Skip to content

ipinto-ai-tools/multi-agents-ocp-builds

Repository files navigation

FlowPilot: Feature SDLC Automation for Shipwright / OpenShift Builds

SDLC orchestration layer that drives a feature from Jira ticket to production-ready Go code, tests, and documentation — using Claude as the execution engine for every stage.


What It Does

You provide a Jira ticket (or a plain feature description). FlowPilot orchestrates five SDLC stages — each powered by Claude — and delivers a complete set of artifacts: architecture design documents, production Go code, code review findings with an auto-fix loop, Ginkgo v2 test files, and PR-ready documentation. Each stage validates its outputs before passing state to the next, so failures surface immediately rather than propagating silently through the pipeline.


The SDLC Flow

┌─────────────────────────────────────────────────────────────────────┐
│                    Feature SDLC Pipeline                            │
├──────────┬──────────┬─────────────┬──────────┬──────────────────────┤
│  Phase 1 │  Phase 2 │   Phase 2.5 │  Phase 3 │       Phase 4        │
│  DESIGN  │  DEVELOP │ CODE REVIEW │  TESTING │   DOCUMENTATION      │
├──────────┼──────────┼─────────────┼──────────┼──────────────────────┤
│ Arch     │ Go code  │ Blocking /  │ Unit     │ PR Summary           │
│ analysis │ PR desc  │ Warning     │ Integr.  │ Release Notes        │
│ Risks    │ API types│ findings    │ E2E      │ SHIP / JTBD docs     │
│ Impl plan│ Tests    │ Auto-fix ↺  │ Ginkgo v2│                      │
└──────────┴──────────┴─────────────┴──────────┴──────────────────────┘
         Input: Jira Ticket or Feature Description
         Output: /output-dir with code/, tests/, design/, docs/

The code review stage includes an auto-fix loop. Blocking findings route the pipeline back to the development stage for a retry, up to MAX_REVIEW_ITERATIONS:

Development → Code Review ──PASS──→ Testing
                   │
                  FAIL (blocking findings)
                   │
                   └──→ Development (retry, max MAX_REVIEW_ITERATIONS)

SDLC Phase Details

Phase Stage Runner Key Inputs Key Outputs
1 · Design Design Agent Jira ticket, repo paths Architecture analysis, impacted components, risks, acceptance criteria, implementation plan
2 · Development Development Agent Design analysis, implementation plan Go source files, PR description, API types
2.5 · Code Review Code Review Agent Generated code BLOCKING/WARNING findings, pass/fail verdict, auto-fix loop
3 · Testing Testing Agent Design + code Unit, integration, e2e Ginkgo v2 test files, test plan
4 · Documentation Documentation Agent All phase outputs PR summary, release notes, SHIP docs, JTBD docs
Publish publish.py Output directory GitHub PR, Jira comments/attachments

Repository Support

Configure repository paths so stage runners can analyze actual Go types, CRDs, and controllers from source. Use repos.yaml for multi-repo setups:

cp repos.yaml.example repos.yaml  # edit with your local clone paths

See Configuration for details. Environment variables (SHIPWRIGHT_REPO_PATH, OPENSHIFT_BUILDS_REPO_PATH) are also supported for single-repo setups.


Quick Start

# 1. Install
git clone <repo> && cd muilti-agents-ocp-builds
uv venv && uv pip install -r requirements.txt
cp .env.example .env  # fill in credentials

# 2. Run the dashboard
uv run python scripts/run_dashboard.py  # http://localhost:8080

Open http://localhost:8080 and click New Run. Enter a Jira ticket ID (e.g. BUILD-1707) or a feature description, then click Run Feature. The pipeline orchestrates all five stages and shows progress in real time.

For CLI and automation usage, see the CLI Reference.

See Quick Start for full credential setup (Vertex AI, Jira, GitHub).


Output Structure

output/BUILD-1707/
├── design/
│   ├── design_analysis.md      # Architecture decisions, impacted components
│   └── implementation_plan.md  # Step-by-step implementation guide
├── code/                       # Production Go source files
├── tests/
│   ├── unit/                   # Ginkgo v2 unit tests
│   ├── integration/            # Integration test files
│   └── e2e/                    # End-to-end test files
├── docs/
│   ├── pr_description.md       # Ready-to-use GitHub PR description
│   └── release_notes.md        # Release notes for the feature
└── state.json                  # Full pipeline state snapshot

Documentation

User Guide — full reference

Section Content
Quick Start Dashboard-first setup: install, configure credentials, run
Dashboard Web UI pages, real-time monitoring, session management
Architecture Pipeline design, state management, security layers
Stage Runners Per-stage reference
CLI Reference All CLI commands for scripting and automation
API Reference REST API endpoints, heartbeat protocol, database schema
Authentication Vertex AI setup
Publishing Push to GitHub / Jira

Requirements

Requirement Detail
Python 3.12+
Claude Google Vertex AI (Application Default Credentials)
Jira API token for ticket fetching
GitHub PAT for PR enrichment and publishing
Qodo Optional — Claude is used as fallback if not installed

About

Multi-agent orchestration system for Shipwright and OpenShift builds development using Claude API + LangGraph

Resources

Stars

Watchers

Forks

Packages

 
 
 

Contributors