修正 PPT 視覺 QA 多 worker 狀態
All checks were successful
CD Pipeline / deploy (push) Successful in 1m2s

This commit is contained in:
OoO
2026-05-19 09:15:31 +08:00
parent 583e318295
commit 21cdbdb75c
4 changed files with 164 additions and 21 deletions

View File

@@ -115,6 +115,7 @@
- V10.226 補 PPT 視覺 QA runtime checklist`/observability/ppt_audit_history` 在視覺模型未就緒時顯示 Feature Flag、LibreOffice、Vision Model 三段檢查與下一步操作,避免只看到「停用」而不知道卡在哪。
- Phase 54 manual sample candidate queue transaction新增 `/api/market_intel/manual_sample_review/candidate_queue_transaction` POST 與 UI transaction preview 按鈕,將 queue row preview 轉成 `market_alert_review_queue` idempotent insert statement、payload hash 與 rollback plan不開 DB connection、不開 transaction、不 commit、不建立 approval record版本同步至 V10.227。
- V10.228 補 PPT 視覺 QA 背景狀態卡:新增 `/observability/ppt_audit/vision_status` 與頁面 Vision QA 狀態卡,讓立即視覺 QA 排入後可看 queued/running/completed/error 與最近審核摘要,不必刷新猜測。
- V10.229 修正 PPT 視覺 QA 多 worker 狀態漂移:將 queued/running/completed/error 寫入 `/app/data/ppt_vision_audit_status.json` runtime state所有 Gunicorn worker 共用同一份狀態並阻擋重複排入。
- Schema smoke`tests/test_market_intel_skeleton.py` 檢查 `Base.metadata` 內含 ADR-035 八張 `market_*` tables。
- Desktop UI QA本機只註冊 `market_intel_bp` 的 Flask harness 載入 `/market_intel`,確認 Phase 15、候選預覽、writer preview、安全 flags、點陣暖紙視覺正常console error 0。
- API QA`/api/market_intel/schema_smoke` 通過 7 張表與 `market_platforms` 必要欄位檢查;`/api/market_intel/platform_seed_writer_plan` 回傳 4 筆 dry-run upsert preview`writes_executed=false`,四平台皆 `blocked_dry_run_only`。

View File

@@ -320,7 +320,7 @@ YOUTUBE_API_KEY = os.getenv('YOUTUBE_API_KEY', '')
# ==========================================
# 系統版本與路徑
# ==========================================
SYSTEM_VERSION = "V10.228"
SYSTEM_VERSION = "V10.229"
LOG_FILE_PATH = os.path.join(BASE_DIR, 'logs/system.log')
public_url = PUBLIC_URL # 用於模板顯示

