Enterprise AI Orchestrator for GitLab

Repository-aware developer workflow agent for release confidence.

Comment

Inspiration

Coding agents can generate implementation quickly, but release confidence still depends on context, tests, evidence, and human review. GitLab is where those signals already live: merge requests, CI status, ownership, dependencies, deployments, and code history.

What it does

Enterprise AI Orchestrator for GitLab demonstrates a repository-aware release workflow. It reviews GitLab Orbit-style context, assesses release risk, proposes targeted test work, assembles evidence, and makes the human approval checkpoint visible before release promotion.

GitLab Duo / Orbit artifact

The repository includes a project-level GitLab Duo Agent Skill at skills/release-risk-orbit/SKILL.md. The skill defines how an agent should use GitLab Orbit context from merge requests, CI, ownership, dependency, and deployment signals to produce release-risk recommendations, test actions, and approval evidence.

The current hosted dashboard is a safe simulated prototype. It does not call live GitLab Orbit or mutate GitLab repositories, branches, merge requests, issues, or CI settings. If live Orbit access is available, the skill is the integration path for replacing simulated context with real Orbit query output.

How we built it

The MVP is a React, TypeScript, and Vite application with local workflow state. It is organized around release risk, repo context, agent workflow, impacted tests, human approval, and evidence. The GitLab-specific artifact is documented in the public repo as an installable skill-style workflow.

Challenges

The challenge was avoiding a generic chatbot demo. The useful workflow is operational: connect repo context to risk, tests, evidence, and approval, while keeping external mutations behind explicit human approval.

Accomplishments

  • Built a responsive dashboard MVP.
  • Added GitLab-specific positioning and workflow language.
  • Added a GitLab Duo Agent Skill artifact for the Orbit release-risk workflow.
  • Prepared a public GitHub repo, GitHub Pages demo, CI, README, MIT license, and submission packet.
  • Kept the integration boundary safe and explicit.

What's next

Next steps are live GitLab Orbit query evidence, merge request context ingestion, CI status ingestion, generated test patches, review comments, and evidence bundles.

GitHub repo: https://github.com/zemeng2015/enterprise-ai-engineering-orchestrator Live demo: https://zemeng2015.github.io/enterprise-ai-engineering-orchestrator/ Demo video: https://youtu.be/8AKMY8VoN7c

Public proof package

GitLab proof

The GitLab-specific artifact is a project-level GitLab Duo Agent Skill:

https://raw.githubusercontent.com/zemeng2015/enterprise-ai-engineering-orchestrator/main/skills/release-risk-orbit/SKILL.md

Boundary: the hosted dashboard uses safe simulated GitLab context. The repository includes the reusable skill artifact and the live demo exposes it in the Submission Proofs panel.

Built With

Share this project:

Updates