QVAC-24253 chore[mod]: bump @qvac/fabric to ^0.10.0 for the npm-runtime addons - #4206
Merged
Merged
Conversation
…me addons classification-ggml and vla-ggml consume the shared runtime through the published npm package @qvac/fabric rather than building the qvac-fabric vcpkg port, so the 10297.1.1 rollout did not reach them. A caret on a 0.x version locks the minor, so neither would ever resolve @qvac/fabric 0.10.0 on its own -- the range has to be bumped by hand, after fabric is actually on npm. @qvac/fabric@0.10.0 is published (npm latest, git tag fabric-v0.10.0), which is what makes this PR resolvable. classification-ggml 0.22.0 -> 0.23.0 @qvac/fabric ^0.8.0 -> ^0.10.0 vla-ggml 0.23.0 -> 0.24.0 @qvac/fabric ^0.9.0 -> ^0.10.0 Both move qvac-fabric 10297.0.0 -> 10297.1.1. classification crosses two fabric minors, so it also picks up 0.9.0's ROCm/HIP DL backend; vla already had it. vla-ggml also releases work that is already on main but unreleased: #3998 migrated it off the vcpkg port onto the npm runtime and deliberately parked its version bump and changelog entry for a follow-up ("Park the fabric-migration changelog and version bump for a follow-up PR after this cycle"). This is that follow-up, so the entry documents the migration, the ASAN test-runner change and the mobile test-group validation alongside the caret bump rather than shipping them silently. classification-ggml's only other unreleased change is a check:generated script tweak, noted in its entry for completeness. Squash-merging this publishes both packages. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Contributor
Review StatusCurrent Status: ❌ PENDING Pending reviews: Needs 1 Management or Team Lead, and 1 more from Management, Team Lead, or Member. |
Contributor
License compliance — cleanNo new dependency license findings in this PR. Warn-only (shadow) mode — this check does not block merges yet. Updated automatically by the canonical license compliance workflow. NOTICE presence (advisory)Missing NOTICE (advisory, does not block):
|
|
Review the following changes in direct dependencies. Learn more about Socket for GitHub.
|
gianni-cor
approved these changes
Sep 1, 2026
donriddo
approved these changes
Sep 1, 2026
This branch was previously deployed
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
🎯 Problem
classification-ggmlandvla-ggmlconsume the shared ggml runtime through the published npm package@qvac/fabric, not theqvac-fabricvcpkg port. Theqvac-fabric 10297.1.1rollout(#4121) therefore never reached them —
.github/fabric-consumers.jsonlists both undernpm_runtimeprecisely because they sit outside it.They cannot pick the new runtime up on their own. A caret on a
0.xversion locks the minor, so^0.8.0and^0.9.0will never resolve0.10.0no matter how long they wait. Left alone, both keepinstalling a superseded fabric indefinitely, and nothing fails loudly to say so.
📝 How
@qvac/fabricclassification-ggml^0.8.0→^0.10.0vla-ggml^0.9.0→^0.10.0Both move
qvac-fabric 10297.0.0→10297.1.1.classification-ggmlcrosses two fabric minors, soit additionally picks up
@qvac/fabric0.9.0's ROCm/HIPGGML_BACKEND_DLmodule;vla-ggmlalready hadit.
Neither addon exposes a GPU or vector-index surface of its own, so no API changes — the bump keeps
them on the current shared runtime rather than a stale one.
vla-ggmlalso releases work already onmain#3998 migrated
vla-ggmloff the vcpkg port onto the npmruntime and deliberately deferred its version bump and changelog entry:
This is that follow-up. Its
0.24.0entry therefore documents the migration, the ASAN C++ test-runnerchange and the mobile test-group validation alongside the caret bump, rather than publishing them
silently under a version whose changelog mentions none of them.
classification-ggml's only other unreleased change is acheck:generatedscript tweak, noted in itsentry for completeness.
🧪 Tested
@qvac/fabric@0.10.0is published and resolvable — npmlatestis0.10.0, git tagfabric-v0.10.0exists. This was the blocking precondition: opening this PR before fabric publishedwould leave its own CI unable to resolve
^0.10.0and go red for an unrelated reason.package.json, 2 ×CHANGELOG.md. No source, no workflows.package.jsonfiles parse, and both changelog headings use the bracketed## [<version>] - <date>form the release-merge-guard requires.latest(0.22.0 / 0.23.0), so neither had an unreleasedversion already pending.
None. Both bumps are minor on
0.x, no API surface changes, and the ROCm/HIP module is fail-safe atruntime — the DL loader skips it on non-AMD hosts and falls back to Vulkan/CPU.
Squash-merging publishes 2 packages:
@qvac/classification-ggml 0.23.0and@qvac/vla-ggml 0.24.0.🤖 Generated with Claude Code