View File

@@ -20,6 +20,7 @@ from __future__ import annotations
import os
import time
import base64
import json
import logging
import shutil
import threading
@@ -39,6 +40,7 @@ PPT_VISION_IMAGE_MAX_EDGE = int(os.getenv('PPT_VISION_IMAGE_MAX_EDGE', '1280'))
PPT_VISION_IMAGE_QUALITY = int(os.getenv('PPT_VISION_IMAGE_QUALITY', '82'))
_AUDIT_LOCK = threading.Lock()
_LAST_AUDIT_RUN: Dict[str, Any] | None = None
_ACTIVE_AUDIT_TTL_SECONDS = int(os.getenv('PPT_VISION_ACTIVE_TTL_SECONDS', '7200'))
def is_ppt_vision_enabled() -> bool:
@@ -106,6 +108,80 @@ def get_ppt_vision_runtime_status() -> Dict[str, Any]:
}
def _audit_state_path() -> str:
return os.getenv(
'PPT_VISION_STATE_PATH',
os.path.join(os.getenv('DATA_DIR', os.path.join(os.getcwd(), 'data')), 'ppt_vision_audit_status.json'),
)
def _now_label() -> str:
return time.strftime('%Y-%m-%d %H:%M:%S')
def _read_persisted_audit_run() -> Dict[str, Any] | None:
path = _audit_state_path()
try:
if not os.path.isfile(path):
return None
with open(path, 'r', encoding='utf-8') as handle:
payload = json.load(handle)
return payload if isinstance(payload, dict) else None
except Exception:
logger.debug("[PPTVision] read audit state failed", exc_info=True)
return None
def _write_persisted_audit_run(run: Dict[str, Any]) -> None:
path = _audit_state_path()
directory = os.path.dirname(path)
try:
os.makedirs(directory, exist_ok=True)
tmp_path = f"{path}.tmp"
with open(tmp_path, 'w', encoding='utf-8') as handle:
json.dump(run, handle, ensure_ascii=False)
os.replace(tmp_path, path)
except Exception:
logger.debug("[PPTVision] write audit state failed", exc_info=True)
def _record_audit_run(run: Dict[str, Any]) -> Dict[str, Any]:
global _LAST_AUDIT_RUN
payload = dict(run)
payload['updated_at'] = payload.get('updated_at') or _now_label()
_LAST_AUDIT_RUN = payload
_write_persisted_audit_run(payload)
return payload
def _load_last_audit_run() -> Dict[str, Any] | None:
persisted = _read_persisted_audit_run()
if not _LAST_AUDIT_RUN:
return persisted
if not persisted:
return _LAST_AUDIT_RUN
if str(persisted.get('updated_at') or '') >= str(_LAST_AUDIT_RUN.get('updated_at') or ''):
return persisted
return _LAST_AUDIT_RUN
def _timestamp_age_seconds(value: str | None) -> float | None:
if not value:
return None
try:
parsed = datetime.strptime(value, '%Y-%m-%d %H:%M:%S')
return max(0.0, (datetime.now() - parsed).total_seconds())
except Exception:
return None
def _is_recent_active_audit_run(run: Dict[str, Any] | None) -> bool:
if not run or run.get('status') not in {'queued', 'running'}:
return False
age = _timestamp_age_seconds(run.get('updated_at') or run.get('started_at') or run.get('queued_at'))
return age is None or age < _ACTIVE_AUDIT_TTL_SECONDS
def _public_audit_run_payload(run: Dict[str, Any] | None) -> Dict[str, Any] | None:
if not run:
return None
@@ -126,6 +202,7 @@ def _public_audit_run_payload(run: Dict[str, Any] | None) -> Dict[str, Any] | No
'queued_at': run.get('queued_at') or '',
'started_at': run.get('started_at') or '',
'finished_at': run.get('finished_at') or '',
'updated_at': run.get('updated_at') or '',
'filenames': [
os.path.basename(str(name))
for name in (run.get('filenames') or [])
@@ -146,8 +223,9 @@ def _public_audit_run_payload(run: Dict[str, Any] | None) -> Dict[str, Any] | No
def get_ppt_vision_audit_status() -> Dict[str, Any]:
"""Return the current/last background visual QA run without touching DB."""
running = _AUDIT_LOCK.locked()
last_run = _public_audit_run_payload(_LAST_AUDIT_RUN)
raw_run = _load_last_audit_run()
running = _AUDIT_LOCK.locked() or _is_recent_active_audit_run(raw_run)
last_run = _public_audit_run_payload(raw_run)
if running:
status = 'running'
status_label = '執行中'
@@ -619,14 +697,13 @@ def start_ppt_vision_audit_background(
filenames: Sequence[str] | None = None,
) -> Dict[str, Any]:
"""Queue a non-blocking PPT vision audit run for the admin UI."""
global _LAST_AUDIT_RUN
if _AUDIT_LOCK.locked():
current_run = _load_last_audit_run()
if _AUDIT_LOCK.locked() or _is_recent_active_audit_run(current_run):
return {
'ok': True,
'status': 'already_running',
'message': 'PPT vision audit is already running.',
'last_run': _public_audit_run_payload(_LAST_AUDIT_RUN),
'last_run': _public_audit_run_payload(current_run),
}
clean_filenames = [
@@ -634,27 +711,26 @@ def start_ppt_vision_audit_background(
for name in (filenames or [])
if str(name).lower().endswith('.pptx')
]
queued_at = time.strftime('%Y-%m-%d %H:%M:%S')
_LAST_AUDIT_RUN = {
queued_at = _now_label()
_record_audit_run({
'ok': True,
'status': 'queued',
'queued_at': queued_at,
'filenames': clean_filenames,
'max_files': max_files,
}
})
def _run():
global _LAST_AUDIT_RUN
with _AUDIT_LOCK:
started_at = time.strftime('%Y-%m-%d %H:%M:%S')
_LAST_AUDIT_RUN = {
started_at = _now_label()
_record_audit_run({
'ok': True,
'status': 'running',
'queued_at': queued_at,
'started_at': started_at,
'filenames': clean_filenames,
'max_files': max_files,
}
})
try:
summary = audit_recent_ppts(
reports_dir=reports_dir,
@@ -662,27 +738,27 @@ def start_ppt_vision_audit_background(
max_files=max_files,
filenames=clean_filenames or None,
)
_LAST_AUDIT_RUN = {
_record_audit_run({
'ok': True,
'status': 'completed',
'queued_at': queued_at,
'started_at': started_at,
'finished_at': time.strftime('%Y-%m-%d %H:%M:%S'),
'finished_at': _now_label(),
'filenames': clean_filenames,
'max_files': max_files,
'summary': summary,
}
})
except Exception as exc:
_LAST_AUDIT_RUN = {
_record_audit_run({
'ok': False,
'status': 'error',
'queued_at': queued_at,
'started_at': started_at,
'finished_at': time.strftime('%Y-%m-%d %H:%M:%S'),
'finished_at': _now_label(),
'filenames': clean_filenames,
'max_files': max_files,
'error': f'{type(exc).__name__}: {str(exc)[:200]}',
}
})
logger.error("[PPTVision] background audit failed: %s", exc, exc_info=True)
thread = threading.Thread(target=_run, name='ppt-vision-audit', daemon=True)

