Spaces:
Running
Running
| #!/usr/bin/env python | |
| """Weekly LOCAL CFR refresh: federal Clean Fuel Regulations approved CI list (ECCC). | |
| Cloud egress unverified for the ECCC endpoint (INS precedent: GSE blocked HF Jobs + GitHub Actions | |
| datacenter IPs); the file is tiny and weekly cadence suffices, so this refreshes HERE, local, and | |
| publishes through the same private-dataset candidate/release flow the daily and INS refreshes use | |
| -- never a direct serving-data write, never auto-activation. | |
| fetch ECCC CSV via /api/file (browser UA) -> schema assert (refuse loudly on drift) | |
| -> normalize (cfr.py) -> corpus_watch row-count bands | |
| (data/cfr_registers_counts.json -- a DEDICATED manifest, see tests/test_cfr_registers_counts.py | |
| for why this never touches the shared data/corpus_counts.json) | |
| -> snapshot-diff for change events -> data_plane.publish | |
| Any download/schema/band failure leaves current serving data untouched (no publish call at all) | |
| and is recorded loudly in the report. Registered as a weekly Windows Scheduled Task | |
| ("Sinetti PC weekly CFR refresh"), weekend slot (see docs/ins-local-refresh-runbook.md for the sibling | |
| INS task this mirrors). | |
| Usage: | |
| python scripts/cfr_local_refresh.py # full run: fetch + publish | |
| python scripts/cfr_local_refresh.py --dry-run # config check only, no network | |
| """ | |
| from __future__ import annotations | |
| import argparse | |
| import json | |
| import os | |
| import shutil | |
| import sys | |
| import tempfile | |
| from datetime import datetime, timezone | |
| from pathlib import Path | |
| ROOT = Path(__file__).parent.parent | |
| sys.path.insert(0, str(ROOT)) | |
| os.environ.setdefault("APP_BACKEND", "local") | |
| def _devnull_headless_streams() -> None: | |
| """Under pythonw.exe (the Scheduled Task launcher) there is no console: sys.stdout/stderr | |
| are None and ANY print/progress-bar write dies with "'NoneType' object has no attribute | |
| 'write'". Route them to devnull so the run only speaks via runner.log + the report file.""" | |
| if sys.stdout is None: | |
| sys.stdout = open(os.devnull, "w", encoding="utf-8") # noqa: SIM115 — lives for the process | |
| if sys.stderr is None: | |
| sys.stderr = open(os.devnull, "w", encoding="utf-8") # noqa: SIM115 | |
| _devnull_headless_streams() | |
| import cfr # noqa: E402 | |
| import data_plane # noqa: E402 | |
| import scripts.refresh_quality as refresh_quality # noqa: E402 | |
| from sinetti_services import corpus_watch # noqa: E402 | |
| from sinetti_services import source_checks # noqa: E402 | |
| REPORT_NAME = "last_cfr_local_refresh.json" | |
| LOG_NAME = "runner.log" | |
| DEFAULT_STAGING_DIR = ROOT / "data" / "refresh" / "cfr_local" | |
| def _manifest_path() -> Path: | |
| return ROOT / "data" / "cfr_registers_counts.json" | |
| def load_env_file(path, environ=os.environ) -> int: | |
| """Load KEY=VALUE lines from a .env file into `environ`. Existing keys are NEVER overwritten | |
| and values are NEVER printed. Returns the number of keys loaded. Missing file -> 0.""" | |
| p = Path(path) | |
| if not p.exists(): | |
| return 0 | |
| loaded = 0 | |
| for line in p.read_text(encoding="utf-8").splitlines(): | |
| line = line.strip() | |
| if not line or line.startswith("#") or "=" not in line: | |
| continue | |
| key, _, value = line.partition("=") | |
| key = key.strip() | |
| value = value.strip().strip('"').strip("'") | |
| if key and key not in environ: | |
| environ[key] = value | |
| loaded += 1 | |
| return loaded | |
| def _cached_login_token() -> str: | |
| try: | |
| from huggingface_hub import get_token | |
| return get_token() or "" | |
| except Exception: | |
| return "" | |
| def _log_line(staging: Path, text: str) -> None: | |
| staging.mkdir(parents=True, exist_ok=True) | |
