mirror of
https://github.com/mudler/LocalAI.git
synced 2026-07-30 18:09:05 -04:00
98a0e3d7fa
The nightly denormalization job edits YAML nodes instead of round-tripping structs so its PR stays small enough for a human to review, but the write path undid that: yaml.Marshal re-encoded the node tree at yaml.v3's default 4-space indent and dropped the leading document marker, reflowing roughly 6000 lines around the handful of real changes. Encode through yaml.NewEncoder at the index's authored 2-space indent and restore the header. A write that changes three fields now changes three lines. Stale inferred_min_vram values were also never cleared. Both skip paths (an authored min_vram is present, or the candidate is the last resort) returned before touching the field, so a candidate that gained a floor or became the last resort after a reorder kept an inferred value that EffectiveMinVRAM reported as a real constraint, failing the meta lint with no way for the job to self-heal. Clear the field before both skips. The workflow discarded a whole night's work on any single failure: the program exits 1 when a candidate cannot be estimated, which aborted the job before the PR step, so one unreachable candidate blocked every other refresh indefinitely. Capture the status, open the PR with what was computed, mark the PR body as partial, and fail the run afterwards so the problem still surfaces. Also preserve the index's existing file mode instead of forcing 0644, and drop the redundant //go:build ignore tag, since Go already skips dot directories and the sibling modelslist.go carries no tag. Assisted-by: Claude:claude-opus-4-8 Signed-off-by: Ettore Di Giacinto <mudler@localai.io>
57 lines
2.4 KiB
YAML
57 lines
2.4 KiB
YAML
name: Denormalize gallery meta entries
|
|
on:
|
|
schedule:
|
|
- cron: 0 22 * * *
|
|
workflow_dispatch:
|
|
jobs:
|
|
denormalize:
|
|
if: github.repository == 'mudler/LocalAI'
|
|
runs-on: ubuntu-latest
|
|
steps:
|
|
- uses: actions/checkout@v7
|
|
- uses: actions/setup-go@v5
|
|
with:
|
|
go-version-file: go.mod
|
|
# The program exits 1 when any candidate could not be estimated, but it
|
|
# still writes everything it did compute. Aborting here would let one
|
|
# unreachable candidate block the refresh of every other one, night after
|
|
# night, so record the status and let the PR open with the partial result.
|
|
- name: Denormalize meta candidates
|
|
id: denormalize
|
|
run: |
|
|
set +e
|
|
go run ./.github/ci/gallery_denormalize.go gallery/index.yaml
|
|
status=$?
|
|
set -e
|
|
echo "status=$status" >> "$GITHUB_OUTPUT"
|
|
if [ "$status" -ne 0 ]; then
|
|
echo "partial=yes" >> "$GITHUB_OUTPUT"
|
|
echo "note=**This run was partial.** At least one candidate could not be estimated, so some fields may be missing or stale. See the workflow log." >> "$GITHUB_OUTPUT"
|
|
else
|
|
echo "partial=no" >> "$GITHUB_OUTPUT"
|
|
echo "note=All candidates were processed successfully." >> "$GITHUB_OUTPUT"
|
|
fi
|
|
- name: Create Pull Request
|
|
uses: peter-evans/create-pull-request@v8
|
|
with:
|
|
token: ${{ secrets.UPDATE_BOT_TOKEN }}
|
|
push-to-fork: ci-forks/LocalAI
|
|
commit-message: ':arrow_up: Denormalize meta model candidates'
|
|
title: 'chore(model-gallery): :arrow_up: refresh meta candidate metadata'
|
|
branch: "update/gallery-denormalize"
|
|
body: |
|
|
Refreshes the denormalized `backend`, `quantization` and
|
|
`inferred_min_vram` fields on meta model entries.
|
|
|
|
Authored `min_vram` values are never modified.
|
|
|
|
${{ steps.denormalize.outputs.note }}
|
|
signoff: true
|
|
# Surface the failure only after the PR exists, so the problem stays
|
|
# visible without discarding the work that did succeed.
|
|
- name: Report estimation failures
|
|
if: steps.denormalize.outputs.partial == 'yes'
|
|
run: |
|
|
echo "gallery_denormalize exited ${{ steps.denormalize.outputs.status }}: one or more candidates need a hand-authored min_vram."
|
|
exit 1
|