Fast path, else Tesseract OCR (with Typhoon OCR option)
 - **Embed Auto-Trigger:** AUTO after commit (parallel), gap covered by DB search
 - **Threshold Recalibration:** After 100-500 docs, based on ai_audit_logs analysis
 
@@ -129,7 +129,7 @@ const DocumentReviewForm = ({ document, aiSuggestions }) => {
 - [ ] BullMQ 2-queue setup (ai-realtime + ai-batch)
 - [ ] QdrantService with projectPublicId enforcement
 - [ ] DocumentReviewForm reusable component
-- [ ] Admin Desktop Ollama (gemma4:e4b Q8_0 + nomic-embed-text) + PaddleOCR setup
+- [ ] Admin Desktop Ollama (typhoon2.5-np-dms:latest + typhoon-np-dms-ocr:latest + nomic-embed-text) setup
 - [ ] n8n workflow orchestration (Migration Phase only)
 - [ ] AI audit logging and monitoring (ai_audit_logs)
 - [ ] Human-in-the-loop validation workflows
@@ -138,3 +138,8 @@ const DocumentReviewForm = ({ document, aiSuggestions }) => {
 
 - `specs/06-Decision-Records/ADR-023-unified-ai-architecture.md` (Base architecture)
 - `specs/06-Decision-Records/ADR-023A-unified-ai-architecture.md` (Model revision - current)
+- `specs/06-Decision-Records/ADR-024-intent-classification-strategy.md` (Pattern→LLM Fallback)
+- `specs/06-Decision-Records/ADR-025-ai-tool-layer-architecture.md` (Tool Registry)
+- `specs/06-Decision-Records/ADR-026-document-chat-ui-pattern.md` (Chat UI)
+- `specs/06-Decision-Records/ADR-027-ai-admin-console-and-dynamic-control.md` (Admin Console)
+- `specs/06-Decision-Records/ADR-028-migration-architecture-refactor.md` (Migration Pipeline)

.agents/rules/README.md

@@ -4,13 +4,14 @@ Critical rules and guidelines for AI agents working on LCBP3-DMS.
 
 ## Version
 
-- **Current:** v1.9.3
+- **Current:** v1.9.6
-- **Last Updated:** 2026-05-15
+- **Last Updated:** 2026-05-22
-- **Synced with:** `AGENTS.md` (v1.9.3)
+- **Synced with:** `AGENTS.md` (v1.9.6)
 
 ## Purpose
 
 This directory contains rule files that define:
+
 - Project context and role expectations
 - Critical Tier 1 rules (CI blockers)
 - Coding standards and patterns
@@ -24,6 +25,7 @@ This directory contains rule files that define:
 ### 🔴 Tier 1 — CRITICAL (CI BLOCKER)
 
 Build fails immediately if violated:
+
 - Security (Auth, RBAC, Validation)
 - UUID Strategy (ADR-019) — no `parseInt` / `Number` / `+` on UUID
 - Database correctness — verify schema before writing queries
@@ -35,15 +37,28 @@ Build fails immediately if violated:
 ### 🟡 Tier 2 — IMPORTANT (CODE REVIEW)
 
 Must fix before merge:
+
 - Architecture patterns (thin controller, business logic in service)
 - Test coverage (80%+ business logic, 70%+ backend overall)
 - Cache invalidation
 - Naming conventions
 - TypeScript Standards: Missing JSDoc, explicit types, or file headers
 
-### 🟢 Tier 3 — GUIDELINES
+### 🟢 Tier 3 — SPECIALIZED WORK
+
+Requires domain-specific knowledge:
+
+- **ADR-021 Integration:** Workflow Engine & Context implementation
+- **AI Infrastructure:** ADR-023/023A boundary enforcement and pipeline usage
+- **AI Runtime Layer:** ADR-024 Intent Classification, ADR-025 Tool Layer, ADR-026 Chat UI, ADR-027 Admin Console
+- **Migration Pipeline:** ADR-028 Staging Queue & post-migration cleanup
+- **Complex Business Logic:** Multi-step workflows with state management
+- **Performance Optimization:** Database queries, caching strategies, bulk operations
+
+### 🔵 Tier 4 — GUIDELINES
 
 Best practice — follow when possible:
+
 - Code style / formatting (Prettier handles)
 - Comment completeness
 - Minor optimizations
@@ -53,7 +68,7 @@ Best practice — follow when possible:
 ### Core Rules (Tier 1 - CRITICAL)
 
 | File                    | Purpose                                                                         |
-|------|---------|
+| ----------------------- | ------------------------------------------------------------------------------- |
 | `00-project-context.md` | Project context, role & persona, tier classification, specs folder organization |
 | `01-adr-019-uuid.md`    | UUID handling strategy — no parseInt, use publicId only                         |
 | `02-security.md`        | Security requirements, checklist, ADR-023/023A AI boundaries                    |
@@ -61,7 +76,7 @@ Best practice — follow when possible:
 ### Coding Standards
 
 | File                      | Purpose                                                 |
-|------|---------|
+| ------------------------- | ------------------------------------------------------- |
 | `03-typescript.md`        | TypeScript rules, file headers, i18n guidelines         |
 | `06-backend-patterns.md`  | NestJS patterns, UUID resolution, API response patterns |
 | `07-frontend-patterns.md` | Next.js patterns, RHF+Zod+TanStack Query, UUID handling |
@@ -69,14 +84,14 @@ Best practice — follow when possible:
 ### Domain & Workflow
 
 | File                       | Purpose                                                       |
-|------|---------|
+| -------------------------- | ------------------------------------------------------------- |
 | `04-domain-terminology.md` | DMS glossary, key spec files priority table                   |
 | `08-development-flow.md`   | Development workflow by work type (Critical/Normal/Quick Fix) |
 
 ### Compliance & Architecture
 
 | File                      | Purpose                                                        |
-|------|---------|
+| ------------------------- | -------------------------------------------------------------- |
 | `05-forbidden-actions.md` | Actions that must never be done, schema changes, UUID handling |
 | `09-commit-checklist.md`  | Pre-commit verification, commit message format                 |
 | `10-error-handling.md`    | ADR-007 error handling strategy, layered classification        |
@@ -87,13 +102,19 @@ Best practice — follow when possible:
 Spec priority: **`06-Decision-Records`** > **`05-Engineering-Guidelines`** > others
 
 | Document                       | Path                                                                        | Use When                          |
-|----------|------|----------|
+| ------------------------------ | --------------------------------------------------------------------------- | --------------------------------- |
 | **Glossary**                   | `specs/00-overview/00-02-glossary.md`                                       | Verify domain terminology         |
 | **Schema Tables**              | `specs/03-Data-and-Storage/lcbp3-v1.9.0-schema-02-tables.sql`               | Before writing any query          |
 | **Data Dictionary**            | `specs/03-Data-and-Storage/03-01-data-dictionary.md`                        | Field meanings + business rules   |
 | **Edge Cases**                 | `specs/01-Requirements/01-06-edge-cases-and-rules.md`                       | Prevent bugs in flows             |
 | **ADR-019 UUID**               | `specs/06-Decision-Records/ADR-019-hybrid-identifier-strategy.md`           | UUID-related work                 |
 | **ADR-023 AI**                 | `specs/06-Decision-Records/ADR-023-unified-ai-architecture.md`              | AI integration work               |
+| **ADR-023A AI Model**          | `specs/06-Decision-Records/ADR-023A-unified-ai-architecture.md`             | 2-model stack, BullMQ 2-queue     |
+| **ADR-024 Intent Class.**      | `specs/06-Decision-Records/ADR-024-intent-classification-strategy.md`       | Pattern→LLM Fallback; Redis cache |
+| **ADR-025 AI Tool Layer**      | `specs/06-Decision-Records/ADR-025-ai-tool-layer-architecture.md`           | Tool Registry; CASL-guarded       |
+| **ADR-026 Chat UI**            | `specs/06-Decision-Records/ADR-026-document-chat-ui-pattern.md`             | Side-panel; streaming SSE         |
+| **ADR-027 AI Admin Console**   | `specs/06-Decision-Records/ADR-027-ai-admin-console-and-dynamic-control.md` | Dynamic control; admin-only       |
+| **ADR-028 Migration Refactor** | `specs/06-Decision-Records/ADR-028-migration-architecture-refactor.md`      | Staging Queue; cleanup            |
 | **Backend Guidelines**         | `specs/05-Engineering-Guidelines/05-02-backend-guidelines.md`               | NestJS patterns                   |
 | **Frontend Guidelines**        | `specs/05-Engineering-Guidelines/05-03-frontend-guidelines.md`              | Next.js patterns                  |
 | **Testing Strategy**           | `specs/05-Engineering-Guidelines/05-04-testing-strategy.md`                 | Coverage goals                    |
@@ -103,10 +124,11 @@ Spec priority: **`06-Decision-Records`** > **`05-Engineering-Guidelines`** > oth
 When updating rules:
 
 1. **Check AGENTS.md version** — Ensure rule files are synced
-2. **Update version numbers** — Bump version in `00-project-context.md` and `03-typescript.md`
+2. **Update version numbers** — Bump version in `00-project-context.md` only (03-typescript.md no longer has version)
-3. **Review ADR references** — Ensure all ADR references are current (ADR-023, ADR-023A, etc.)
+3. **Review ADR references** — Ensure all ADR references are current (ADR-023, ADR-023A, ADR-024~028)
 4. **Add new forbidden actions** — When new patterns are identified as violations
 5. **Update key spec files table** — When new ADRs or guidelines are added
+6. **Update Tier 3 SPECIALIZED WORK** — When new domain-specific workflows are added
 
 ## Related Documents
 

.agents/scripts/bash/agent-registry.sh

@@ -42,7 +42,7 @@ init_agent_registry() {
         [qwen]="Qwen Code"
         [opencode]="opencode"
         [codex]="Codex CLI"
-        [windsurf]="Windsurf"
+        [devin]="Devin"
         [kilocode]="Kilo Code"
         [auggie]="Auggie CLI"
         [roo]="Roo Code"

.agents/scripts/bash/update-agent-context.sh

@@ -30,12 +30,12 @@
 #
 # 5. Multi-Agent Support
 #    - Handles agent-specific file paths and naming conventions
-#    - Supports: Claude, Gemini, Copilot, Cursor, Qwen, opencode, Codex, Windsurf, Kilo Code, Auggie CLI, Roo Code, CodeBuddy CLI, Qoder CLI, Amp, SHAI, or Amazon Q Developer CLI
+#    - Supports: Claude, Gemini, Copilot, Cursor, Qwen, opencode, Codex, Devin, Kilo Code, Auggie CLI, Roo Code, CodeBuddy CLI, Qoder CLI, Amp, SHAI, or Amazon Q Developer CLI
 #    - Can update single agents or all existing agent files
 #    - Creates default Claude file if no agent files exist
 #
 # Usage: ./update-agent-context.sh [agent_type]
-# Agent types: claude|gemini|copilot|cursor-agent|qwen|opencode|codex|windsurf|kilocode|auggie|shai|q|bob|qoder
+# Agent types: claude|gemini|copilot|cursor-agent|qwen|opencode|codex|devin|kilocode|auggie|shai|q|bob|qoder
 # Leave empty to update all existing agent files
 
 set -e
@@ -609,8 +609,8 @@ update_specific_agent() {
         codex)
             update_agent_file "$AGENTS_FILE" "Codex CLI"
             ;;
-        windsurf)
+        devin)
-            update_agent_file "$WINDSURF_FILE" "Windsurf"
+            update_agent_file "$DEVIN_FILE" "Devin"
             ;;
         kilocode)
             update_agent_file "$KILOCODE_FILE" "Kilo Code"
@@ -681,8 +681,8 @@ update_all_existing_agents() {
         found_agent=true
     fi
 
-    if [[ -f "$WINDSURF_FILE" ]]; then
+    if [[ -f "$DEVIN_FILE" ]]; then
-        update_agent_file "$WINDSURF_FILE" "Windsurf"
+        update_agent_file "$DEVIN_FILE" "Devin"
         found_agent=true
     fi
 

.agents/skills/README.md

@@ -1,8 +1,8 @@
 # `.agents/skills/` — LCBP3 Agent Skill Pack
 
-**Version:** 1.9.0 | **Last Updated:** 2026-05-17 | **Total Skills:** 23
+**Version:** 1.9.0 | **Last Updated:** 2026-06-07 | **Total Skills:** 24
 
-Agent skills for AI-assisted development in **Windsurf IDE** (and compatible agents: Codex CLI, opencode, Amp, Antigravity, AGENTS.md-aware tools).
+Agent skills for AI-assisted development in **Devin IDE** (and compatible agents: Codex CLI, opencode, Amp, Antigravity, AGENTS.md-aware tools).
 