| stamp = datetime.now(timezone.utc).isoformat(timespec="seconds") | |
| with open(staging / LOG_NAME, "a", encoding="utf-8") as fh: | |
| fh.write(f"{stamp} {text}\n") | |
| def _write_report(staging: Path, report: dict) -> None: | |
| staging.mkdir(parents=True, exist_ok=True) | |
| (staging / REPORT_NAME).write_text(json.dumps(report, indent=2, default=str), encoding="utf-8") | |
| _SERVING_PAIR = ("sinetti_app.sqlite.gz", "sinetti_data_release.json") | |
| def _artifact_pathway_count(artifact_gz: Path) -> int: | |
| return _artifact_row_count(artifact_gz, "v_app_pathways") | |
| def _artifact_row_count(artifact_gz: Path, table: str) -> int: | |
| import gzip as _gzip | |
| import sqlite3 as _sqlite3 | |
| with tempfile.NamedTemporaryFile(suffix=".sqlite", delete=False) as tmp: | |
| tmp_path = Path(tmp.name) | |
| try: | |
| with _gzip.open(artifact_gz, "rb") as fi, open(tmp_path, "wb") as fo: | |
| shutil.copyfileobj(fi, fo) | |
| conn = _sqlite3.connect(tmp_path) | |
| try: | |
| return int(conn.execute(f"SELECT COUNT(*) FROM {table}").fetchone()[0]) | |
| finally: | |
| conn.close() | |
| finally: | |
| tmp_path.unlink(missing_ok=True) | |
| def _rebuild_candidate_serving_pair(candidate_root: Path, rebuild_fn=None) -> dict: | |
| """Rebuild the candidate's serving pair from its own data/, validate, and swap only the | |
| verified pair back into the candidate. Any failure leaves the candidate's previous pair | |
| untouched and refuses the publish.""" | |
| if rebuild_fn is None: | |
| import scripts.daily_refresh as daily_refresh | |
| rebuild_fn = daily_refresh.rebuild_bundled_app_database | |
| import sinetti_data_release | |
| candidate_data = candidate_root / "data" | |
| try: | |
| with tempfile.TemporaryDirectory(prefix="sinetti-cfr-rebuild-") as temp_name: | |
| staged_root = Path(temp_name) / "candidate" | |
| staged_data = staged_root / "data" | |
| shutil.copytree(candidate_data, staged_data) | |
| app_db = rebuild_fn(root=staged_root, replace=True) | |
| if app_db.get("status") != "ok": | |
| raise RuntimeError(str(app_db.get("error") or "candidate rebuild did not complete")) | |
| validation = sinetti_data_release.validate_data_release( | |
| staged_data / "sinetti_app.sqlite.gz", | |
| manifest_path=staged_data / "sinetti_data_release.json", | |
| inspect_database=True, | |
| ) | |
| if not validation.get("ok"): | |
| raise RuntimeError(str(validation.get("reason") or "rebuilt serving pair is invalid")) | |
| pathway_count = _artifact_pathway_count(staged_data / "sinetti_app.sqlite.gz") | |
| if pathway_count <= 0: | |
| raise RuntimeError( | |
| "rebuilt serving pair has zero v_app_pathways rows " | |
| "(structured PDF layer missing from the candidate workspace?)" | |
| ) | |
| sites_count = _artifact_row_count(staged_data / "sinetti_app.sqlite.gz", "sites") | |
| if sites_count <= 0: | |
| raise RuntimeError( | |
| "rebuilt serving pair has zero sites rows " | |
| "(structured DB predates the sites layer, or structured_sites is empty?)" | |
| ) | |
| for name in _SERVING_PAIR: | |
| shutil.copy2(staged_data / name, candidate_data / name) | |
| return { | |
| "status": "ok", | |
| "certificates": app_db.get("certificates"), | |
| } | |
| except Exception as exc: # noqa: BLE001 -- refuse the publish, never crash the run | |
| return {"status": "error", "error": f"{type(exc).__name__}: {str(exc)[:300]}"} | |
| def _fetch_and_normalize_cfr(fetch_fn) -> dict: | |
| """Fetch + schema-assert + normalize CFR. Returns a refresh-leg report dict; never raises -- | |
| a fetch or schema-drift failure comes back as decision='kept_previous'.""" | |
| try: | |
| header, rows = fetch_fn() | |
| cfr.assert_schema(header) | |
| except Exception as exc: # noqa: BLE001 -- any failure here must refuse, never crash the run | |
| return {"status": "error", "decision": "kept_previous", | |
