Skip to content

Encrypted Projects, Expanded Infrastructure, and MLflow Gateway

Latest
Latest

Choose a tag to compare

View all tags
@nicococo nicococo released this 12 Jun 11:39
· 24 commits to main since this release
v0.3.0
578a9a8
This commit was created on GitHub.com and signed with GitHub’s verified signature.
GPG key ID: B5690EEEBB952194
Verified
Learn about vigilant mode.

MLOX v0.3.0 is a major step toward a unified platform for deploying and managing connected ML/AI infrastructure.

Highlights

  • Encrypted project files: Projects now use portable SQLCipher-encrypted .mlox databases containing metadata, infrastructure state, and secrets.
  • Unified workspace API: ProjectWorkspace replaces the previous session and application APIs with a clearer persistence and lifecycle model.
  • Expanded server support: New Multipass-based Ubuntu servers support Native, Docker, and K3s environments. Lightweight connector servers represent externally hosted services.
  • Service capabilities: Service templates now declare and validate their infrastructure requirements, improving placement and compatibility checks.
  • MLflow Gateway: A lightweight model registry gateway adds model-reference resolution, alias management, prediction support, and LRU/TTL model caching.
  • New Ollama service: Deploy standalone Ollama instances through MLOX.
  • Improved secrets management: OpenBao receives hardened initialization, Raft storage, audit logging, periodic credentials, and improved token handling.
  • Better operations: The TUI now offers richer service details, application logs, lifecycle actions, and direct local or SSH terminal access.
  • Improved service integrations: Updates include PostgreSQL catalog inspection, Registry UI support, KubeApps improvements, GCP connector integration, and enhanced OpenTelemetry resource metrics.
  • Architecture and documentation: The project structure, public APIs, guides, examples, and test coverage have been substantially revised.

Important Migration Notice

This release replaces the legacy encrypted .project file and externally stored infrastructure state with a single SQLCipher-encrypted .mlox file.

Legacy projects must be migrated explicitly:

python scripts/migrate_legacy_project.py \
  ./legacy.project ./projects/demo.mlox \
  --legacy-password "$OLD_PASSWORD" \
  --new-password "$NEW_PASSWORD"

The migration copies the project without modifying the original. Keep the legacy project and its secret store until the migrated project has been verified.

See Encrypted Project Files for creation, backup, security, and migration details.

Assets 2
Loading
0 Join discussion