 ---
 
@@ -14,6 +14,7 @@ Agent skills for AI-assisted development in **Windsurf IDE** (and compatible age
 ├── skills.md                    # Overview + dependency matrix + health monitoring
 ├── _LCBP3-CONTEXT.md            # Shared LCBP3 context injected into every speckit-* skill
 ├── README.md                    # (this file)
+├── save-memory/                 # Session log & project memory update
 ├── nestjs-best-practices/       # Backend rules (40 rules across 10 categories)
 ├── next-best-practices/         # Frontend rules (Next.js 15+)
 ├── e2e-testing/                 # Playwright E2E testing patterns (POM, flaky tests, CI/CD)
@@ -30,12 +31,10 @@ Each skill directory contains:
 
 ---
 
-## 🚀 How Windsurf Invokes These Skills
+## 🚀 How Devin Invokes These Skills
 
-Windsurf exposes two entry points:
+1. **Skill tool** — Devin discovers skills by scanning `.agents/skills/*/SKILL.md` frontmatter. Skills marked `user-invocable: false` are used silently by Cascade.
-
+2. **Slash commands** — `.devin/workflows/*.md` wraps each skill as a slash command (e.g. `/04-speckit.plan`). The workflow file is short; the heavy lifting is delegated to the skill via `skill` tool.
-1. **Skill tool** — Windsurf discovers skills by scanning `.agents/skills/*/SKILL.md` frontmatter. Skills marked `user-invocable: false` are used silently by Cascade.
-2. **Slash commands** — `.windsurf/workflows/*.md` wraps each skill as a slash command (e.g. `/04-speckit.plan`). The workflow file is short; the heavy lifting is delegated to the skill via `skill` tool.
 
 Both paths end up executing the same `SKILL.md` instructions.
 
@@ -66,13 +65,13 @@ Use `/00-speckit.all` to run specify → clarify → plan → tasks → analyze
 From repo root:
 
 | Script                                                 | Purpose                                                    |
-| --------------------------------------------------------- | ----------------------------------------------------------- |
+| ------------------------------------------------------ | ---------------------------------------------------------- |
 | `./.agents/scripts/bash/check-prerequisites.sh --json` | Emit `FEATURE_DIR` + `AVAILABLE_DOCS` for a feature branch |
 | `./.agents/scripts/bash/setup-plan.sh --json`          | Emit `FEATURE_SPEC`, `IMPL_PLAN`, `SPECS_DIR`, `BRANCH`    |
-| `./.agents/scripts/bash/update-agent-context.sh windsurf` | Append tech entries to `AGENTS.md`                          |
+| `./.agents/scripts/bash/update-agent-context.sh devin` | Append tech entries to `AGENTS.md`                         |
 | `./.agents/scripts/bash/audit-skills.sh`               | Validate all `SKILL.md` frontmatter + presence             |
 | `./.agents/scripts/bash/validate-versions.sh`          | Version consistency check                                  |
-| `./.agents/scripts/bash/sync-workflows.sh`                | Verify every skill has a `.windsurf/workflows/*.md` wrapper |
+| `./.agents/scripts/bash/sync-workflows.sh`             | Verify every skill has a `.devin/workflows/*.md` wrapper   |
 
 All scripts mirror to `.agents/scripts/powershell/*.ps1` for Windows.
 
@@ -97,7 +96,7 @@ To add a new skill:
 
 1. Create `NAME/SKILL.md` with frontmatter: `name`, `description`, `version: 1.9.0`, `scope`, `depends-on`.
 2. Append an LCBP3 context reference pointing to `_LCBP3-CONTEXT.md`.
-3. Wrap with `.windsurf/workflows/NAME.md` so it becomes a slash command.
+3. Wrap with `.devin/workflows/NAME.md` so it becomes a slash command.
 4. Update [`skills.md`](./skills.md) dependency matrix.
 5. Run `./.agents/scripts/bash/audit-skills.sh` → must pass.
 

.agents/skills/_LCBP3-CONTEXT.md

@@ -5,14 +5,14 @@
 
 **Project:** NAP-DMS (LCBP3) — Laem Chabang Port Phase 3 Document Management System
 **Stack:** NestJS 11 + Next.js 16 + TypeScript + MariaDB 11.8 + Redis + BullMQ + Elasticsearch + Ollama (on-prem AI)
-**Version:** 1.8.9 (2026-04-18)
+**Version:** 1.9.7 (2026-05-25)
 
 ---
 
 ## 📌 Canonical Rule Sources (read in this order)
 
 1. **`AGENTS.md`** (repo root) — primary rule file for AI agents; supersedes legacy `GEMINI.md`.
-2. **`specs/06-Decision-Records/`** — architectural decisions (22 ADRs); ADR priority > Engineering Guidelines.
+2. **`specs/06-Decision-Records/`** — architectural decisions (29 ADRs); ADR priority > Engineering Guidelines.
 3. **`specs/05-Engineering-Guidelines/`** — backend/frontend/testing/i18n/git patterns.
 4. **`specs/00-Overview/00-02-glossary.md`** — domain terminology (Correspondence / RFA / Transmittal / Circulation).
 5. **`specs/00-Overview/00-03-product-vision.md`** — project constitution (Vision, Strategic Pillars, Guardrails).
@@ -29,6 +29,7 @@
 - **ADR-002 Document Numbering:** Redis Redlock + TypeORM `@VersionColumn` (double-lock). Never use application-side counter alone.
 - **ADR-008 Notifications:** BullMQ queue — never inline email/notification in a request thread.
 - **ADR-023/023A AI Boundary:** Ollama on Admin Desktop only; AI → DMS API → DB (never direct DB/storage). 2-model stack: `gemma4:e4b Q8_0` + `nomic-embed-text`. BullMQ `ai-realtime` / `ai-batch` queues. Human-in-the-loop validation required. (ADR-018 superseded by ADR-023)
+- **ADR-029 Dynamic Prompt Management:** Prompt templates in DB (`ai_prompts`), never hardcoded in processor; Redis cache `ai:prompt:active:{type}` TTL 60s; `activate()` runs in DB transaction + Redis DEL after commit; `system.manage_all` guard on all mutations.
 - **ADR-007 Error Handling:** Layered (Validation / Business / System); `BusinessException` hierarchy; user-friendly `userMessage` + `recoveryAction`; technical stack only in logs.
 - **TypeScript Strict:** Zero `any`, zero `console.log` (use NestJS `Logger`).
 - **i18n:** No hardcoded Thai/English strings in components — use i18n keys (see `05-08-i18n-guidelines.md`).
@@ -54,7 +55,7 @@
 ## 📁 Key Files for Generating / Validating Artifacts
 
 | When you need...           | Read                                                                                                                |
-| ------------------------- | ------------------------------------------------------------------------------------------------------------------- |
+| -------------------------- | ------------------------------------------------------------------------------------------------------------------- |
 | A new feature spec         | `.agents/skills/speckit-specify/templates/spec-template.md` + `specs/01-Requirements/01-06-edge-cases-and-rules.md` |
 | A plan                     | `.agents/skills/speckit-plan/templates/plan-template.md` + relevant ADRs                                            |
 | Task breakdown             | `.agents/skills/speckit-tasks/templates/tasks-template.md` + existing patterns in `specs/08-Tasks/`                 |
@@ -62,6 +63,12 @@
 | Schema / table definition  | `specs/03-Data-and-Storage/lcbp3-v1.9.0-schema-02-tables.sql` + `03-01-data-dictionary.md`                          |
 | RBAC / permissions         | `specs/03-Data-and-Storage/lcbp3-v1.8.0-seed-permissions.sql` + `01-02-01-rbac-matrix.md`                           |
 | Release / hotfix           | `specs/04-Infrastructure-OPS/04-08-release-management-policy.md`                                                    |
+| ADR-024 Intent Class.      | `specs/06-Decision-Records/ADR-024-intent-classification-strategy.md`                                               |
+| ADR-025 AI Tool Layer      | `specs/06-Decision-Records/ADR-025-ai-tool-layer-architecture.md`                                                   |
+| ADR-026 Chat UI            | `specs/06-Decision-Records/ADR-026-document-chat-ui-pattern.md`                                                     |
+| ADR-027 AI Admin Console   | `specs/06-Decision-Records/ADR-027-ai-admin-console-and-dynamic-control.md`                                         |
+| ADR-028 Migration Refactor | `specs/06-Decision-Records/ADR-028-migration-architecture-refactor.md`                                              |
+| ADR-029 Dynamic Prompts    | `specs/06-Decision-Records/ADR-029-dynamic-prompt-management.md`                                                    |
 
 ---
 
@@ -83,7 +90,7 @@
 - [ ] Business comments in Thai, code identifiers in English
 - [ ] Schema changes via SQL directly (not migration)
 - [ ] Test coverage meets targets (Backend 70%+, Business Logic 80%+)
-- [ ] Relevant ADRs referenced (007/008/009/016/019/021/023/023A for AI work)
+- [ ] Relevant ADRs referenced (007/008/009/016/019/021/023/023A/024-029 for AI work)
 - [ ] Domain glossary terms used correctly
 - [ ] Error handling: `Logger` + `HttpException` / `BusinessException`
 - [ ] i18n keys used (no hardcode text)

.agents/skills/nestjs-best-practices/AGENTS.md

@@ -6454,7 +6454,7 @@ CREATE TABLE ai_audit_log (
   user_id INT NOT NULL,
   action VARCHAR(64) NOT NULL,       -- 'ai.extract_metadata', 'ai.classify', etc.
   file_id INT,
-  model VARCHAR(64),                  -- 'gemma-4:7b', 'paddleocr-v3'
+  model VARCHAR(64),                  -- 'gemma-4:7b', 'typhoon-np-dms-ocr', 'tesseract-ocr'
   confidence DECIMAL(4,3),
   input_hash CHAR(64),                -- SHA-256 of input for replay detection
   output_summary JSON,

.agents/skills/nestjs-best-practices/rules/lcbp3-ai-boundary.md

@@ -137,7 +137,7 @@ CREATE TABLE ai_audit_log (
   user_id INT NOT NULL,
   action VARCHAR(64) NOT NULL,       -- 'ai.extract_metadata', 'ai.classify', etc.
   file_id INT,
-  model VARCHAR(64),                  -- 'gemma-4:7b', 'paddleocr-v3'
+  model VARCHAR(64),                  -- 'gemma-4:7b', 'typhoon-np-dms-ocr', 'tesseract-ocr'
   confidence DECIMAL(4,3),
   input_hash CHAR(64),                -- SHA-256 of input for replay detection
   output_summary JSON,