View File

@@ -696,10 +696,11 @@ def test_ppt_audit_run_vision_queues_background_audit(client, monkeypatch):
assert captured['max_files'] == 1
def test_ppt_vision_audit_status_sanitizes_last_run(monkeypatch):
def test_ppt_vision_audit_status_sanitizes_last_run(monkeypatch, tmp_path):
"""背景視覺 QA 狀態只回檔名與摘要,不把 reports_dir 絕對路徑曝露到頁面。"""
from services import ppt_vision_service as svc
monkeypatch.setenv('PPT_VISION_STATE_PATH', str(tmp_path / 'vision_state.json'))
monkeypatch.setattr(svc, '_LAST_AUDIT_RUN', {
'ok': True,
'status': 'completed',
@@ -731,6 +732,71 @@ def test_ppt_vision_audit_status_sanitizes_last_run(monkeypatch):
assert '/app/data/reports' not in str(status)
def test_ppt_vision_audit_status_reads_persisted_state(monkeypatch, tmp_path):
"""多 worker 下狀態需從 runtime state 檔讀取,不能只靠單一 worker 記憶體。"""
import json
from services import ppt_vision_service as svc
state_path = tmp_path / 'vision_state.json'
monkeypatch.setenv('PPT_VISION_STATE_PATH', str(state_path))
monkeypatch.setattr(svc, '_LAST_AUDIT_RUN', None)
state_path.write_text(json.dumps({
'ok': True,
'status': 'completed',
'queued_at': '2026-05-19 12:00:00',
'started_at': '2026-05-19 12:00:01',
'finished_at': '2026-05-19 12:00:05',
'updated_at': '2026-05-19 12:00:05',
'filenames': ['/app/data/reports/ocbot_daily_20260518.pptx'],
'max_files': 1,
'summary': {
'audited_files': [{
'path': '/app/data/reports/ocbot_daily_20260518.pptx',
'slides_checked': 1,
'issues': 0,
'error': None,
}],
'total_issues': 0,
'errors': [],
},
}), encoding='utf-8')
status = svc.get_ppt_vision_audit_status()
assert status['status'] == 'completed'
assert status['last_run']['updated_at'] == '2026-05-19 12:00:05'
assert status['last_run']['filenames'] == ['ocbot_daily_20260518.pptx']
assert '/app/data/reports' not in str(status)
def test_ppt_vision_audit_background_respects_persisted_running(monkeypatch, tmp_path):
"""另一個 worker 已在跑時,新 request 應回 already_running避免重複開模型任務。"""
import json
from services import ppt_vision_service as svc
state_path = tmp_path / 'vision_state.json'
monkeypatch.setenv('PPT_VISION_STATE_PATH', str(state_path))
monkeypatch.setattr(svc, '_LAST_AUDIT_RUN', None)
state_path.write_text(json.dumps({
'ok': True,
'status': 'running',
'queued_at': '2999-05-19 12:00:00',
'started_at': '2999-05-19 12:00:01',
'updated_at': '2999-05-19 12:00:01',
'filenames': ['/app/data/reports/ocbot_daily_20260518.pptx'],
'max_files': 1,
}), encoding='utf-8')
result = svc.start_ppt_vision_audit_background(
filenames=['ocbot_daily_20260518.pptx'],
max_files=1,
)
assert result['ok'] is True
assert result['status'] == 'already_running'
assert result['last_run']['filenames'] == ['ocbot_daily_20260518.pptx']
def test_ppt_audit_vision_status_route_returns_json(client, monkeypatch):
"""頁面輪詢用 status endpoint 要能回最近一次背景視覺 QA 狀態。"""
from services import ppt_vision_service as svc