07 · Real-World Use Cases

How the components from earlier sections combine to address real engineering workflow automation.

Use Case Overview

Use Case Trigger Output Human Gate
JIRA → Pull Request New ticket assigned to AI agent GitHub PR with code + tests PR review before merge
Playwright RCA & Fix CI test failure Fix commit or RCA document on PR Developer reviews fix
Spring Boot Code Generation Architect provides OpenAPI spec Scaffold controller + service + tests PR review

Architecture Overview — Both Cases

graph TD
    subgraph Ingestion
        A[JIRA Webhook] --> Orch
        B[CI Failure Webhook] --> Orch
    end
    subgraph Orch[Orchestration Layer · LangGraph]
        C[State Machine]
        D[Agent Nodes]
        E[Interrupt Gates]
    end
    subgraph Tools[Tool Layer]
        F[GitHub MCP]
        G[JIRA MCP]
        H[Playwright MCP]
        I[Vector Code Index]
    end
    subgraph Output
        J[GitHub PR]
        K[PR Comment · RCA]
        L[JIRA Comment]
    end
    Orch --> Tools
    Tools --> Orch
    Orch --> Output

Key Design Decisions

Decision Recommendation Reason
LLM provider Claude 3.5 Sonnet or GPT-4o Best instruction following for code tasks
Orchestration LangGraph Stateful, interrupt support, parallel branches
Code index pgvector + nomic-embed-code Simple, stays in existing Postgres
Ticket integration JIRA MCP server Portable, reusable across projects
Human gate LangGraph interrupt before PR create Non-negotiable safety constraint
Storage PostgreSQL checkpointer Resumable across server restarts

→ Deep Dive: JIRA to Pull Request (Case 1)
→ Deep Dive: Playwright RCA & Auto-Fix (Case 2)
→ Deep Dive: Spring Boot Code Generation

What Else Can AI Automate?

Beyond the two primary use cases, the same stack supports:

Capability Trigger Agent Output
Code review assistance PR opened Inline review comments on security, style, coverage
Documentation generation Code merged Updated Confluence / README from code changes
Dependency upgrade Security advisory PR bumping dependency version with test validation
On-call runbook execution PagerDuty alert Diagnose, apply known fix, post incident summary
API contract testing Spec changed Generate and run contract tests between services
Test gap detection Weekly scheduled Identify code paths without test coverage, generate tests
Sprint retrospective summary End of sprint Read completed tickets, generate engineer-readable summary