GitHub - mitkox/ai-coding-factory: AI Coding Factory. Replace your software outsourcing vendor contracts with your own AI. (original) (raw)

Enterprise-grade internal software delivery platform for private, traceable, and governed .NET development. Open source and GitHub-ready, with optional Azure DevOps support.

Repository: https://github.com/mitkox/ai-coding-factory

Features | Quick Start | Governance | How To Verify

Overview

AI Coding Factory turns local inference into an internal delivery platform that replaces outsourcing with auditable, automated, and private software delivery. The platform enforces quality, security, traceability, and governance by default while staying fully offline-capable.

Features

Privacy-First AI Delivery

• Local inference with vLLM or LM Studio (OpenAI-compatible API)
• .env.example for configuration; no secrets in repo
• Air-gapped deployment guidance and least-privilege agent permissions

Enterprise Governance and Traceability

• Definition of Done/Ready templates with explicit traceability rules
• Story -> Test -> Commit -> Release enforcement
• Automated traceability reports and release notes
• Governance policy covering branching, reviews, ownership, and risk

Enterprise DevOps Choice (GitHub or Azure DevOps)

• GitHub Issues/Projects or Azure Boards for backlog and story tracking
• GitHub or Azure Repos for PRs and code review
• GitHub Actions or Azure Pipelines for CI quality gates and release readiness

.NET 8+ Clean Architecture Templates

• Clean Architecture solution template with DDD/CQRS ready structure
• Microservice template with Kubernetes manifests
• Documentation and testing requirements baked in

Agile-Native Delivery with AI Scrum Teams

• Scrum Team as Code: PO, Scrum Master, Dev, QA, Security, DevOps agents
• Sprint planning, execution, review, and retrospective workflows
• Traceability and quality gates enforced by agents

Agents and Skills

Stage agents: ideation, prototype, poc, pilot, product
Scrum team agents: product-owner, scrum-master, developer, qa, security, devops
Skills: 12 reusable .NET skills in .opencode/skill

Repository Map

ai-coding-factory/
├── .opencode/                      # Agents, skills, prompts, templates
├── docs/                           # Governance, traceability, testing, Scrum
├── templates/                      # Clean Architecture + microservice templates
├── scripts/                        # Validation, traceability, scaffold verification
├── artifacts/                      # Traceability outputs (sample + generated)
├── azure-pipelines.yml             # Azure DevOps CI pipeline
├── .github/workflows/quality-gates.yml # GitHub Actions CI pipeline (optional)
└── README.md

Governance and Traceability

• Corporate R&D Development Policy (authoritative): CORPORATE_RND_POLICY.md
• Governance policy: docs/governance/GOVERNANCE.md
• Traceability model: docs/traceability/TRACEABILITY.md
• Scrum Team as Code: docs/agile/SCRUM-TEAM-AS-CODE.md
• Testing strategy: docs/testing/TESTING-STRATEGY.md
• Documentation requirements: docs/documentation/DOCUMENTATION-REQUIREMENTS.md

Quick Start

Prerequisites

• .NET 8 SDK
• Python 3
• Docker (for container verification)
• Local inference server (vLLM or LM Studio)
• GitHub or Azure DevOps project (Issues/Boards, Repos, Actions/Pipelines)

0) Clone the Repository (GitHub)

git clone https://github.com/mitkox/ai-coding-factory.git cd ai-coding-factory

1) Configure Local Inference

Update .opencode/opencode.json or set values in .env.example:

{ "provider": { "local-inference": { "type": "openai-compatible", "baseUrl": "http://localhost:8000/v1", "apiKey": "your-api-key" } } }

2) Start Inference Server

vLLM:

vllm serve GLM-4.7 --dtype auto --api-key your-api-key

LM Studio:

• Load a model
• Enable OpenAI-compatible server

3) Run OpenCode

opencode /agent product-owner

4) Connect to GitHub or Azure DevOps

• Create a GitHub or Azure DevOps project
• Use GitHub Issues/Projects or Azure Boards for story tracking
• Configure GitHub Actions or Azure Pipelines to run CI checks

How To Verify

Run the full lifecycle verification (build, test, coverage, traceability, container):

chmod +x scripts/scaffold-and-verify.sh ./scripts/scaffold-and-verify.sh

Additional checks:

./scripts/validate-project.sh ./scripts/validate-documentation.sh ./scripts/validate-rnd-policy.sh python3 scripts/traceability/traceability.py validate --commit-range origin/main..HEAD

If you use Azure DevOps, the same checks run in azure-pipelines.yml. For GitHub, use .github/workflows/quality-gates.yml.

Autopilot (Story → PR → Evidence Pack)

Autopilot turns an ACF-### story into a consistent delivery workflow: branch naming, PR creation (GitHub or Azure DevOps), and an auditable Evidence/Review Pack.

• Script: scripts/autopilot/autopilot.py
• Docs: scripts/autopilot/README.md

Security and Offline Guidance

• No secrets committed; use .env locally
• Inference endpoints should bind to localhost or a protected subnet
• Use least-privilege permissions in .opencode/opencode.json
• Air-gapped use: mirror dependencies and disable external MCP integrations

License

MIT License. See LICENSE.

Made with ❤️

Made with ❤️ by Mitko X.