.agents/skills/save-memory/SKILL.md

@@ -0,0 +1,198 @@
+---
+name: save-memory
+description: บันทึก session log และอัปเดต project memory ตามโครงสร้างใหม่
+version: 1.9.0
+scope: project-management
+depends-on: []
+user-invocable: true
+---
+
+# บันทึก Memory (Save Memory)
+
+Skill นี้ใช้สำหรับบันทึก session log และอัปเดต project memory ตามโครงสร้างใหม่ที่ reorganization แล้ว
+
+## โครงสร้าง Memory ใหม่
+
+```
+memory/
+├── README.md (index + overview)
+├── mcp-tools.md (MCP MariaDB + Memory Tools)
+└── project-memory-override.md (OS rules, Current Decisions, Environment, Next Session Focus)
+
+specs/88-logs/
+├── rollouts.md (Recent rollouts table)
+└── session-YYYY-MM-DD-[topic].md (Session logs)
+```
+
+## ขั้นตอนการบันทึก Memory
+
+### 1. สร้าง Session Log (ถ้ามีงาน session ใหม่)
+
+เมื่อทำงาน session ใหม่ให้:
+
+1. **สร้างไฟล์ session log ใหม่** ใน `specs/88-logs/`
+   - ชื่อไฟล์: `session-YYYY-MM-DD-[topic].md`
+   - ตัวอย่าง: `session-2026-06-07-memory-reorganization.md`
+
+2. **บันทึกเนื้อหาใน session log**:
+
+   ```markdown
+   # Session [N] — YYYY-MM-DD ([Topic])
+
+   ## Summary
+
+   [สรุปสิ่งที่ทำใน session นี้]
+
+   ## ปัญหาที่พบ (Root Cause)
+
+   [อธิบายปัญหาและสาเหตุ]
+
+   ## การแก้ไข (Fix)
+
+   | ไฟล์           | การเปลี่ยนแปลง         |
+   | -------------- | ---------------------- |
+   | [path/to/file] | [อธิบายการเปลี่ยนแปลง] |
+
+   ## กฎที่ Lock แล้ว
+
+   [บันทึก pattern หรือ decision ที่ตกลง]
+
+   ## Verification
+
+   [วิธีตรวจสอบว่างานสำเร็จ]
+   ```
+
+3. **อัปเดต `specs/88-logs/rollouts.md`**
+   - เพิ่ม entry ใหม่ในตาราง Recent Rollouts
+   - รูปแบบ: `| วันที่ | Version | รายการ | สถานะ |`
+
+### 2. อัปเดต Project Memory (ถ้ามี decision ใหม่)
+
+เมื่อมีการตัดสินใจสำคัญใหม่ให้:
+
+1. **เปิดไฟล์ `memory/project-memory-override.md`**
+
+2. **อัปเดตตาราง "Current Decisions (Locked)"**
+   - เพิ่ม entry ใหม่ถ้ามี decision ใหม่
+   - รูปแบบ: `| ID | Decision | ADR |`
+
+3. **อัปเดต "Next Session Focus"**
+   - เพิ่มงานใหม่ถ้ามี
+   - ทำเครื่องหมาย `[ ]` สำหรับงานที่ยังไม่เสร็จ
+   - ทำเครื่องหมาย `[X]` สำหรับงานที่เสร็จแล้ว
+
+4. **อัปเดต "Environment & Services"** (ถ้ามีการเปลี่ยนแปลง)
+   - อัปเดต URL, port, หรือ notes ถ้ามีการเปลี่ยน infrastructure
+
+### 3. อัปเดต MCP Tools (ถ้ามี tools ใหม่)
+
+เมื่อมี MCP tools ใหม่ให้:
+
+1. **เปิดไฟล์ `memory/mcp-tools.md`**
+
+2. **เพิ่ม tool ใหม่ในตาราง "Available Tools"**
+   - รูปแบบ: `| Tool | Purpose | Example Usage |`
+
+3. **เพิ่ม usage example และ warnings** ถ้าจำเป็น
+
+### 4. อัปเดต Root Documentation (ถ้ามีการเปลี่ยนแปลง)
+
+เมื่อมีการเปลี่ยนแปลงที่ส่งผลต่อเอกสารระดับ root ให้:
+
+1. **ARCHITECTURE.md** — อัปเดตเมื่อ:
+   - เปลี่ยน architecture หลัก
+   - เพิ่ม/ลบ component สำคัญ
+   - เปลี่ยน data flow หรือ integration pattern
+
+2. **CHANGELOG.md** — อัปเดตเมื่อ:
+   - Deploy version ใหม่
+   - เพิ่ม feature หรือ breaking change สำคัญ
+   - รูปแบบ: `## [version] (YYYY-MM-DD)` → `### feat(scope): description`
+
+3. **CONTEXT.md** — อัปเดตเมื่อ:
+   - เปลี่ยน domain terminology หลัก
+   - เพิ่ม concept ใหม่ที่ใช้ทั่ว project
+   - อัปเดต glossary หรือ business rules
+
+4. **CONTRIBUTING.md** — อัปเดตเมื่อ:
+   - เปลี่ยน workflow การทำงาน
+   - เพิ่ม/เปลี่ยน coding standards
+   - อัปเดต CI/CD process
+
+5. **README.md** — อัปเดตเมื่อ:
+   - เปลี่ยน project structure
+   - เพิ่ม/เปลี่ยน installation steps
+   - อัปเดต feature overview หรือ tech stack
+
+## Template สำหรับ Session Log
+
+```markdown
+# Session [N] — YYYY-MM-DD ([Topic])
+
+## Summary
+
+[สรุปสิ่งที่ทำใน session นี้ใน 1-2 ประโยค]
+
+## ปัญหาที่พบ (Root Cause)
+
+[อธิบายปัญหาและสาเหตุหลัก]
+
+## การแก้ไข (Fix)
+
+| ไฟล์           | การเปลี่ยนแปลง         |
+| -------------- | ---------------------- |
+| `path/to/file` | [อธิบายการเปลี่ยนแปลง] |
+
+## กฎที่ Lock แล้ว
+
+[บันทึก pattern หรือ decision ที่ตกลงและไม่ควรเปลี่ยน]
+
+## Verification
+
+- [ ] [check 1]
+- [ ] [check 2]
+```
+
+## ข้อควรระวัง
+
+- **ห้าม** บันทึก rules ที่ซ้ำกับ specs/ (ADRs, glossary, guidelines)
+- **ห้าม** บันทึก commands ที่ซ้ำกับ specs/05-Engineering-Guidelines/
+- **ห้าม** บันทึก environment ที่ซ้ำกับ specs/04-Infrastructure-OPS/
+- **ใช้** `specs/88-logs/` สำหรับ session history และ rollouts
+- **ใช้** `memory/project-memory-override.md` สำหรับ OS rules, decisions, environment ที่ไม่มีใน specs
+- **ใช้** `memory/mcp-tools.md` สำหรับ MCP tools documentation
+- **อัปเดต Root Documentation** (ARCHITECTURE.md, CHANGELOG.md, CONTEXT.md, CONTRIBUTING.md, README.md) เฉพาะเมื่อมีการเปลี่ยนแปลงที่ส่งผลต่อ project architecture, version, terminology, workflow หรือ structure
+
+## ตัวอย่างการใช้งาน
+
+### กรณีที่ 1: ทำงาน session ใหม่
+
+```
+1. สร้างไฟล์ specs/88-logs/session-2026-06-07-bug-fix.md
+2. บันทึกปัญหา, การแก้ไข, verification
+3. อัปเดต specs/88-logs/rollouts.md
+```
+
+### กรณีที่ 2: มี decision ใหม่
+
+```
+1. เปิด memory/project-memory-override.md
+2. เพิ่ม entry ใหม่ในตาราง Current Decisions
+3. อัปเดต Next Session Focus
+```
+
+### กรณีที่ 3: เปลี่ยน infrastructure
+
+```
+1. เปิด memory/project-memory-override.md
+2. อัปเดตตาราง Environment & Services
+3. อัปเดต Key Environment Variables ถ้าจำเป็น
+```
+
+### กรณีที่ 4: อัปเดต Root Documentation
+
+```
+1. ตรวจสอบว่ามีการเปลี่ยนแปลงที่ส่งผลต่อ ARCHITECTURE.md, CHANGELOG.md, CONTEXT.md, CONTRIBUTING.md, หรือ README.md
+2. อัปเดตไฟล์ที่เกี่ยวข้องตามรูปแบบที่กำหนด
+3. ตรวจสอบว่าการเปลี่ยนแปลงสอดคล้องกับ specs/ และ ADRs
+```

.agents/skills/skills.md

@@ -2,7 +2,7 @@
 
 ไฟล์นี้กำหนดทักษะและความสามารถเฉพาะทางของ Document Intelligence Engine สำหรับโครงการ LCBP3 v1.9.0 เพื่อรักษามาตรฐานสูงสุดด้าน Security และ Data Integrity
 
-**Status**: Production Ready | **Last Updated**: 2026-05-17 | **Total Skills**: 23
+**Status**: Production Ready | **Last Updated**: 2026-06-07 | **Total Skills**: 24
 
 > 📌 Shared context for all speckit-\* skills: see [`_LCBP3-CONTEXT.md`](./_LCBP3-CONTEXT.md).
 
@@ -76,6 +76,7 @@
 | **speckit-status**         | None                 | None                                     | Progress tracking             |
 | **speckit-taskstoissues**  | speckit-tasks        | None                                     | Issue sync                    |
 | **speckit-checklist**      | speckit-plan         | None                                     | Requirements validation       |
+| **save-memory**            | None                 | None                                     | Session log & memory update   |
 | **nestjs-best-practices**  | None                 | speckit-implement                        | Backend patterns              |
 | **next-best-practices**    | None                 | speckit-implement                        | Frontend patterns             |
 | **speckit-security-audit** | None                 | speckit-reviewer                         | Security validation           |
@@ -99,7 +100,7 @@
 
 ### Health Metrics
 
-- **Total Skills**: 23 implemented
+- **Total Skills**: 24 implemented
 - **Version Alignment**: v1.9.0 across all skills
 - **Template Coverage**: 100% for skills requiring templates
 - **Documentation**: Complete front matter + shared `_LCBP3-CONTEXT.md` appendix

.agents/tests/skill-integration.test.js

@@ -19,7 +19,7 @@ class SkillTestSuite {
     this.results = {
       passed: 0,
       failed: 0,
-            errors: []
+      errors: [],
     };
   }
 
@@ -29,7 +29,7 @@ class SkillTestSuite {
       pass: '\x1b[32m', // Green
       fail: '\x1b[31m', // Red
       warn: '\x1b[33m', // Yellow
-            reset: '\x1b[0m'
+      reset: '\x1b[0m',
     };
 
     const color = colors[type] || colors.info;
@@ -119,11 +119,20 @@ function runAllTests() {
 
   // Test 3: Script Files
   testSuite.log('Test 3: Validation Scripts', 'info');
-    testSuite.testFileExists(path.join(AGENTS_DIR, 'scripts', 'bash', 'validate-versions.sh'), 'bash validate-versions.sh');
+  testSuite.testFileExists(
+    path.join(AGENTS_DIR, 'scripts', 'bash', 'validate-versions.sh'),
+    'bash validate-versions.sh'
+  );
   testSuite.testFileExists(path.join(AGENTS_DIR, 'scripts', 'bash', 'audit-skills.sh'), 'bash audit-skills.sh');
   testSuite.testFileExists(path.join(AGENTS_DIR, 'scripts', 'bash', 'sync-workflows.sh'), 'bash sync-workflows.sh');
-    testSuite.testFileExists(path.join(AGENTS_DIR, 'scripts', 'powershell', 'validate-versions.ps1'), 'powershell validate-versions.ps1');
+  testSuite.testFileExists(
-    testSuite.testFileExists(path.join(AGENTS_DIR, 'scripts', 'powershell', 'audit-skills.ps1'), 'powershell audit-skills.ps1');
+    path.join(AGENTS_DIR, 'scripts', 'powershell', 'validate-versions.ps1'),
+    'powershell validate-versions.ps1'
+  );
+  testSuite.testFileExists(
+    path.join(AGENTS_DIR, 'scripts', 'powershell', 'audit-skills.ps1'),
+    'powershell audit-skills.ps1'
+  );
   testSuite.log('');
 
   // Test 4: Version Consistency
@@ -131,21 +140,29 @@ function runAllTests() {
   testSuite.testFileContent(path.join(AGENTS_DIR, 'README.md'), /v1\.8\.6/, 'README.md version');
   testSuite.testFileContent(path.join(SKILLS_DIR, 'VERSION'), /version: 1\.8\.6/, 'skills VERSION file');
   testSuite.testFileContent(path.join(SKILLS_DIR, 'skills.md'), /v1\.8\.6/, 'skills.md version');
-    testSuite.testFileContent(path.join(AGENTS_DIR, 'rules', '00-project-context.md'), /v1\.8\.6/, 'project context version');
+  testSuite.testFileContent(
+    path.join(AGENTS_DIR, 'rules', '00-project-context.md'),
+    /v1\.8\.6/,
+    'project context version'
+  );
   testSuite.log('');
 
   // Test 5: Skills Structure
   testSuite.log('Test 5: Skills Structure', 'info');
-    const skillDirs = fs.readdirSync(SKILLS_DIR).filter(item => {
+  const skillDirs = fs.readdirSync(SKILLS_DIR).filter((item) => {
     const itemPath = path.join(SKILLS_DIR, item);
-        return fs.statSync(itemPath).isDirectory() && item.startsWith('speckit-') || item === 'nestjs-best-practices' || item === 'next-best-practices';
+    return (
+      (fs.statSync(itemPath).isDirectory() && item.startsWith('speckit-')) ||
+      item === 'nestjs-best-practices' ||
+      item === 'next-best-practices'
+    );
   });
 
   testSuite.assert(skillDirs.length >= 20, `Found at least 20 skill directories (found ${skillDirs.length})`);
 
   // Test a few key skills
   const keySkills = ['speckit-plan', 'speckit-implement', 'speckit-specify', 'speckit-validate'];
-    keySkills.forEach(skill => {
+  keySkills.forEach((skill) => {
     const skillPath = path.join(SKILLS_DIR, skill);
     const skillMdPath = path.join(skillPath, 'SKILL.md');
     testSuite.testDirectoryExists(skillPath, `${skill} directory`);
@@ -163,12 +180,12 @@ function runAllTests() {
 
   // Test 6: Workflows Structure
   testSuite.log('Test 6: Workflows Structure', 'info');
-    const workflowFiles = fs.readdirSync(WORKFLOWS_DIR).filter(item => item.endsWith('.md'));
+  const workflowFiles = fs.readdirSync(WORKFLOWS_DIR).filter((item) => item.endsWith('.md'));
   testSuite.assert(workflowFiles.length >= 20, `Found at least 20 workflow files (found ${workflowFiles.length})`);
 
   // Test key workflows
   const keyWorkflows = ['00-speckit.all.md', '02-speckit.specify.md', '04-speckit.plan.md', '07-speckit.implement.md'];
-    keyWorkflows.forEach(workflow => {
+  keyWorkflows.forEach((workflow) => {
     const workflowPath = path.join(WORKFLOWS_DIR, workflow);
     testSuite.testFileExists(workflowPath, `${workflow} file`);
   });
@@ -177,12 +194,12 @@ function runAllTests() {
   // Test 7: Rules Structure
   testSuite.log('Test 7: Rules Structure', 'info');
   const rulesDir = path.join(AGENTS_DIR, 'rules');
-    const ruleFiles = fs.readdirSync(rulesDir).filter(item => item.endsWith('.md'));
+  const ruleFiles = fs.readdirSync(rulesDir).filter((item) => item.endsWith('.md'));
   testSuite.assert(ruleFiles.length >= 10, `Found at least 10 rule files (found ${ruleFiles.length})`);
 
   // Test key rules
   const keyRules = ['00-project-context.md', '01-adr-019-uuid.md', '02-security.md'];
-    keyRules.forEach(rule => {
+  keyRules.forEach((rule) => {
     const rulePath = path.join(rulesDir, rule);
     testSuite.testFileExists(rulePath, `${rule} file`);
   });
@@ -209,9 +226,21 @@ function runAllTests() {
 