| "error": f"{type(exc).__name__}: {exc}"} | |
| certs, excluded = cfr.normalize_pathways(rows) | |
| return {"status": "ok", "decision": "accepted", "rows_total": len(rows), | |
| "certs": len(certs), "excluded": len(excluded), "raw_rows": rows, "certs_list": certs} | |
| def _snapshot_status_counts(certs_list: list[dict]) -> dict[str, int]: | |
| return { | |
| "cfr.rows_total": 0, # filled by caller from rows_total | |
| "cfr.active_in_snapshot": sum(1 for c in certs_list if c.get("status") == "active"), | |
| "cfr.inactive_in_snapshot": sum(1 for c in certs_list if c.get("status") == "withdrawn"), | |
| } | |
| def main(argv: list[str] | None = None, *, fetch_cfr_fn=None, probe_cfr_fn=None, | |
| rebuild_fn=None, environ=os.environ, env_file=None, | |
| _candidate_root: Path | None = None) -> int: | |
| from sinetti_operator.registration import register_live_providers | |
| register_live_providers() # jobs never reach full_registration (SIN-0999) | |
| env_path = env_file if env_file is not None else ROOT / ".env" | |
| load_env_file(env_path, environ=environ) | |
| ap = argparse.ArgumentParser(description=__doc__) | |
| ap.add_argument("--dry-run", action="store_true", help="print config, no network calls") | |
| ap.add_argument("--staging-dir", default=str(DEFAULT_STAGING_DIR)) | |
| args = ap.parse_args(argv) | |
| staging = Path(args.staging_dir) | |
| token = (environ.get("HF_TOKEN", "") or environ.get("HF_JOB_TOKEN", "") | |
| or environ.get("HF_CREATE_TOKEN", "") or _cached_login_token()) | |
| data_repo_id = environ.get("SINETTI_DATA_REPO_ID", "") | |
| if args.dry_run: | |
| print(json.dumps({"dry_run": True, "data_repo_id": data_repo_id, | |
| "staging_dir": str(staging), "hf_token_present": bool(token)}, indent=2)) | |
| return 0 | |
| if not data_repo_id: | |
| print("REFUSING: SINETTI_DATA_REPO_ID not set -- this pipeline publishes only through " | |
| "the private dataset candidate/release flow, no legacy fallback.") | |
| _log_line(staging, "refused: SINETTI_DATA_REPO_ID missing") | |
| return 2 | |
| if not token: | |
| print("REFUSING: no HF token (HF_TOKEN / HF_JOB_TOKEN / HF_CREATE_TOKEN) in environment, " | |
| ".env, or a cached `huggingface-cli login`.") | |
| _log_line(staging, "refused: HF_TOKEN missing") | |
| return 2 | |
| if _candidate_root is None: | |
| workspace = tempfile.TemporaryDirectory(prefix="sinetti-cfr-candidate-") | |
| try: | |
| return main( | |
| argv, fetch_cfr_fn=fetch_cfr_fn, rebuild_fn=rebuild_fn, | |
| environ=environ, env_file=env_file, | |
| _candidate_root=Path(workspace.name), | |
| ) | |
| finally: | |
| workspace.cleanup() | |
| candidate_root = Path(_candidate_root) | |
| candidate_data = candidate_root / "data" | |
| baseline = data_plane.boot_sync( | |
| repo_id=data_repo_id, token=token, cache_dir=candidate_root / "data-cache", | |
| data_dir=candidate_data, materialize=True, | |
| ) | |
| baseline_failed = baseline.get("status") != "ok" or not baseline.get("hub_revision") | |
| started = datetime.now(timezone.utc) | |
| run_id = started.isoformat() | |
| if baseline_failed: | |
| report = { | |
| "started_at": run_id, "runner": "cfr-local-refresh", | |
| "baseline": baseline, | |
| "refresh": {"cfr": {"status": "error", "decision": "kept_previous", | |
| "error": "verified dataset baseline unavailable"}}, | |
| "upload": {"status": "skipped-baseline-error"}, | |
| "status": "error", | |
| } | |
| _write_report(staging, report) | |
| _log_line(staging, "status=error rc=1 reason=baseline_failed") | |
| return 1 | |
| fetch_cfr_fn = fetch_cfr_fn or cfr.fetch_cfr_rows | |
| # Ask ECCC before downloading. Measured 2026-08-14: the CI list's Last-Modified was | |
| # 2025-12-04 and a live fetch was row-identical to our snapshot -- this weekly task had | |
| # been re-parsing an eight-month-old file every week since it was registered. | |
