precommitEU ops · Enterprise CI
Run your own AI.
On your infrastructure.
Under your control.
Run the supplied open-source models on your own Azure runner to review every pull request for EU regulatory risks. All AI inference stays on that runner; no external AI service or API key is needed.
Quality was measured on 450 synthetic cases. Cost and timing assume one regulation pack and roughly 10 changed files per PR. Read the evidence, assumptions and limitations.
From pull request to review
A pull request queues a scan. An ephemeral GPU runner checks the changed files with local models, writes findings with quoted code evidence, and returns a check result and downloadable reports. Your organization decides which checks must pass before merging.
The scanner, weights and deployment code are Apache-2.0. There is no software subscription or LLM API fee; your organization provides the compute and controls its software and model versions.
From evaluation to everyday use
The current deployment uses ops 0.2.0, scanner 0.3.2 and models v0.3.0.
The scanner can run air-gapped after installation. This CI deployment connects to GitHub while keeping AI inference on your own runner. See network access for the boundary.