   // Test 9: Documentation Quality
   testSuite.log('Test 9: Documentation Quality', 'info');
-    testSuite.testFileContent(path.join(AGENTS_DIR, 'README.md'), /## Troubleshooting/, 'README.md has troubleshooting section');
+  testSuite.testFileContent(
-    testSuite.testFileContent(path.join(SKILLS_DIR, 'skills.md'), /## Skill Dependency Matrix/, 'skills.md has dependency matrix');
+    path.join(AGENTS_DIR, 'README.md'),
-    testSuite.testFileContent(path.join(AGENTS_DIR, 'README.md'), /## Architecture/, 'README.md has architecture section');
+    /## Troubleshooting/,
+    'README.md has troubleshooting section'
+  );
+  testSuite.testFileContent(
+    path.join(SKILLS_DIR, 'skills.md'),
+    /## Skill Dependency Matrix/,
+    'skills.md has dependency matrix'
+  );
+  testSuite.testFileContent(
+    path.join(AGENTS_DIR, 'README.md'),
+    /## Architecture/,
+    'README.md has architecture section'
+  );
   testSuite.log('');
 
   // Results Summary
@@ -221,7 +250,7 @@ function runAllTests() {
 
   if (testSuite.results.errors.length > 0) {
     testSuite.log('Errors:', 'fail');
-        testSuite.results.errors.forEach(error => {
+    testSuite.results.errors.forEach((error) => {
       testSuite.log(`  - ${error}`, 'fail');
     });
   }

.agents/tests/workflow-validation.test.js

@@ -17,7 +17,7 @@ class WorkflowTestSuite {
     this.results = {
       passed: 0,
       failed: 0,
-            errors: []
+      errors: [],
     };
   }
 
@@ -27,7 +27,7 @@ class WorkflowTestSuite {
       pass: '\x1b[32m', // Green
       fail: '\x1b[31m', // Red
       warn: '\x1b[33m', // Yellow
-            reset: '\x1b[0m'
+      reset: '\x1b[0m',
     };
 
     const color = colors[type] || colors.info;
@@ -67,7 +67,7 @@ class WorkflowTestSuite {
 
       // Check for proper markdown formatting
       const lines = content.split('\n');
-            const nonEmptyLines = lines.filter(line => line.trim().length > 0);
+      const nonEmptyLines = lines.filter((line) => line.trim().length > 0);
       this.assert(nonEmptyLines.length >= 5, `${expectedName} has sufficient content`);
 
       return true;
@@ -121,7 +121,7 @@ const expectedWorkflows = {
   'util-speckit.migrate.md': 'Migration utility',
   'util-speckit.quizme.md': 'Quiz utility',
   'util-speckit.status.md': 'Status utility',
-    'util-speckit.taskstoissues.md': 'Task to issues utility'
+  'util-speckit.taskstoissues.md': 'Task to issues utility',
 };
 
 // Test suite implementation
@@ -169,7 +169,6 @@ function runWorkflowTests() {
 
         // Validate skill dependencies
         workflowTestSuite.validateWorkflowDependency(filename, content);
-                
       } catch (error) {
         workflowTestSuite.assert(false, `${filename} - content validation error: ${error.message}`);
       }
@@ -179,7 +178,7 @@ function runWorkflowTests() {
 
   // Test 4: Workflow naming consistency
   workflowTestSuite.log('Test 4: Naming Consistency', 'info');
-    const actualFiles = fs.readdirSync(WORKFLOWS_DIR).filter(file => file.endsWith('.md'));
+  const actualFiles = fs.readdirSync(WORKFLOWS_DIR).filter((file) => file.endsWith('.md'));
 
   for (const actualFile of actualFiles) {
     if (!expectedWorkflows[actualFile]) {
@@ -211,11 +210,14 @@ function runWorkflowTests() {
   // Results Summary
   workflowTestSuite.log('=== Workflow Test Results Summary ===', 'info');
   workflowTestSuite.log(`Passed: ${workflowTestSuite.results.passed}`, 'pass');
-    workflowTestSuite.log(`Failed: ${workflowTestSuite.results.failed}`, workflowTestSuite.results.failed > 0 ? 'fail' : 'pass');
+  workflowTestSuite.log(
+    `Failed: ${workflowTestSuite.results.failed}`,
+    workflowTestSuite.results.failed > 0 ? 'fail' : 'pass'
+  );
 
   if (workflowTestSuite.results.errors.length > 0) {
     workflowTestSuite.log('Errors:', 'fail');
-        workflowTestSuite.results.errors.forEach(error => {
+    workflowTestSuite.results.errors.forEach((error) => {
       workflowTestSuite.log(`  - ${error}`, 'fail');
     });
   }

.agents/workflows/save-memory.md

@@ -0,0 +1,11 @@
+---
+description: บันทึก session log และอัปเดต project memory
+---
+
+# บันทึก Memory
+
+ใช้ skill `save-memory` เพื่อบันทึก session log และอัปเดต project memory ตามโครงสร้างใหม่
+
+```bash
+skill save-memory
+```

.devin/rules/00-project-context.md

@@ -0,0 +1,168 @@
+---
+trigger: always_on
+---
+
+# NAP-DMS Project Context & Rules
+
+- For: Devin Cascade (and compatible: Codex CLI, opencode, Amp, Antigravity, AGENTS.md tools)
+- Version: 1.9.10 | Last synced from repo: 2026-06-06
+- Repo: [https://git.np-dms.work/np-dms/lcbp3](https://git.np-dms.work/np-dms/lcbp3)
+- Skill pack: `.agents/skills/` (v1.9.0, 21 skills) — see [`skills/README.md`](./.agents/skills/README.md) + [`skills/_LCBP3-CONTEXT.md`](./.agents/skills/_LCBP3-CONTEXT.md)
+
+## 🧠 Role & Persona
+
+Act as **Senior Full Stack Developer** specialized in NestJS, Next.js, TypeScript, DMS. Focus: Data Integrity, Security, Maintainability, Performance.
+
+You are a **Document Intelligence Engine** — not a general chatbot. Every response must be **precise**, **spec-compliant**, and **production-ready**.
+
+---
+
+## 🧩 Thought & Planning Protocol (Powered by Everything-Claude-Code)
+
+Before writing any code or taking any action in Tier 1 and Tier 2, the AI must demonstrate the following thinking process:
+
+### 1. Analysis Phase (Explore & Analyze)
+
+Problem Understanding: Restate what the user wants in clear, unambiguous terms.
+Context Search: Identify the relevant Spec files or ADRs from the "Key Spec Files" table that must be read before starting.
+Constraints Identification: Identify key constraints (e.g. Security rules, UUID patterns, or Domain terminology).
+
+### 2. Planning Phase (Plan)
+
+Alternative Exploration: Present at least 2 solution approaches (where possible) with pros/cons analysis.
+Step-by-Step Roadmap: Write a file-by-file plan of changes before executing.
+Verification Plan: Specify how to verify the work is complete (e.g. "which unit tests to write" or "which file to check the schema in").
+
+### 3. Execution & Refinement (Execute & Refine)
+
+Follow the plan step by step, and pause to ask if any uncertainty arises.
+If significant logic changes are made, summarize what was done for the user after completion.
+
+---
+
+## ⚙️ DMS Workflow Engine Protocol
+
+กฎนี้ใช้คุม Logic การไหลของเอกสาร (RFA, Transmittal, Correspondence) เพื่อป้องกัน Race Condition และรักษาความถูกต้องของสถานะ:
+
+- **State Management:** ตรวจสอบสถานะปัจจุบันจาก DB ก่อนเสมอ เพื่อป้องกันการอนุมัติซ้ำซ้อน (ดู `05-06-code-snippets.md` `[workflow-transition]`)
+- **Concurrency Control:** การจอนเลขที่เอกสารต้องใช้ **Redis Redlock** หรือ **TypeORM `@VersionColumn`** เท่านั้น (ADR-002)
+- **Background Jobs:** งานนานหรือการแจ้งเตือนต้องส่งไปทำที่ **BullMQ** ห้ามเขียนแบบ Inline (ADR-008)
+- **Term Consistency:** ห้ามใช้ "Approval Flow" ให้ใช้ **"Workflow Engine"** และห้ามใช้ "Letter" ให้ใช้ **"Correspondence"** (หมายเหตุ: "จดหมาย" ในคอมเมนต์ภาษาไทย = Correspondence ที่ครอบคลุมทุกประเภท)
+
+---
+
+## 🛡️ Security & Integrity Audit Protocol
+
+กฎนี้ให้ AI เป็น Gatekeeper ก่อน Commit โดยเน้น **Tier 1 — CRITICAL**:
+
+- **UUID Validation:** ตรวจสอบว่าเป็น **UUIDv7** และห้ามใช้ `parseInt()` บน UUID (ADR-019)
+- **RBAC Check:** API ใหม่ต้องมี **CASL Guard** และตรวจสอบ 4-Level RBAC Matrix (ADR-016)
+- **Data Isolation:** AI ต้องรันผ่าน **Ollama บน Admin Desktop** เท่านั้น ห้ามเข้าถึง DB/storage โดยตรง (ADR-023)
+- **Input Sanitization:** ไฟล์อัปโหลดต้องผ่าน **Two-Phase** (Temp → Commit) และสแกนด้วย **ClamAV** (ADR-016)
+
+---
+
+## 🧭 Rule Enforcement Tiers
+
+### 🔴 Tier 1 — CRITICAL (CI BLOCKER)
+
+Build fails หากละเมิด:
+
+- Security (Auth, RBAC, Validation)
+- UUID Strategy (ADR-019) — no `parseInt` / `Number` / `+` on UUID
+- Database correctness — verify schema before writing queries
+- File upload security (ClamAV + whitelist)
+- AI validation boundary (ADR-023)
+- Error handling strategy (ADR-007)
+- Forbidden patterns: `any`, `console.log`, UUID misuse, `id ?? ''` fallback
+
+### 🟡 Tier 2 — IMPORTANT (CODE REVIEW)
+
+Must fix ก่อน merge:
+
+- Architecture patterns (thin controller, business logic in service)
+- Test coverage (80%+ business logic, 70%+ backend overall)
+- Cache invalidation
+- Naming conventions
+- **TypeScript Standards:** Missing JSDoc, explicit types, or file headers
+
+### 🟢 Tier 3 — SPECIALIZED WORK
+
+Requires domain-specific knowledge:
+
+- **ADR-021 Integration:** Workflow Engine & Context implementation
+- **AI Infrastructure:** ADR-023/023A boundary enforcement and pipeline usage
+- **AI Runtime Layer:** ADR-024 Intent Classification, ADR-025 Tool Layer, ADR-026 Chat UI, ADR-027 Admin Console
+- **Migration Pipeline:** ADR-028 Staging Queue & post-migration cleanup
+- **Complex Business Logic:** Multi-step workflows with state management
+- **Performance Optimization:** Database queries, caching strategies, bulk operations
+
+### 🔵 Tier 4 — GUIDELINES
+
+Best practice — follow when possible:
+
+- Code style / formatting (Prettier handles)
+- Comment completeness
+- Minor optimizations
+
+---
+
+## 📐 TypeScript Rules & Coding Standards
+
+### 📝 Core Standards
+
+- **Strict Mode** — all strict checks enforced.
+- **ZERO `any` types** — use proper types or `unknown` + narrowing.
+- **ZERO `console.log`** — use NestJS `Logger` (backend) or remove (frontend).
+- **English for Code** — use English for all code identifiers, variables, and logic.
+- **Thai for Comments** — use Thai for comments, documentation, and JSDoc.
+- **Explicit Typing** — explicitly define types for all variables, parameters, and return values.
+- **JSDoc** — use JSDoc for all public classes and methods.
+
+### 🏗️ File & Function Structure
+
+- **File Headers** — every file MUST start with `// File: path/filename` on the first line.
+- **Change Log** — include `// Change Log` at the top of the file to track modifications.
+- **Single Export** — export **only one main symbol** per file.
+- **Function Style** — avoid unnecessary blank lines inside functions.
+
+---
+
+## 🚫 Forbidden Actions
+
+| ❌ Forbidden                                    | ✅ Correct Approach                                     | ⚠️ Why                                               |
+| ----------------------------------------------- | ------------------------------------------------------- | ---------------------------------------------------- |
+| SQL Triggers for business logic                 | NestJS Service methods                                  | Untestable; bypasses audit log                       |
+| `.env` files in production                      | `docker-compose.yml` environment section                | Secrets exposed in version control                   |
+| TypeORM migration files                         | Edit schema SQL directly (ADR-009)                      | Migration drift risk; schema managed via SQL delta   |
+| Inventing table/column names                    | Verify against `schema-02-tables.sql`                   | Schema mismatch causes silent runtime errors         |
+| `any` TypeScript type                           | Proper types / generics                                 | Defeats strict mode; hides runtime type errors       |
+| `console.log` in committed code                 | NestJS Logger (backend) / remove (frontend)             | Log flooding in production; risk of data leakage     |
+| `req: any` in controllers                       | `RequestWithUser` typed interface                       | Type safety lost; auth context unreachable           |
+| `parseInt()` on UUID values                     | Use UUID string directly (ADR-019)                      | `"0195…"` parsed to integer `19` — silently wrong    |
+| Exposing INT PK in API responses                | UUIDv7 `publicId` (ADR-019)                             | Leaks row count; enables DB enumeration attacks      |
+| AI accessing DB/storage directly                | AI → DMS API → DB (ADR-023/023A)                        | Bypasses RBAC, audit trail, and validation layer     |
+| Direct file operations bypassing StorageService | `StorageService` for all file moves                     | Orphaned files; broken ClamAV scan; no audit trail   |
+| Inline email/notification sending               | BullMQ queue job (ADR-008)                              | Blocks request thread; no retry on transient failure |
+| Deploying without Release Gates                 | Complete `04-08-release-management-policy.md`           | Unverified deploy risks data loss in production      |
+| AI direct cloud API calls                       | On-premises Ollama only (ADR-023/023A)                  | Data privacy violation; no audit control             |
+| AI outputs without human validation             | Human-in-the-loop validation required (ADR-023/023A)    | Unvalidated AI metadata corrupts document records    |
+| n8n calling Ollama/Qdrant directly              | n8n → DMS API → BullMQ → Ollama (ADR-023A)              | Bypasses audit log, RBAC, and error handling layer   |
+| Qdrant query without `projectPublicId` filter   | `QdrantService.search(projectPublicId, ...)` (ADR-023A) | Cross-project data leak via vector search            |
+
+---
+
+## 🚧 Out of Scope — Never Do Without Explicit Approval
+
+| ❌ Never Do Autonomously                                        | ⚠️ Why Approval Is Required                                      |
+| --------------------------------------------------------------- | ---------------------------------------------------------------- |
+| `DROP` or `RENAME` a column / table                             | Irreversible data loss — requires DBA + PM sign-off              |
+| Push directly to `main` / `master` branch                       | Bypasses CI, code review, and release gates                      |
+| Generate or insert seed data into production database           | May corrupt live data or violate business state invariants       |
+| Delete files from permanent storage                             | Files may be referenced in active documents or audit trails      |
+| Modify RBAC permission matrix without security team approval    | Defines access control for all users — security boundary change  |
+| Upgrade major library versions (NestJS, Next.js, TypeORM, etc.) | Breaking changes require full regression test cycle              |
+| Disable or modify authentication / authorization guards         | Creates unguarded endpoints — immediate security risk            |
+| Change Redis lock TTL or disable Redlock                        | Risk of document number race condition (ADR-002)                 |
+| Create or supersede an ADR unilaterally                         | Architecture decisions require team consensus and review process |
+| Add new columns to production tables without schema review      | Must update Data Dictionary + downstream queries simultaneously  |

