Luồng Dữ Liệu — DocKit Master

Luồng Dữ Liệu

Tham Khảo Nhanh

• Pattern: Sequential Pipeline (6 bước)
• Input: Câu trả lời cấu hình + source code
• Output: Astro Starlight site hoặc Markdown files
• Serialization: Markdown + YAML frontmatter

Pipeline Tổng Quan

DocKit Master xử lý dữ liệu theo pipeline 6 bước tuần tự. Mỗi bước nhận output của bước trước làm input.

graph TB
    style S1 fill:#2d333b,stroke:#6d5dfc,color:#e6edf3
    style S2 fill:#2d333b,stroke:#6d5dfc,color:#e6edf3
    style S3 fill:#2d333b,stroke:#8b949e,color:#e6edf3
    style S4 fill:#2d333b,stroke:#6d5dfc,color:#e6edf3
    style S5 fill:#2d333b,stroke:#3fb950,color:#e6edf3
    style S6 fill:#2d333b,stroke:#3fb950,color:#e6edf3

    S1["📥 Step 1<br/>Gather Input"] --> S2["🔍 Step 2<br/>Analyze Codebase"]
    S2 --> S3["📝 Step 3<br/>Apply Guidelines"]
    S3 --> S4["⚙️ Step 4<br/>Generate Documents"]
    S4 --> S5["📤 Step 5<br/>Export + Sitemap"]
    S5 --> S6["✅ Step 6<br/>Summary"]

Step 1: Gather Input

Orchestrator (SKILL.md:41-118) hiển thị form 10 câu hỏi cho người dùng. Output là execution config.

sequenceDiagram
    participant U as 👤 User
    participant O as 📋 SKILL.md

    O->>U: Hiển thị form cấu hình
    Note over U: Trả lời: "all, astro, full, yes, yes"
    U->>O: Cấu hình

    Note over O: Auto-detect language<br/>từ ngôn ngữ chat
    Note over O: Generate execution plan<br/>DOC_TYPE=all, FORMAT=astro...

    O->>U: Hiển thị kế hoạch triển khai
    Note over O: Proceed to Step 2

Input: Câu trả lời người dùng (text) Output: Execution config object (10 fields)

Step 2: Analyze Codebase

Analyzer (skills/analyze-codebase.md) quét toàn bộ codebase và tạo metadata.

graph LR
    style SCAN fill:#2d333b,stroke:#6d5dfc,color:#e6edf3
    style DETECT fill:#2d333b,stroke:#6d5dfc,color:#e6edf3
    style MAP fill:#2d333b,stroke:#6d5dfc,color:#e6edf3
    style OUTPUT fill:#2d333b,stroke:#3fb950,color:#e6edf3

    SCAN["📂 Scan Files<br/>root, dirs, extensions"] --> DETECT["🔎 Detect Stack<br/>frameworks, languages"]
    DETECT --> MAP["🗺️ Map Architecture<br/>layers, entry points"]
    MAP --> OUTPUT["📄 analysis.md<br/>structured metadata"]

Quy trình quét:

1. Đọc README.md, package.json, config files
2. Đếm files theo extension → detect languages
3. Detect frameworks từ signal files
4. Map 6 architecture layers (presentation, business, data, infra, integrations, testing)
5. Extract routes, database schema, dependencies

Input: Source code files Output: docs/analysis.md với Quick Reference, Architecture diagram, tables

Step 3: Apply Content Guidelines

Content engine áp dụng 3 bộ quy tắc song song:

Bộ quy tắc	File	Mục đích
Content Guidelines	skills/content-guidelines.md	UX laws, Markdown rules, frontmatter schema
SEO Writing	skills/content-writing.md	Keyword placement, inverted pyramid, active voice
LLM Optimization	skills/llm-optimization.md	AI-readable heading, self-contained sections

Các quy tắc này không tạo output file — chúng là constraints cho Step 4.