| cfr_gate = source_checks.gate( | |
| "CFR", | |
| probe_fn=probe_cfr_fn or cfr.probe_cfr_source, | |
| bundle_path=candidate_data / "cfr" / "pathways.json.gz", | |
| checked_at=run_id, | |
| ) | |
| unchanged = bool(cfr_gate["skip"]) | |
| if unchanged: | |
| cfr_leg = {"status": "unchanged", "decision": "kept_previous", | |
| "reason": cfr_gate["reason"], "signal": cfr_gate["signal"], | |
| "source_last_modified": cfr_gate["source_last_modified"]} | |
| else: | |
| cfr_leg = _fetch_and_normalize_cfr(fetch_cfr_fn) | |
| cfr_leg["probe"] = cfr_gate["reason"] | |
| leg_ok = cfr_leg["status"] == "ok" | |
| bands_report: dict = {"status": "skipped", "reason": "refresh not accepted"} | |
| if leg_ok: | |
| counts = _snapshot_status_counts(cfr_leg["certs_list"]) | |
| counts["cfr.rows_total"] = cfr_leg["rows_total"] | |
| manifest = corpus_watch.load_manifest(_manifest_path()) | |
| bands_report = corpus_watch.counts_report(counts, manifest) | |
| ok = leg_ok and bands_report.get("status") == "ok" | |
| change_events: list = [] | |
| if ok: | |
| old_cfr = cfr.load_cfr_certs(path=candidate_data / "cfr" / "pathways.json.gz") | |
| change_events = refresh_quality.build_register_diff_events( | |
| run_id, "CFR", old_cfr, cfr_leg["certs_list"], | |
| cfr.diff_cfr_snapshot(old_cfr, cfr_leg["certs_list"]), | |
| detected_at=run_id, | |
| ) | |
| cfr.write_snapshot( | |
| cfr_leg["certs_list"], path=candidate_data / "cfr" / "pathways.json.gz", | |
| fetched_at=run_id, source="cfr_local_refresh", | |
| source_last_modified=cfr_gate["source_last_modified"] or None, | |
| ) | |
| app_db_report: dict = {"status": "skipped", "reason": "refresh not accepted"} | |
| if ok: | |
| app_db_report = _rebuild_candidate_serving_pair(candidate_root, rebuild_fn) | |
| ok = app_db_report.get("status") == "ok" | |
| def _leg_report(leg: dict) -> dict: | |
| return {k: v for k, v in leg.items() if k not in ("raw_rows", "certs_list")} | |
| report: dict = { | |
| "started_at": run_id, "runner": "cfr-local-refresh", | |
| "baseline": baseline, | |
| "refresh": { | |
| "cfr": {**_leg_report(cfr_leg), "change_events": change_events} if cfr_leg["status"] == "ok" else cfr_leg, | |
| "bands": bands_report, | |
| "app_db": app_db_report, | |
| }, | |
| "upload": {"status": "pending"}, | |
| } | |
| rc = 0 | |
| if unchanged: | |
| report["upload"] = {"status": "skipped-unchanged", | |
| "reason": "ECCC reported the CI list unchanged; nothing to publish"} | |
| report["status"] = "unchanged" | |
| elif not leg_ok: | |
| report["upload"] = {"status": "skipped-refresh-error"} | |
| rc = 1 | |
| elif bands_report.get("status") != "ok": | |
| report["upload"] = {"status": "skipped-band-violation"} | |
| rc = 1 | |
| elif app_db_report.get("status") != "ok": | |
| report["upload"] = {"status": "skipped-rebuild-error"} | |
| rc = 1 | |
| else: | |
| try: | |
| result = data_plane.publish( | |
| candidate_root, repo_id=data_repo_id, token=token, | |
| producer="cfr_local_refresh", expected_parent=baseline.get("hub_revision"), | |
| ) | |
| report["upload"] = result | |
| if result.get("status") != "ok": | |
| rc = 1 | |
| except Exception as exc: # noqa: BLE001 | |
| report["upload"] = {"status": "error", | |
| "exception": {"class": type(exc).__name__, "message": str(exc)}} | |
| rc = 1 | |
| finished = datetime.now(timezone.utc) | |
| if not (rc == 0 and report.get("status") == "unchanged"): | |
| report["status"] = "ok" if rc == 0 else "error" | |
| report["finished_at"] = finished.isoformat() | |
| report["duration_s"] = round((finished - started).total_seconds(), 2) | |
| _write_report(staging, report) | |
| refresh_quality.append_history_files(staging, report) | |
| _log_line(staging, f"status={report['status']} rc={rc} upload={report['upload'].get('status')}") | |
| return rc | |
| if __name__ == "__main__": | |
| raise SystemExit(main()) | |