.devin/rules/02-security.md

@@ -13,9 +13,9 @@ trigger: always_on
 5. **Password:** bcrypt 12 salt rounds, min 8 chars, rotate every 90 days
 6. **Rate Limiting:** `ThrottlerGuard` on all auth endpoints
 7. **File Upload:** Whitelist PDF/DWG/DOCX/XLSX/ZIP, max 50MB, ClamAV scan
-8. **AI Isolation (ADR-018):** Ollama on Admin Desktop ONLY — NO direct DB/storage access
+8. **AI Isolation (ADR-023/023A):** Ollama on Admin Desktop ONLY — NO direct DB/storage access; 2-model stack `gemma4:e4b Q8_0` + `nomic-embed-text`; all inference via BullMQ (`ai-realtime` / `ai-batch`)
 9. **Error Handling (ADR-007):** Use layered error classification with user-friendly messages
-10. **AI Integration (ADR-020):** RFA-First approach with unified pipeline architecture
+10. **AI Integration (ADR-023/023A):** RFA-First approach; n8n orchestrates Migration Phase only via DMS API — never calls Ollama directly; `QdrantService.search()` requires `projectPublicId` as mandatory param
 11. **AI Audit Trail:** Log all AI interactions and human validations
 12. **Rate Limiting:** Apply to AI endpoints to prevent abuse
 
@@ -30,7 +30,7 @@ trigger: always_on
 - [ ] No SQL injection vulnerabilities
 - [ ] File upload validation (whitelist + ClamAV)
 - [ ] Rate limiting applied to auth endpoints
-- [ ] AI boundary enforcement (ADR-023) - no direct DB/storage access
+- [ ] AI boundary enforcement (ADR-023/023A) - no direct DB/storage access
 - [ ] AI audit logging implemented for AI interactions
 - [ ] AI outputs validated before use (human-in-the-loop)
 - [ ] Error handling follows ADR-007 layered classification

.devin/rules/03-typescript.md

@@ -1,4 +1,8 @@
-# TypeScript Rules (v1.9.3)
+---
+trigger: always_on
+---
+
+# TypeScript Rules
 
 ## Core Standards
 

.devin/rules/04-domain-terminology.md

@@ -0,0 +1,45 @@
+---
+trigger: always_on
+---
+
+# Domain Terminology
+
+## DMS Glossary
+
+| ✅ Use             | ❌ Don't Use                          | คำอธิบายเพิ่มเติม                                |
+| ------------------ | ------------------------------------- | ------------------------------------------------ |
+| Correspondence     | Letter, Communication, Document       | ครอบคลุมทุกประเภท: Letter, RFA, Memo, ฯลฯ        |
+| RFA                | Approval Request, Submit for Approval | เอกสารขออนุมัติ (ชนิดหนึ่งของ Correspondence)    |
+| Transmittal        | Delivery Note, Cover Letter           | เอกสารนำส่ง (ชนิดหนึ่งของ Correspondence)        |
+| Circulation        | Distribution, Routing                 | ใบเวียนเอกสารภายใน (ชนิดหนึ่งของ Correspondence) |
+| Shop Drawing       | Construction Drawing                  | แบบก่อสร้าง                                      |
+| Contract Drawing   | Design Drawing, Blueprint             | แบบคู่สัญญา                                      |
+| Workflow Engine    | Approval Flow, Process Engine         | เครื่องมือจัดการลำดับงาน                         |
+| Document Numbering | Document ID, Auto Number              | ระบบจัดการเลขที่เอกสาร                           |
+| RBAC               | Permission System (generic)           | การควบคุมสิทธิ์ตามบทบาท                          |
+
+## Full Glossary
+
+`specs/00-overview/00-02-glossary.md`
+
+## Key Spec Files Priority
+
+Spec priority: **`06-Decision-Records`** > **`05-Engineering-Guidelines`** > others
+
+| Document                       | Path                                                                        | Use When                          |
+| ------------------------------ | --------------------------------------------------------------------------- | --------------------------------- |
+| **Glossary**                   | `specs/00-overview/00-02-glossary.md`                                       | Verify domain terminology         |
+| **Schema Tables**              | `specs/03-Data-and-Storage/lcbp3-v1.9.0-schema-02-tables.sql`               | Before writing any query          |
+| **Data Dictionary**            | `specs/03-Data-and-Storage/03-01-data-dictionary.md`                        | Field meanings + business rules   |
+| **Edge Cases**                 | `specs/01-Requirements/01-06-edge-cases-and-rules.md`                       | Prevent bugs in flows             |
+| **ADR-019 UUID**               | `specs/06-Decision-Records/ADR-019-hybrid-identifier-strategy.md`           | UUID-related work                 |
+| **ADR-023 AI**                 | `specs/06-Decision-Records/ADR-023-unified-ai-architecture.md`              | AI integration work               |
+| **ADR-023A AI Model**          | `specs/06-Decision-Records/ADR-023A-unified-ai-architecture.md`             | 2-model stack, BullMQ 2-queue     |
+| **ADR-024 Intent Class.**      | `specs/06-Decision-Records/ADR-024-intent-classification-strategy.md`       | Pattern→LLM Fallback; Redis cache |
+| **ADR-025 AI Tool Layer**      | `specs/06-Decision-Records/ADR-025-ai-tool-layer-architecture.md`           | Tool Registry; CASL-guarded       |
+| **ADR-026 Chat UI**            | `specs/06-Decision-Records/ADR-026-document-chat-ui-pattern.md`             | Side-panel; streaming SSE         |
+| **ADR-027 AI Admin Console**   | `specs/06-Decision-Records/ADR-027-ai-admin-console-and-dynamic-control.md` | Dynamic control; admin-only       |
+| **ADR-028 Migration Refactor** | `specs/06-Decision-Records/ADR-028-migration-architecture-refactor.md`      | Staging Queue; cleanup            |
+| **Backend Guidelines**         | `specs/05-Engineering-Guidelines/05-02-backend-guidelines.md`               | NestJS patterns                   |
+| **Frontend Guidelines**        | `specs/05-Engineering-Guidelines/05-03-frontend-guidelines.md`              | Next.js patterns                  |
+| **Testing Strategy**           | `specs/05-Engineering-Guidelines/05-04-testing-strategy.md`                 | Coverage goals                    |

.devin/rules/05-forbidden-actions.md

@@ -0,0 +1,43 @@
+---
+trigger: always_on
+---
+
+# Forbidden Actions
+
+## ❌ Never Do This
+
+| ❌ Forbidden                                    | ✅ Correct Approach                                     | ⚠️ Why                                               |
+| ----------------------------------------------- | ------------------------------------------------------- | ---------------------------------------------------- |
+| SQL Triggers for business logic                 | NestJS Service methods                                  | Untestable; bypasses audit log                       |
+| `.env` files in production                      | `docker-compose.yml` environment section                | Secrets exposed in version control                   |
+| TypeORM migration files                         | Edit schema SQL directly (ADR-009)                      | Migration drift risk; schema managed via SQL delta   |
+| Inventing table/column names                    | Verify against `schema-02-tables.sql`                   | Schema mismatch causes silent runtime errors         |
+| `any` TypeScript type                           | Proper types / generics                                 | Defeats strict mode; hides runtime type errors       |
+| `console.log` in committed code                 | NestJS Logger (backend) / remove (frontend)             | Log flooding in production; risk of data leakage     |
+| `req: any` in controllers                       | `RequestWithUser` typed interface                       | Type safety lost; auth context unreachable           |
+| `parseInt()` on UUID values                     | Use UUID string directly (ADR-019)                      | `"0195…"` parsed to integer `19` — silently wrong    |
+| Exposing INT PK in API responses                | UUIDv7 `publicId` (ADR-019)                             | Leaks row count; enables DB enumeration attacks      |
+| AI accessing DB/storage directly                | AI → DMS API → DB (ADR-023/023A)                        | Bypasses RBAC, audit trail, and validation layer     |
+| Direct file operations bypassing StorageService | `StorageService` for all file moves                     | Orphaned files; broken ClamAV scan; no audit trail   |
+| Inline email/notification sending               | BullMQ queue job (ADR-008)                              | Blocks request thread; no retry on transient failure |
+| Deploying without Release Gates                 | Complete `04-08-release-management-policy.md`           | Unverified deploy risks data loss in production      |
+| AI direct cloud API calls                       | On-premises Ollama only (ADR-023/023A)                  | Data privacy violation; no audit control             |
+| AI outputs without human validation             | Human-in-the-loop validation required (ADR-023/023A)    | Unvalidated AI metadata corrupts document records    |
+| n8n calling Ollama/Qdrant directly              | n8n → DMS API → BullMQ → Ollama (ADR-023A)              | Bypasses audit log, RBAC, and error handling layer   |
+| Qdrant query without `projectPublicId` filter   | `QdrantService.search(projectPublicId, ...)` (ADR-023A) | Cross-project data leak via vector search            |
+
+## Schema Changes (ADR-009)
+
+- **NO TypeORM migrations** — edit SQL schema directly
+- Always check `specs/03-Data-and-Storage/lcbp3-v1.9.0-schema-02-tables.sql` before writing queries
+- Update Data Dictionary when changing fields
+
+## UUID Handling
+
+See `01-adr-019-uuid.md` for complete UUID rules.
+
+Quick reminder:
+
+- ❌ `parseInt(uuid)` → NEVER
+- ❌ `Number(uuid)` → NEVER
+- ✅ Use UUID string directly

