Skip to content

Merge pull request #26 from sensein/claude/schema-registry-meta-model… #3

Merge pull request #26 from sensein/claude/schema-registry-meta-model…

Merge pull request #26 from sensein/claude/schema-registry-meta-model… #3

name: Rebuild & Publish Registry
# Trigger manually or weekly (Sunday 03:00 UTC).
# A schema submission (schema_submission.yml) handles per-PR incremental updates;
# this workflow does a full clean rebuild on a schedule.
on:
push:
branches: [main]
paths:
- "schemas/**"
- "neuro_ghost/**"
- "requirements.txt"
workflow_dispatch:
inputs:
bump:
description: "Version bump (patch / minor / major)"
required: false
default: "patch"
type: choice
options: [patch, minor, major]
skip_converters:
description: "Skip fetching external schemas (BIDS/NWB/DANDI/…)"
required: false
default: false
type: boolean
schedule:
- cron: "0 3 * * 0" # weekly, Sunday 03:00 UTC
jobs:
rebuild:
runs-on: ubuntu-latest
permissions:
contents: write # needed to push data/ back to main
steps:
# -----------------------------------------------------------------------
- name: Checkout
uses: actions/checkout@v4
with:
# Fetch full history so the bump commit doesn't squash provenance
fetch-depth: 0
# -----------------------------------------------------------------------
- name: Set up Python
uses: actions/setup-python@v5
with:
python-version: "3.12"
# -----------------------------------------------------------------------
- name: Cache pip
uses: actions/cache@v4
with:
path: ~/.cache/pip
key: ${{ runner.os }}-pip-${{ hashFiles('requirements.txt') }}
restore-keys: ${{ runner.os }}-pip-
# Sentence-transformers downloads all-MiniLM-L6-v2 (~90 MB) on first use.
# Cache it so rebuilds don't re-download from Hugging Face every week.
- name: Cache sentence-transformers model
uses: actions/cache@v4
with:
path: ~/.cache/huggingface/hub
key: hf-minilm-v2
restore-keys: hf-
# Embeddings are computed once per unique set of schema content and stored
# in data/embeddings.parquet. Cache keyed on schema file hashes so a
# schema change forces a recompute while an unchanged run reuses it.
- name: Cache embeddings
uses: actions/cache@v4
with:
path: data/embeddings.parquet
key: embeddings-${{ hashFiles('schemas/*.yml') }}
restore-keys: embeddings-
# -----------------------------------------------------------------------
- name: Install dependencies
run: pip install -r requirements.txt
# -----------------------------------------------------------------------
- name: Determine pipeline flags
id: flags
run: |
BUMP="${{ inputs.bump || 'patch' }}"
SKIP=""
if [ "${{ inputs.skip_converters }}" = "true" ]; then
SKIP="--skip-converters"
fi
echo "bump=$BUMP" >> $GITHUB_OUTPUT
echo "skip=$SKIP" >> $GITHUB_OUTPUT
# -----------------------------------------------------------------------
# Full pipeline: wipe → seed → (converters) → ingest all → align → export
- name: Seed, ingest, align, export
run: |
python neuro_ghost/pipeline.py \
--fresh \
${{ steps.flags.outputs.skip }} \
--bump ${{ steps.flags.outputs.bump }} \
--agent "github-actions"
# -----------------------------------------------------------------------
- name: Commit & push updated registry
run: |
git config user.name "github-actions[bot]"
git config user.email "github-actions[bot]@users.noreply.github.com"
# Stage registry outputs + any new converted schemas
git add data/registry.json data/versions/ data/provenance.json
git add schemas/ --ignore-errors
if git diff --cached --quiet; then
echo "Registry unchanged — nothing to commit."
else
VERSION=$(python - <<'PY'
import json, pathlib
d = pathlib.Path("data/registry.json")
if d.exists():
print(json.loads(d.read_text()).get("registry_version", ""))
PY
)
git commit -m "chore(registry): rebuild v${VERSION} [skip ci]"
git push origin HEAD:main
fi