lab_code: demo_411
title_en: Build an Agentic Branching Workflow
title_zh: 构建 Agentic 分支工作流
title_ja: Agentic Branch ワークフローの構築
title_ko: Agentic 분기 워크플로 구축
description_en: "In this lab, you will learn: Understand why product-grade agent systems need database branching, using Manus as a reference; Explain TiDB Cloud Branching and create a branch manually; Use Amazon Strands Agents in a Jupyter Notebook to explore three competing Todo app schemas on three branches, then branch again from the chosen baseline to add a users table and relationships; Optionally, use TiDB Cloud Branching for GitHub pull request testing and manage branches via the Branch API; Review a research paper on why branching matters in agentic systems."
description_zh: 在本次实验中，你将学习如何：以 Manus 为参考理解为什么真实的 Agent 系统需要数据库分支；理解 TiDB Cloud Branching 并在控制台手动创建分支；在 Jupyter Notebook 中使用 Amazon Strands Agents，在 3 个分支上并行探索 Todo 应用的 3 种 Schema 方案，并从选定基线继续分叉以增加 users 表与关联；可选地，将分支用于 GitHub Pull Request 的真实分支测试，并通过 Branch API 管理分支；阅读一篇论文理解分支机制对 Agentic 系统的重要性。
description_ja: この演習では、次のことを学びます：Manus を参考に、実運用の Agent システムでデータベース分岐が必要となる理由を理解します；TiDB Cloud Branching の概要を把握し、コンソールでブランチを手動作成します；Jupyter Notebook で Amazon Strands Agents を用い、3 つのブランチ上で Todo アプリの 3 つの Schema 案を並行に検討し、選択した基準線から再分岐して users テーブルとリレーションを追加します；任意で、GitHub Pull Request の検証にブランチを活用し、Branch API でブランチを管理します；分岐が Agentic システムに重要な理由を論文で確認します。
description_ko: "이 실습에서는 다음을 배우게 됩니다: Manus를 참고해 실제 Agent 시스템에서 데이터베이스 분기가 왜 필요한지 이해합니다; TiDB Cloud Branching을 이해하고 콘솔에서 수동으로 브랜치를 생성합니다; Jupyter Notebook에서 Amazon Strands Agents를 사용해 3개의 브랜치에서 Todo 앱의 3가지 스키마 접근을 병렬로 탐색하고, 선택한 기준 브랜치에서 다시 분기하여 users 테이블과 관계를 추가합니다; 선택 사항으로 GitHub Pull Request 테스트에 브랜치를 활용하고 Branch API로 브랜치를 관리합니다; 에이전틱 시스템에서 분기가 중요한 이유를 논문으로 확인합니다."
splash_image_en: https://lab-static.pingcap.com/images/2025/12/18/6b62baa64f40214a35226bdd964e11ed90e5d6cb.jpg
splash_image_zh: https://lab-static.pingcap.com/images/2025/12/18/6b62baa64f40214a35226bdd964e11ed90e5d6cb.jpg
splash_image_ja: https://lab-static.pingcap.com/images/2025/12/18/6b62baa64f40214a35226bdd964e11ed90e5d6cb.jpg
splash_image_ko: https://lab-static.pingcap.com/images/2025/12/18/6b62baa64f40214a35226bdd964e11ed90e5d6cb.jpg
external_materials_en:
  - name: "Branching for Agentic AI Deck Template"
    url: https://docs.google.com/presentation/d/1rR5mM1fivaxLBxF-yDXM-GhivTLNMAqV/edit?slide=id.p1#slide=id.p1
    employee_only: true
external_materials_ja:
  - name: "Agentic AI ブランチワークフローデッキテンプレート"
    url: https://docs.google.com/presentation/d/1rR5mM1fivaxLBxF-yDXM-GhivTLNMAqV/edit?slide=id.p1#slide=id.p1
    employee_only: true
external_materials_zh:
  - name: "Agentic AI 分支工作流演示文稿模板"
    url: https://docs.google.com/presentation/d/1rR5mM1fivaxLBxF-yDXM-GhivTLNMAqV/edit?slide=id.p1#slide=id.p1
    employee_only: true
external_materials_ko:
  - name: "Agentic AI 분기 워크플로 템플릿"
    url: https://docs.google.com/presentation/d/1rR5mM1fivaxLBxF-yDXM-GhivTLNMAqV/edit?slide=id.p1#slide=id.p1
    employee_only: true
lab_level: Advanced
category_en:
  - TiDB Cloud
  - Generative AI
category_zh:
  - TiDB Cloud
  - 生成式 AI
category_ja:
  - TiDB Cloud
  - 生成 AI
category_ko:
  - TiDB Cloud
  - 생성형 AI
authors:
  - Cheese Wong (CEG)
delivery_type: Self-Paced
tidb_environments:
  - TiDB Cloud Starter
tidb_versions: []
target_audience:
  - GenAI Developer
regions:
  aws:
    - oregon
    - singapore
    - frankfurt
credits: 1
credit_redeemable: true
employee_only: false
estimated_duration_minutes: 120

requirements:
  - type: "CloudStack"
  - type: "TiDBServerlessCluster"

topology:
  - role: monitor
    node_count: 1
    ssh_user: ubuntu
    web_ssh_name: "Client Machine"
    web_ssh: true
    https_reverse_proxy: true

bootstrap:
  - name: "prepare monitor"
    type: "ShellScript"
    target:
      - monitor1
    script: ./scripts/prepare-monitor.sh

features:
  enable_content_explain: false
  disable_serverless_sandbox: true
  enable_environment_variables_fallback: true