.devin/rules/08-development-flow.md

@@ -0,0 +1,214 @@
+---
+trigger: always_on
+---
+
+# Development Flow
+
+## 🔴 Critical Work — DB / API / Security / Workflow Engine
+
+**MUST complete all steps:**
+
+1. **Glossary check** — verify domain terms in `00-02-glossary.md`
+2. **Read the spec** — select from Key Spec Files table
+3. **Check schema** — verify table/column in `lcbp3-v1.9.0-schema-02-tables.sql`
+4. **Check data dictionary** — confirm field meanings + business rules
+5. **Scan edge cases** — `01-06-edge-cases-and-rules.md`
+6. **Check ADRs** — verify decisions align (ADR-009, ADR-019, ADR-023)
+7. **Write code** — TypeScript strict, no `any`, no `console.log`
+
+## 🟡 Normal Work — UI / Feature / Integration
+
+- Follow existing patterns in codebase.
+- Check spec for relevant module only.
+- **Hybrid Specs Organization:**
+  - Place new Infrastructure tasks in `specs/100-Infrastructures/`
+  - Place new Feature/Workflow tasks in `specs/200-fullstacks/`
+  - Place Documentation/Research in `specs/300-others/`
+- Ensure no forbidden patterns (`any`, `console.log`, UUID misuse) are introduced.
+
+## 🟢 Quick Fix — Bug Fix / Typo / Style
+
+- Fix directly
+- Add minimal test if logic changed
+- Check forbidden patterns before commit
+
+### 🟢 Specialized Work — ADR-021, AI Runtime Layer, Complex Logic
+
+**MUST complete:**
+
+1. **Domain Knowledge Check** - Read relevant ADRs (ADR-021, ADR-023/023A, ADR-024~028)
+2. **Pattern Verification** - Check existing implementations in codebase
+3. **Specialized Requirements** - Follow domain-specific patterns
+4. **Complex Logic Testing** - Multi-scenario test coverage
+5. **Performance Validation** - Load testing if applicable
+
+**For ADR-021 Integration:**
+
+- Read ADR-021 - Integrated workflow & step attachments
+- Check ADR-001 - Unified workflow engine patterns
+- Verify WorkflowEngineService - Polymorphic instance handling
+- Add workflow fields - Expose workflowInstanceId, workflowState, availableActions
+- Include IntegratedBanner - Frontend workflow lifecycle display
+- Test workflow transitions - State changes and action validation
+
+**For AI Infrastructure (ADR-023/023A):**
+
+- Verify AI boundary enforcement - No direct DB/storage access
+- Check BullMQ 2-queue setup - ai-realtime + ai-batch
+- Validate Qdrant multi-tenancy - projectPublicId filter required
+- Test human-in-the-loop validation workflows
+- Audit AI interaction logging to ai_audit_logs
+
+**For AI Runtime Layer (ADR-024/025/026/027):**
+
+- ADR-024: Pattern Layer first (ai_intent_patterns DB + Redis cache 5 min) → LLM Fallback (gemma4:e4b, semaphore max=3)
+- ADR-025: Tool Registry dispatch — AI Gateway → Tool → Business Service; ToolResult DTO must use publicId only
+- ADR-026: useAiChat() hook + side-panel UI; streaming response via SSE; TanStack Query cache
+- ADR-027: Admin Console — dynamic model/prompt/intent control; CASL-guarded admin-only endpoints
+
+**For Migration Pipeline (ADR-028):**
+
+- Use Staging Queue pattern — never write directly to production tables
+- Post-migration cleanup process required after each batch
+- Migration Validation Gates must pass before promoting to production
+
+**Expected output:**
+
+- Backend services expose specialized context fields
+- Frontend components use domain-specific patterns
+- Complex state management with proper validation
+- Performance metrics within acceptable thresholds
+- Comprehensive test coverage for edge cases
+
+---
+
+## Context-Aware Triggers
+
+| Request                     | Files to Check                                                                        | Expected Response                                                       |
+| --------------------------- | ------------------------------------------------------------------------------------- | ----------------------------------------------------------------------- |
+| "สร้าง API ใหม่"            | `05-02-backend-guidelines.md`, `lcbp3-v1.9.0-schema-02-tables.sql`                    | NestJS Controller + Service + DTO + CASL Guard                          |
+| "แก้ฟอร์ม frontend"         | `05-03-frontend-guidelines.md`, `01-06-edge-cases-and-rules.md`                       | RHF+Zod + TanStack Query + Thai comments                                |
+| "เพิ่ม field ใหม่"          | `ADR-009`, `03-01-data-dictionary.md`, `lcbp3-v1.9.0-schema-02-tables.sql`            | Edit SQL directly + update Data Dictionary + Entity                     |
+| "ตรวจสอบ UUID"              | `ADR-019`, `05-07-hybrid-uuid-implementation-plan.md`                                 | UUIDv7 MariaDB native UUID + TransformInterceptor                       |
+| "สร้าง migration"           | `ADR-009`, `03-06-migration-business-scope.md`                                        | Edit SQL schema directly + n8n workflow                                 |
+| "ตรวจสอบ permission"        | `lcbp3-v1.9.0-seed-permissions.sql`, `ADR-016`                                        | CASL 4-Level RBAC matrix                                                |
+| "deploy production"         | `04-08-release-management-policy.md`, `ADR-015`                                       | Release Gates + Blue-Green strategy                                     |
+| "เพิ่ม test"                | `05-04-testing-strategy.md`                                                           | Coverage goals + test patterns                                          |
+| "AI integration"            | `ADR-023`, `ADR-023A`, `ADR-024`, `ADR-025`                                           | AI boundary + 2-model stack + BullMQ queue policy + Intent/Tool Layer   |
+| "Error handling"            | `ADR-007`                                                                             | Layered error classification + recovery                                 |
+| "File upload"               | `ADR-016`, `05-02-backend-guidelines.md`, `03-Data-and-Storage/03-03-file-storage.md` | Two-phase upload → temp → commit; ClamAV + whitelist                    |
+| "Notifications / Queue"     | `ADR-008`, `05-02-backend-guidelines.md`                                              | BullMQ job — never inline; check retry + dead-letter                    |
+| "Add i18n / translate"      | `05-08-i18n-guidelines.md`                                                            | i18n keys only — no hardcoded text                                      |
+| "Workflow / DSL"            | `ADR-001`, `01-03-modules/01-03-06-unified-workflow.md`                               | DSL state machine + WorkflowEngineService                               |
+| "Document numbering"        | `ADR-002`, `01-02-business-rules/01-02-02-doc-numbering-rules.md`                     | Redis Redlock + DB optimistic lock (double-lock)                        |
+| "ตรวจสอบ Workflow"          | `01-06-edge-cases-and-rules.md`, `05-02-backend-guidelines.md`, `ADR-001`, `ADR-002`  | เช็คการเปลี่ยน State, คิว BullMQ และการล็อกเลขที่เอกสาร                 |
+| "Transmittal submit"        | `ADR-021`, `specs/200-fullstacks/201-transmittals-circulation/`                       | submit() with EC-RFA-004 validation                                     |
+| "Circulation reassign"      | `ADR-021`, `specs/200-fullstacks/201-transmittals-circulation/`                       | reassignRouting() with EC-CIRC-001                                      |
+| "สร้าง workflow ใหม่"       | `ADR-001`, `ADR-021`, `specs/200-fullstacks/203-unified-workflow-engine/`             | DSL workflow definition + WorkflowEngineService setup                   |
+| "ตรวจสอบ AI boundary"       | `ADR-023`, `ADR-023A`                                                                 | Verify Ollama isolation + BullMQ queues + Qdrant projectPublicId filter |
+| "Intent classification"     | `ADR-024`, `specs/200-fullstacks/224-intent-classification/`                          | Pattern Layer → LLM Fallback; ai_intent_patterns; Redis cache 5 min     |
+| "AI Tool Layer"             | `ADR-025`, `specs/200-fullstacks/225-ai-tool-layer-architecture/`                     | Tool Registry; CASL-guarded dispatch; ToolResult publicId only          |
+| "Document Chat UI"          | `ADR-026`, `specs/200-fullstacks/226-document-chat-ui-pattern/`                       | Side-panel; useAiChat() hook; streaming SSE; TanStack Query cache       |
+| "AI Admin Console"          | `ADR-027`, `specs/200-fullstacks/227-ai-admin-console/`                               | Dynamic model/prompt/intent control; admin-only CASL endpoints          |
+| "Migration refactor"        | `ADR-028`, `specs/200-fullstacks/228-migration-arch-refactor/`                        | Staging Queue; post-migration cleanup; validation gates                 |
+| "จัดการ document numbering" | `ADR-002`, `specs/03-Data-and-Storage/03-04-document-numbering.md`                    | Redis Redlock + template system + preview/override workflows            |
+| "Audit ความปลอดภัย"         | `ADR-016`, `ADR-019`, `ADR-023`, `ADR-023A`                                           | ตรวจสอบ UUID pattern, CASL Guard, AI Boundary และ Qdrant multi-tenancy  |
+| "แก้ bug / bugfix"          | `.agents/workflows/bugfix.md`, `error-catalog.md`                                     | ใช้ bugfix workflow สำหรับเคสที่สาเหตุชัดเจน                            |
+| "ตรวจแอปจริง"               | `.windsurf/workflows/check-real-app.md`                                               | ตรวจ endpoint/UI/console หลัง build pass — No Fake Evidence             |
+| "งานค้าง / resume"          | `.windsurf/workflows/resume-pending-work.md`                                          | อ่าน checkpoint เดิม → ตรวจ build → วางแผนต่อโดยไม่ทำงานซ้ำ             |
+
+---
+
+## 🔌 MCP MariaDB Tools
+
+MCP MariaDB server ให้เครื่องมือสำหรับตรวจสอบและจัดการ database โดยตรง ใช้สำหรับ:
+
+- ตรวจสอบ schema กับ spec file `specs/03-Data-and-Storage/lcbp3-v1.9.0-schema-02-tables.sql`
+- Debug ปัญหา database โดยไม่ต้องเข้า MySQL client
+- ตรวจสอบ data ใน production/staging
+- Validate การเปลี่ยนแปลง schema ก่อน deploy
+
+### Available Tools
+
+| Tool | หน้าที่ | ตัวอย่างการใช้งาน |
+|------|----------|------------------|
+| `mcp1_mysql_test_connection` | ทดสอบ connection กับ database | ตรวจสอบว่า MCP server เชื่อมต่อได้ |
+| `mcp1_mysql_show_databases` | แสดง databases ทั้งหมด | ดูว่ามี database อะไรบ้าง |
+| `mcp1_mysql_show_tables` | แสดง tables ทั้งหมดใน database | ดูรายชื่อ tables ใน `lcbp3` |
+| `mcp1_mysql_describe_table` | ดู structure/columns ของ table | ตรวจสอบ columns, types, keys ของ `correspondences` |
+| `mcp1_mysql_query` | รัน SELECT query | ดู data ใน table หรือ join query |
+| `mcp1_mysql_insert` | INSERT data | เพิ่ม seed data หรือ test data |
+| `mcp1_mysql_update` | UPDATE data | แก้ไข data ใน table |
+| `mcp1_mysql_delete` | DELETE data | ลบ data ใน table |
+
+### การใช้งานร่วมกับ Development Flow
+
+**เมื่อเขียน query ใหม่:**
+1. ใช้ `mcp1_mysql_describe_table` เพื่อตรวจสอบ columns และ types
+2. เปรียบเทียบกับ `specs/03-Data-and-Storage/lcbp3-v1.9.0-schema-02-tables.sql`
+3. ใช้ `mcp1_mysql_query` เพื่อทดสอบ query ก่อน implement
+
+**เมื่อเปลี่ยน schema (ADR-009):**
+1. ใช้ `mcp1_mysql_describe_table` เพื่อดู structure ปัจจุบัน
+2. สร้าง SQL delta ใน `specs/03-Data-and-Storage/deltas/`
+3. ใช้ `mcp1_mysql_query` เพื่อตรวจสอบผลลัพธ์หลัง apply delta
+
+**เมื่อ debug ปัญหา database:**
+1. ใช้ `mcp1_mysql_query` เพื่อดู data จริง
+2. เปรียบเทียบกับ spec และ data dictionary
+3. ตรวจสอบ foreign keys และ constraints
+
+### ข้อควรระวัง
+
+- **❌ ห้ามใช้ MCP MariaDB สำหรับ DDL operations** (CREATE/ALTER/DROP) โดยตรง — ต้องใช้ SQL delta ตาม ADR-009
+- **✅ ใช้สำหรับ DQL/DML operations** (SELECT/INSERT/UPDATE/DELETE) เพื่อ debug และ test เท่านั้น
+- **⚠️ ระวัง DELETE operations** — อาจทำให้เสีย data ใน production
+- **✅ ตรวจสอบ schema กับ spec file เสมอ** ก่อนเขียน query
+
+---
+
+## 🧠 MCP Memory Tools
+
+MCP Memory server ให้เครื่องมือสำหรับจัดการ Knowledge Graph และ Long-term Memory ใช้สำหรับ:
+
+- จัดเก็บความรู้และ context ของโปรเจกต์ในรูปแบบ Graph (Entities + Relations + Observations)
+- ค้นหาและดึงข้อมูล context จาก memory ที่บันทึกไว้ใน session ก่อนหน้า
+- สร้าง/แก้ไข/ลบ entities, relations, และ observations ใน knowledge graph
+
+### Available Tools
+
+| Tool | หน้าที่ | ตัวอย่างการใช้งาน |
+|------|----------|------------------|
+| `mcp3_create_entities` | สร้าง entities ใหม่หลายตัวพร้อม observations | สร้าง entity ใหม่เช่น Project, User, Task |
+| `mcp3_create_relations` | สร้าง relations ระหว่าง entities | สร้าง relation: Project → has → User |
+| `mcp3_add_observations` | เพิ่ม observations ให้ entity ที่มีอยู่แล้ว | เพิ่ม context เพิ่มเติมให้ entity |
+| `mcp3_delete_entities` | ลบ entities และ relations ที่เกี่ยวข้อง | ลบ entity ที่ไม่ใช้แล้ว |
+| `mcp3_delete_relations` | ลบ relations ระหว่าง entities | ลบ relation ที่ผิดหรือไม่ใช้แล้ว |
+| `mcp3_delete_observations` | ลบ observations จาก entity | ลบ context ที่ผิดหรือล้าสุด |
+| `mcp3_open_nodes` | ดึงข้อมูล entities ตามชื่อ | ดึง entity ที่ระบุชื่อ |
+| `mcp3_read_graph` | อ่าน knowledge graph ทั้งหมด | ดูทั้ง graph structure |
+| `mcp3_search_nodes` | ค้นหา entities ตาม query | ค้นหา entity จากชื่อ, type, หรือ observation |
+
+### การใช้งานร่วมกับ Development Flow
+
+**เมื่อบันทึก context ใหม่:**
+1. ใช้ `mcp3_create_entities` เพื่อสร้าง entities ใหม่ (ถ้ายังไม่มี)
+2. ใช้ `mcp3_create_relations` เพื่อเชื่อมโยง entities
+3. ใช้ `mcp3_add_observations` เพื่อเพิ่ม context/observations
+
+**เมื่อค้นหา context:**
+1. ใช้ `mcp3_search_nodes` เพื่อค้นหา entities ที่เกี่ยวข้อง
+2. ใช้ `mcp3_open_nodes` เพื่อดึงข้อมูล entities ที่ต้องการ
+3. ใช้ `mcp3_read_graph` เพื่อดู relations ระหว่าง entities
+
+**เมื่อแก้ไข context:**
+1. ใช้ `mcp3_add_observations` เพื่อเพิ่ม observations ใหม่
+2. ใช้ `mcp3_delete_observations` เพื่อลบ observations ที่ผิด
+3. ใช้ `mcp3_create_relations` หรือ `mcp3_delete_relations` เพื่อปรับ relations
+
+### ข้อควรระวัง
+
+- **✅ ใช้สำหรับบันทึก context ที่ต้องใช้ร่วมกันหลาย session** — เช่น การตัดสินใจสำคัญ, architecture decisions, rollout history
+- **⚠️ ระวังการลบ entities** — อาจทำให้เสีย context ที่ยังใช้งานอยู่
+- **✅ ตรวจสอบว่า entity มีอยู่แล้วก่อนสร้าง** — ใช้ `mcp3_search_nodes` หรือ `mcp3_open_nodes` ก่อน
+- **✅ ใช้ชื่อ entity ที่ชัดเจนและไม่ซ้ำกัน** — เพื่อป้องกันความสับสน

