Skip to content

VLStream Cloud v1.2.7

Choose a tag to compare

@github-actions github-actions released this 18 Oct 05:53
· 58 commits to main since this release
Immutable release. Only release title and notes can be modified.

VLStream Cloud v1.2.7

Built from the current mainline at d4a39cb1, with the production Flyway history
continued from v1.2.6.

Highlights

  • Add a unified protocol-aware device square with clearer source-availability
    status and safer handling of offline devices.
  • Expand tenant-scoped GPU compute with company management, shared task queues,
    provider-backed inventory and guarded marketplace endpoints.
  • Persist and validate training datasets asynchronously; improve model artifact
    handling and add an optional isolated CPU worker for Hi3519DV500 OM conversion.
  • Harden local-account and tenant identity flows, and enforce tenant-scoped event
    media access and deletion.
  • Improve workflow guidance and localization across the console.

Compatibility and database upgrades

  • Deploy WVP separately using VLStream WVP Lite v1.0.8,
    image ghcr.io/oortcloudgroup/vlstream-cloud-lite:1.0.8. WVP remains the
    device center and owns ZLMediaKit; the VLS Compose project does not start a
    second WVP or ZLMediaKit stack.
  • Backend and frontend images:
    ghcr.io/oortcloudgroup/vlstream-backend:1.2.7 and
    ghcr.io/oortcloudgroup/vlstream-frontend:1.2.7.
  • Production upgrades continue the public v1.2.6 Flyway history and apply
    V1_2_0_029 through V1_2_0_048. prod selects
    db/migration/release; dev and local retain the separate mainline
    history. Never change migration locations for a database that already has
    Flyway history.
  • Fresh installations import the sanitized VLS bootstrap schema and the
    reviewed WVP v1.0.8 schema; Flyway then applies the release migrations. The
    deployment archive includes the optional OM-conversion Compose overlay, but
    not vendor SVP/ATC software or its licenses.

Cloud-compute behavior depends on tenant configuration, provider access, network
reachability and the target GPU runtime. The optional OM worker is isolated and
disabled unless explicitly enabled; successful conversion still requires
device-side validation. Release checks do not establish model quality or
compatibility for every GPU or device.

Full Changelog: https://github.com/OortCloudGroup/VLStream-Cloud/commits/v1.2.7