Step 4: Generate Documents

3 generators chạy tuần tự, mỗi generator đọc analysis.md + source code:

graph TB
    style ANALYSIS fill:#2d333b,stroke:#8b949e,color:#e6edf3
    style TECH fill:#2d333b,stroke:#6d5dfc,color:#e6edf3
    style SOP fill:#2d333b,stroke:#6d5dfc,color:#e6edf3
    style API fill:#2d333b,stroke:#6d5dfc,color:#e6edf3
    style OUT1 fill:#2d333b,stroke:#3fb950,color:#e6edf3
    style OUT2 fill:#2d333b,stroke:#3fb950,color:#e6edf3
    style OUT3 fill:#2d333b,stroke:#3fb950,color:#e6edf3

    ANALYSIS["📄 analysis.md"] --> TECH["📐 tech-docs.md"]
    ANALYSIS --> SOP["📋 sop-guide.md"]
    ANALYSIS --> API["🔌 api-reference.md"]

    TECH --> OUT1["architecture.md<br/>database.md<br/>deployment.md<br/>data-flow.md"]
    SOP --> OUT2["sop/index.md<br/>sop/[feature].md"]
    API --> OUT3["api/index.md<br/>api/[resource].md"]

Tech Docs Generator

Sinh 4 file kỹ thuật với ≥2 Mermaid diagrams mỗi file:

• architecture.md — Sơ đồ kiến trúc, ADR
• database.md — Schema, ER diagram, indexes
• deployment.md — Cài đặt, cấu hình, CI/CD
• data-flow.md — Pipeline, sequence diagrams

SOP Generator

Quét UI routes/components → nhóm theo module → sinh 1 file SOP per feature:

• Quick Reference card
• Step-by-step numbered guide
• Form field tables
• Troubleshooting (progressive disclosure)
• FAQ (schema-ready)

API Generator

Quét route files → nhóm theo resource → sinh 1 file per resource:

• Endpoints overview table
• Parameters + response schemas
• Multi-language examples (cURL, Python, JS, Go)
• Error codes + rate limiting

Step 5: Export

Tuỳ theo FORMAT, exporter scaffold output:

Astro Path

graph LR
    style DOCS fill:#2d333b,stroke:#6d5dfc,color:#e6edf3
    style SCAFFOLD fill:#2d333b,stroke:#6d5dfc,color:#e6edf3
    style TEMPLATE fill:#2d333b,stroke:#8b949e,color:#e6edf3
    style BUILD fill:#2d333b,stroke:#3fb950,color:#e6edf3
    style SEO fill:#2d333b,stroke:#d29922,color:#e6edf3

    DOCS["📄 docs/*.md"] --> SCAFFOLD["🚀 npm create astro"]
    TEMPLATE["🎯 templates/<br/>config + CSS"] --> SCAFFOLD
    SCAFFOLD --> BUILD["📦 npm run build<br/>→ dist/"]
    BUILD --> SEO["🔍 SEO Audit<br/>+ Sitemap"]

1. npm create astro@latest — scaffold Starlight project
2. Copy premium template (config + CSS)
3. Copy docs into src/content/docs/
4. npm install && npm run build
5. Generate sitemap-urls.txt + robots.txt
6. Run SEO checklist trên mọi trang

Markdown Path

Copy docs trực tiếp, tạo docs/README.md làm index.

Step 6: Summary

Orchestrator hiển thị cho người dùng:

• Danh sách file đã tạo + kích thước
• Hướng dẫn xem/serve docs
• Next steps (customize, deploy)

Tích Hợp Bên Ngoài

Dịch vụ	Protocol	Hướng	Dữ liệu
Google Antigravity	Prompt/Response	Bidirectional	Markdown text
Clipboard (macOS)	pbcopy	Output only	Prompt text
npm Registry	HTTP	Download	Astro packages
NotebookLM	URL paste	Output only	sitemap-urls.txt

---

Xem thêm: Kiến trúc hệ thống · Phân tích mã nguồn