.devin/rules/11-ai-integration.md

@@ -7,7 +7,7 @@ trigger: always_on
 ## CRITICAL RULES
 
 - **ALWAYS** follow ADR-023 AI boundary policy (isolation on Admin Desktop)
-- **ALWAYS** use ADR-023A 2-model stack (gemma4:e4b Q8_0 + nomic-embed-text)
+- **ALWAYS** use ADR-023A 2-model stack (gemma4:e2b + nomic-embed-text)
 - **ALWAYS** use BullMQ 2-queue (ai-realtime + ai-batch) for GPU overload prevention
 - **NEVER** allow AI direct database/storage access
 - **ALWAYS** implement human-in-the-loop validation
@@ -30,7 +30,7 @@ n8n (Migration) → DMS API → BullMQ → Admin Desktop (Ollama) → Backend Va
 | ----------------- | ------------------------- | ------------------------------------------------------------------------ |
 | **AI Gateway**    | Backend (NestJS)          | API endpoints, validation, audit logging                                 |
 | **BullMQ Queues** | Backend (NestJS)          | ai-realtime (RAG/Suggest), ai-batch (OCR/Extract/Embed)                  |
-| **Ollama Engine** | Admin Desktop (Desk-5439) | gemma4:e4b Q8_0 (LLM) + nomic-embed-text (Embedding)                     |
+| **Ollama Engine** | Admin Desktop (Desk-5439) | gemma4:e2b (LLM) + nomic-embed-text (Embedding)                          |
 | **OCR Engine**    | Admin Desktop (Desk-5439) | PaddleOCR + PyThaiNLP (Thai/English text extraction)                     |
 | **Orchestrator**  | QNAP NAS (n8n)            | Migration Phase orchestrator only (calls DMS API, never Ollama directly) |
 
@@ -80,7 +80,7 @@ export class AiService {
   async extractMetadata(documentId: string): Promise<AIMetadata> {
     // 1. Validate permissions
     // 2. Queue job to BullMQ (ai-batch or ai-realtime)
-    // 3. Worker sends to Admin Desktop AI (gemma4:e4b Q8_0)
+    // 3. Worker sends to Admin Desktop AI (gemma4:e2b)
     // 4. Validate AI response
     // 5. Log audit trail to ai_audit_logs
     // 6. Return validated results
@@ -119,7 +119,7 @@ const DocumentReviewForm = ({ document, aiSuggestions }) => {
 
 ## ADR-023A Specific Rules
 
-- **2-Model Stack:** gemma4:e4b Q8_0 (~4.0GB) + nomic-embed-text (~0.3GB) = ~4.3GB VRAM peak
+- **2-Model Stack:** gemma4:e2b + nomic-embed-text
 - **PDF 3-Page Limit:** Classification/Tagging uses first 3 pages only (NOT RAG embedding)
 - **RAG Embedding:** Full document chunked at 512 tokens/64 tokens overlap
 - **OCR Auto-Detect:** PyMuPDF chars > 100 → Fast path, else PaddleOCR
@@ -133,7 +133,7 @@ const DocumentReviewForm = ({ document, aiSuggestions }) => {
 - [ ] BullMQ 2-queue setup (ai-realtime + ai-batch)
 - [ ] QdrantService with projectPublicId enforcement
 - [ ] DocumentReviewForm reusable component
-- [ ] Admin Desktop Ollama (gemma4:e4b Q8_0 + nomic-embed-text) + PaddleOCR setup
+- [ ] Admin Desktop Ollama (gemma4:e2b + nomic-embed-text) + PaddleOCR setup
 - [ ] n8n workflow orchestration (Migration Phase only)
 - [ ] AI audit logging and monitoring (ai_audit_logs)
 - [ ] Human-in-the-loop validation workflows
@@ -142,3 +142,8 @@ const DocumentReviewForm = ({ document, aiSuggestions }) => {
 
 - `specs/06-Decision-Records/ADR-023-unified-ai-architecture.md` (Base architecture)
 - `specs/06-Decision-Records/ADR-023A-unified-ai-architecture.md` (Model revision - current)
+- `specs/06-Decision-Records/ADR-024-intent-classification-strategy.md` (Pattern→LLM Fallback)
+- `specs/06-Decision-Records/ADR-025-ai-tool-layer-architecture.md` (Tool Registry)
+- `specs/06-Decision-Records/ADR-026-document-chat-ui-pattern.md` (Chat UI)
+- `specs/06-Decision-Records/ADR-027-ai-admin-console-and-dynamic-control.md` (Admin Console)
+- `specs/06-Decision-Records/ADR-028-migration-architecture-refactor.md` (Migration Pipeline)

.devin/rules/README.md

@@ -0,0 +1,141 @@
+---
+trigger: always_on
+---
+
+# LCBP3 Agent Rules
+
+Critical rules and guidelines for AI agents working on LCBP3-DMS.
+
+## Version
+
+- **Current:** v1.9.6
+- **Last Updated:** 2026-05-22
+- **Synced with:** `AGENTS.md` (v1.9.6)
+
+## Purpose
+
+This directory contains rule files that define:
+
+- Project context and role expectations
+- Critical Tier 1 rules (CI blockers)
+- Coding standards and patterns
+- Domain terminology and glossary
+- Development workflows
+- Security requirements
+- AI integration architecture (ADR-023/023A)
+
+## Rule Enforcement Tiers
+
+### 🔴 Tier 1 — CRITICAL (CI BLOCKER)
+
+Build fails immediately if violated:
+
+- Security (Auth, RBAC, Validation)
+- UUID Strategy (ADR-019) — no `parseInt` / `Number` / `+` on UUID
+- Database correctness — verify schema before writing queries
+- File upload security (ClamAV + whitelist)
+- AI validation boundary (ADR-023)
+- Error handling strategy (ADR-007)
+- Forbidden patterns: `any`, `console.log`, UUID misuse, `id ?? ''` fallback
+
+### 🟡 Tier 2 — IMPORTANT (CODE REVIEW)
+
+Must fix before merge:
+
+- Architecture patterns (thin controller, business logic in service)
+- Test coverage (80%+ business logic, 70%+ backend overall)
+- Cache invalidation
+- Naming conventions
+- TypeScript Standards: Missing JSDoc, explicit types, or file headers
+
+### 🟢 Tier 3 — SPECIALIZED WORK
+
+Requires domain-specific knowledge:
+
+- **ADR-021 Integration:** Workflow Engine & Context implementation
+- **AI Infrastructure:** ADR-023/023A boundary enforcement and pipeline usage
+- **AI Runtime Layer:** ADR-024 Intent Classification, ADR-025 Tool Layer, ADR-026 Chat UI, ADR-027 Admin Console
+- **Migration Pipeline:** ADR-028 Staging Queue & post-migration cleanup
+- **Complex Business Logic:** Multi-step workflows with state management
+- **Performance Optimization:** Database queries, caching strategies, bulk operations
+
+### 🔵 Tier 4 — GUIDELINES
+
+Best practice — follow when possible:
+
+- Code style / formatting (Prettier handles)
+- Comment completeness
+- Minor optimizations
+
+## Rule Files
+
+### Core Rules (Tier 1 - CRITICAL)
+
+| File                    | Purpose                                                                         |
+| ----------------------- | ------------------------------------------------------------------------------- |
+| `00-project-context.md` | Project context, role & persona, tier classification, specs folder organization |
+| `01-adr-019-uuid.md`    | UUID handling strategy — no parseInt, use publicId only                         |
+| `02-security.md`        | Security requirements, checklist, ADR-023/023A AI boundaries                    |
+
+### Coding Standards
+
+| File                      | Purpose                                                 |
+| ------------------------- | ------------------------------------------------------- |
+| `03-typescript.md`        | TypeScript rules, file headers, i18n guidelines         |
+| `06-backend-patterns.md`  | NestJS patterns, UUID resolution, API response patterns |
+| `07-frontend-patterns.md` | Next.js patterns, RHF+Zod+TanStack Query, UUID handling |
+
+### Domain & Workflow
+
+| File                       | Purpose                                                       |
+| -------------------------- | ------------------------------------------------------------- |
+| `04-domain-terminology.md` | DMS glossary, key spec files priority table                   |
+| `08-development-flow.md`   | Development workflow by work type (Critical/Normal/Quick Fix) |
+
+### Compliance & Architecture
+
+| File                      | Purpose                                                        |
+| ------------------------- | -------------------------------------------------------------- |
+| `05-forbidden-actions.md` | Actions that must never be done, schema changes, UUID handling |
+| `09-commit-checklist.md`  | Pre-commit verification, commit message format                 |
+| `10-error-handling.md`    | ADR-007 error handling strategy, layered classification        |
+| `11-ai-integration.md`    | ADR-023/023A AI architecture, 2-model stack, BullMQ 2-queue    |
+
+## Key Spec Files Priority
+
+Spec priority: **`06-Decision-Records`** > **`05-Engineering-Guidelines`** > others
+
+| Document                       | Path                                                                        | Use When                          |
+| ------------------------------ | --------------------------------------------------------------------------- | --------------------------------- |
+| **Glossary**                   | `specs/00-overview/00-02-glossary.md`                                       | Verify domain terminology         |
+| **Schema Tables**              | `specs/03-Data-and-Storage/lcbp3-v1.9.0-schema-02-tables.sql`               | Before writing any query          |
+| **Data Dictionary**            | `specs/03-Data-and-Storage/03-01-data-dictionary.md`                        | Field meanings + business rules   |
+| **Edge Cases**                 | `specs/01-Requirements/01-06-edge-cases-and-rules.md`                       | Prevent bugs in flows             |
+| **ADR-019 UUID**               | `specs/06-Decision-Records/ADR-019-hybrid-identifier-strategy.md`           | UUID-related work                 |
+| **ADR-023 AI**                 | `specs/06-Decision-Records/ADR-023-unified-ai-architecture.md`              | AI integration work               |
+| **ADR-023A AI Model**          | `specs/06-Decision-Records/ADR-023A-unified-ai-architecture.md`             | 2-model stack, BullMQ 2-queue     |
+| **ADR-024 Intent Class.**      | `specs/06-Decision-Records/ADR-024-intent-classification-strategy.md`       | Pattern→LLM Fallback; Redis cache |
+| **ADR-025 AI Tool Layer**      | `specs/06-Decision-Records/ADR-025-ai-tool-layer-architecture.md`           | Tool Registry; CASL-guarded       |
+| **ADR-026 Chat UI**            | `specs/06-Decision-Records/ADR-026-document-chat-ui-pattern.md`             | Side-panel; streaming SSE         |
+| **ADR-027 AI Admin Console**   | `specs/06-Decision-Records/ADR-027-ai-admin-console-and-dynamic-control.md` | Dynamic control; admin-only       |
+| **ADR-028 Migration Refactor** | `specs/06-Decision-Records/ADR-028-migration-architecture-refactor.md`      | Staging Queue; cleanup            |
+| **Backend Guidelines**         | `specs/05-Engineering-Guidelines/05-02-backend-guidelines.md`               | NestJS patterns                   |
+| **Frontend Guidelines**        | `specs/05-Engineering-Guidelines/05-03-frontend-guidelines.md`              | Next.js patterns                  |
+| **Testing Strategy**           | `specs/05-Engineering-Guidelines/05-04-testing-strategy.md`                 | Coverage goals                    |
+
+## Maintenance
+
+When updating rules:
+
+1. **Check AGENTS.md version** — Ensure rule files are synced
+2. **Update version numbers** — Bump version in `00-project-context.md` only (03-typescript.md no longer has version)
+3. **Review ADR references** — Ensure all ADR references are current (ADR-023, ADR-023A, ADR-024~028)
+4. **Add new forbidden actions** — When new patterns are identified as violations
+5. **Update key spec files table** — When new ADRs or guidelines are added
+6. **Update Tier 3 SPECIALIZED WORK** — When new domain-specific workflows are added
+
+## Related Documents
+
+- `AGENTS.md` — Master agent configuration and context
+- `specs/06-Decision-Records/` — All Architecture Decision Records
+- `specs/05-Engineering-Guidelines/` — Backend, frontend, and testing guidelines

