Skip to main content

GitLab Model Registry Integration

GitLab Model Registry Integration

Overview

The gitlab_components project now includes native GitLab Model Registry integration (GA in GitLab 17.6+).

Components

1. [object Object]

Native GitLab Model Registry integration with:

  • Model registration via GitLab API
  • Semantic versioning (auto-generated or manual)
  • CI/CD job linking
  • Multi-stage promotion (development staging production)

2. [object Object]

Bidirectional sync between MLflow and GitLab Model Registry:

  • Sync MLflow experiments to GitLab
  • Migrate artifacts to GitLab Package Registry
  • Stage mapping (Production production, Staging staging)

3. [object Object]

Automated model promotion with quality gates:

  • Staging: accuracy 0.85, F1 0.80
  • Production: accuracy 0.90, F1 0.85 (manual approval)
  • Champion/challenger comparison

Usage

include:
  - component: gitlab.com/blueflyio/agent-platform/gitlab_components/ml/gitlab-model-registry@main
    inputs:
      model_name: "my-model"
      model_version: "1.0.0"

Neo4j Knowledge Graph

The Neo4j integration has been enhanced with ML entities:

  • Nodes: MLExperiment, MLRun, MLModel, MLMetric
  • Relationships: TRAINS, HAS_VERSION, SUPERSEDES, EVALUATES