.devin/rules/specify-rules.md

@@ -0,0 +1,30 @@
+# lcbp3 Development Guidelines
+
+Auto-generated from all feature plans. Last updated: 2026-05-30
+
+## Active Technologies
+
+- TypeScript 5.x (NestJS 11 backend, Next.js 16 frontend), Python 3.11 (OCR sidecar)
 + Ollama (AI runtime), BullMQ (job queues), TypeORM (ORM), Redis (caching/locks), MariaDB 11.8 (database)
 (232-typhoon-ocr-integration)
+
+## Project Structure
+
+```text
+backend/
+frontend/
+tests/
+```
+
+## Commands
+
+cd src [ONLY COMMANDS FOR ACTIVE TECHNOLOGIES][ONLY COMMANDS FOR ACTIVE TECHNOLOGIES] pytest [ONLY COMMANDS FOR ACTIVE TECHNOLOGIES][ONLY COMMANDS FOR ACTIVE TECHNOLOGIES] ruff check .
+
+## Code Style
+
+TypeScript 5.x (NestJS 11 backend, Next.js 16 frontend), Python 3.11 (OCR sidecar)
: Follow standard conventions
+
+## Recent Changes
+
+- 232-typhoon-ocr-integration: Added TypeScript 5.x (NestJS 11 backend, Next.js 16 frontend), Python 3.11 (OCR sidecar)
 + Ollama (AI runtime), BullMQ (job queues), TypeORM (ORM), Redis (caching/locks), MariaDB 11.8 (database)
+
+<!-- MANUAL ADDITIONS START -->
+<!-- MANUAL ADDITIONS END -->

.devin/skills/README.md

@@ -1,8 +1,8 @@
 # `.agents/skills/` — LCBP3 Agent Skill Pack
 
-**Version:** 1.9.0 | **Last Updated:** 2026-05-17 | **Total Skills:** 23
+**Version:** 1.9.0 | **Last Updated:** 2026-06-07 | **Total Skills:** 24
 
-Agent skills for AI-assisted development in **Windsurf IDE** (and compatible agents: Codex CLI, opencode, Amp, Antigravity, AGENTS.md-aware tools).
+Agent skills for AI-assisted development in **Devin IDE** (and compatible agents: Codex CLI, opencode, Amp, Antigravity, AGENTS.md-aware tools).
 
 ---
 
@@ -14,6 +14,7 @@ Agent skills for AI-assisted development in **Windsurf IDE** (and compatible age
 ├── skills.md                    # Overview + dependency matrix + health monitoring
 ├── _LCBP3-CONTEXT.md            # Shared LCBP3 context injected into every speckit-* skill
 ├── README.md                    # (this file)
+├── save-memory/                 # Session log & project memory update
 ├── nestjs-best-practices/       # Backend rules (40 rules across 10 categories)
 ├── next-best-practices/         # Frontend rules (Next.js 15+)
 ├── e2e-testing/                 # Playwright E2E testing patterns (POM, flaky tests, CI/CD)
@@ -30,12 +31,10 @@ Each skill directory contains:
 
 ---
 
-## 🚀 How Windsurf Invokes These Skills
+## 🚀 How Devin Invokes These Skills
 
-Windsurf exposes two entry points:
+1. **Skill tool** — Devin discovers skills by scanning `.agents/skills/*/SKILL.md` frontmatter. Skills marked `user-invocable: false` are used silently by Cascade.
-
+2. **Slash commands** — `.devin/workflows/*.md` wraps each skill as a slash command (e.g. `/04-speckit.plan`). The workflow file is short; the heavy lifting is delegated to the skill via `skill` tool.
-1. **Skill tool** — Windsurf discovers skills by scanning `.agents/skills/*/SKILL.md` frontmatter. Skills marked `user-invocable: false` are used silently by Cascade.
-2. **Slash commands** — `.windsurf/workflows/*.md` wraps each skill as a slash command (e.g. `/04-speckit.plan`). The workflow file is short; the heavy lifting is delegated to the skill via `skill` tool.
 
 Both paths end up executing the same `SKILL.md` instructions.
 
@@ -66,13 +65,13 @@ Use `/00-speckit.all` to run specify → clarify → plan → tasks → analyze
 From repo root:
 
 | Script                                                 | Purpose                                                    |
-| --------------------------------------------------------- | ----------------------------------------------------------- |
+| ------------------------------------------------------ | ---------------------------------------------------------- |
 | `./.agents/scripts/bash/check-prerequisites.sh --json` | Emit `FEATURE_DIR` + `AVAILABLE_DOCS` for a feature branch |
 | `./.agents/scripts/bash/setup-plan.sh --json`          | Emit `FEATURE_SPEC`, `IMPL_PLAN`, `SPECS_DIR`, `BRANCH`    |
-| `./.agents/scripts/bash/update-agent-context.sh windsurf` | Append tech entries to `AGENTS.md`                          |
+| `./.agents/scripts/bash/update-agent-context.sh devin` | Append tech entries to `AGENTS.md`                         |
 | `./.agents/scripts/bash/audit-skills.sh`               | Validate all `SKILL.md` frontmatter + presence             |
 | `./.agents/scripts/bash/validate-versions.sh`          | Version consistency check                                  |
-| `./.agents/scripts/bash/sync-workflows.sh`                | Verify every skill has a `.windsurf/workflows/*.md` wrapper |
+| `./.agents/scripts/bash/sync-workflows.sh`             | Verify every skill has a `.devin/workflows/*.md` wrapper   |
 
 All scripts mirror to `.agents/scripts/powershell/*.ps1` for Windows.
 
@@ -97,7 +96,7 @@ To add a new skill:
 
 1. Create `NAME/SKILL.md` with frontmatter: `name`, `description`, `version: 1.9.0`, `scope`, `depends-on`.
 2. Append an LCBP3 context reference pointing to `_LCBP3-CONTEXT.md`.
-3. Wrap with `.windsurf/workflows/NAME.md` so it becomes a slash command.
+3. Wrap with `.devin/workflows/NAME.md` so it becomes a slash command.
 4. Update [`skills.md`](./skills.md) dependency matrix.
 5. Run `./.agents/scripts/bash/audit-skills.sh` → must pass.
 

.devin/skills/_LCBP3-CONTEXT.md

@@ -5,14 +5,14 @@
 
 **Project:** NAP-DMS (LCBP3) — Laem Chabang Port Phase 3 Document Management System
 **Stack:** NestJS 11 + Next.js 16 + TypeScript + MariaDB 11.8 + Redis + BullMQ + Elasticsearch + Ollama (on-prem AI)
-**Version:** 1.8.9 (2026-04-18)
+**Version:** 1.9.7 (2026-05-25)
 
 ---
 
 ## 📌 Canonical Rule Sources (read in this order)
 
 1. **`AGENTS.md`** (repo root) — primary rule file for AI agents; supersedes legacy `GEMINI.md`.
-2. **`specs/06-Decision-Records/`** — architectural decisions (22 ADRs); ADR priority > Engineering Guidelines.
+2. **`specs/06-Decision-Records/`** — architectural decisions (29 ADRs); ADR priority > Engineering Guidelines.
 3. **`specs/05-Engineering-Guidelines/`** — backend/frontend/testing/i18n/git patterns.
 4. **`specs/00-Overview/00-02-glossary.md`** — domain terminology (Correspondence / RFA / Transmittal / Circulation).
 5. **`specs/00-Overview/00-03-product-vision.md`** — project constitution (Vision, Strategic Pillars, Guardrails).
@@ -29,6 +29,7 @@
 - **ADR-002 Document Numbering:** Redis Redlock + TypeORM `@VersionColumn` (double-lock). Never use application-side counter alone.
 - **ADR-008 Notifications:** BullMQ queue — never inline email/notification in a request thread.
 - **ADR-023/023A AI Boundary:** Ollama on Admin Desktop only; AI → DMS API → DB (never direct DB/storage). 2-model stack: `gemma4:e4b Q8_0` + `nomic-embed-text`. BullMQ `ai-realtime` / `ai-batch` queues. Human-in-the-loop validation required. (ADR-018 superseded by ADR-023)
+- **ADR-029 Dynamic Prompt Management:** Prompt templates in DB (`ai_prompts`), never hardcoded in processor; Redis cache `ai:prompt:active:{type}` TTL 60s; `activate()` runs in DB transaction + Redis DEL after commit; `system.manage_all` guard on all mutations.
 - **ADR-007 Error Handling:** Layered (Validation / Business / System); `BusinessException` hierarchy; user-friendly `userMessage` + `recoveryAction`; technical stack only in logs.
 - **TypeScript Strict:** Zero `any`, zero `console.log` (use NestJS `Logger`).
 - **i18n:** No hardcoded Thai/English strings in components — use i18n keys (see `05-08-i18n-guidelines.md`).
@@ -54,7 +55,7 @@
 ## 📁 Key Files for Generating / Validating Artifacts
 
 | When you need...           | Read                                                                                                                |
-| ------------------------- | ------------------------------------------------------------------------------------------------------------------- |
+| -------------------------- | ------------------------------------------------------------------------------------------------------------------- |
 | A new feature spec         | `.agents/skills/speckit-specify/templates/spec-template.md` + `specs/01-Requirements/01-06-edge-cases-and-rules.md` |
 | A plan                     | `.agents/skills/speckit-plan/templates/plan-template.md` + relevant ADRs                                            |
 | Task breakdown             | `.agents/skills/speckit-tasks/templates/tasks-template.md` + existing patterns in `specs/08-Tasks/`                 |
@@ -62,6 +63,12 @@
 | Schema / table definition  | `specs/03-Data-and-Storage/lcbp3-v1.9.0-schema-02-tables.sql` + `03-01-data-dictionary.md`                          |
 | RBAC / permissions         | `specs/03-Data-and-Storage/lcbp3-v1.8.0-seed-permissions.sql` + `01-02-01-rbac-matrix.md`                           |
 | Release / hotfix           | `specs/04-Infrastructure-OPS/04-08-release-management-policy.md`                                                    |
+| ADR-024 Intent Class.      | `specs/06-Decision-Records/ADR-024-intent-classification-strategy.md`                                               |
+| ADR-025 AI Tool Layer      | `specs/06-Decision-Records/ADR-025-ai-tool-layer-architecture.md`                                                   |
+| ADR-026 Chat UI            | `specs/06-Decision-Records/ADR-026-document-chat-ui-pattern.md`                                                     |
+| ADR-027 AI Admin Console   | `specs/06-Decision-Records/ADR-027-ai-admin-console-and-dynamic-control.md`                                         |
+| ADR-028 Migration Refactor | `specs/06-Decision-Records/ADR-028-migration-architecture-refactor.md`                                              |
+| ADR-029 Dynamic Prompts    | `specs/06-Decision-Records/ADR-029-dynamic-prompt-management.md`                                                    |
 
 ---
 
@@ -83,7 +90,7 @@
 - [ ] Business comments in Thai, code identifiers in English
 - [ ] Schema changes via SQL directly (not migration)
 - [ ] Test coverage meets targets (Backend 70%+, Business Logic 80%+)
-- [ ] Relevant ADRs referenced (007/008/009/016/019/021/023/023A for AI work)
+- [ ] Relevant ADRs referenced (007/008/009/016/019/021/023/023A/024-029 for AI work)
 - [ ] Domain glossary terms used correctly
 - [ ] Error handling: `Logger` + `HttpException` / `BusinessException`
 - [ ] i18n keys used (no hardcode text)

.devin/skills/nestjs-best-practices/AGENTS.md

@@ -6454,7 +6454,7 @@ CREATE TABLE ai_audit_log (
   user_id INT NOT NULL,
   action VARCHAR(64) NOT NULL,       -- 'ai.extract_metadata', 'ai.classify', etc.
   file_id INT,
-  model VARCHAR(64),                  -- 'gemma-4:7b', 'paddleocr-v3'
+  model VARCHAR(64),                  -- 'gemma-4:7b', 'typhoon-np-dms-ocr', 'tesseract-ocr'
   confidence DECIMAL(4,3),
   input_hash CHAR(64),                -- SHA-256 of input for replay detection
   output_summary JSON,

.devin/skills/nestjs-best-practices/rules/lcbp3-ai-boundary.md

@@ -137,7 +137,7 @@ CREATE TABLE ai_audit_log (
   user_id INT NOT NULL,
   action VARCHAR(64) NOT NULL,       -- 'ai.extract_metadata', 'ai.classify', etc.
   file_id INT,
-  model VARCHAR(64),                  -- 'gemma-4:7b', 'paddleocr-v3'
+  model VARCHAR(64),                  -- 'gemma-4:7b', 'typhoon-np-dms-ocr', 'tesseract-ocr'
   confidence DECIMAL(4,3),
   input_hash CHAR(64),                -- SHA-256 of input for replay detection
   output_summary JSON,