eezd
4929a32a0a
Extract reusable literal keywords once per source revision, assign complete news packages to long-lived events, and research all event members with immutable evidence and references. Add additive research storage, vocabulary snapshots, the two-stage workspace, isolated acceptance coverage, and a paused rollout with an initial fixed 50-news production trial.
65 lines
3.9 KiB
Python
65 lines
3.9 KiB
Python
"""Export the accepted twenty-package baseline through a read-only DB transaction.
|
|
|
|
Run inside the API environment; stdout contains news and AI results, never provider
|
|
credentials. This exports inputs only and cannot start workers or call models.
|
|
"""
|
|
import asyncio
|
|
import json
|
|
from datetime import UTC, datetime
|
|
from sqlalchemy import select, text
|
|
from alphapulse.core.db import SessionLocal, engine
|
|
from alphapulse.models import NewsItem
|
|
from alphapulse.models.risk import ArticleAnalysis, NewsSummary, SummaryArticle, SummaryVersion, PackageEvidence, RiskSettings
|
|
|
|
SAMPLE_IDS = [126, 140, 763, 774, 781, 769, 765, 94, 96, 97, 776, 623, 7, 2, 227, 88, 768, 770, 762, 761]
|
|
|
|
|
|
async def export():
|
|
start = datetime.fromisoformat('2026-11-01T00:00:00+08:00')
|
|
end = datetime.fromisoformat('2026-11-03T00:00:00+08:00')
|
|
packages, news = [], {}
|
|
try:
|
|
async with SessionLocal() as session:
|
|
await session.execute(text('SET TRANSACTION ISOLATION LEVEL REPEATABLE READ, READ ONLY'))
|
|
settings = await session.get(RiskSettings, 1)
|
|
if settings.enabled:
|
|
raise ValueError('请先暂停线上AI再导出固定样本')
|
|
for identifier in SAMPLE_IDS:
|
|
package = await session.get(NewsSummary, identifier)
|
|
if package is None or package.replacement_ids or package.state != 'done':
|
|
raise ValueError(f'样本包#{identifier}已变化，请重新审核样本')
|
|
version = await session.scalar(select(SummaryVersion).where(SummaryVersion.summary_id == identifier,
|
|
SummaryVersion.revision == package.published_revision, SummaryVersion.state == 'done'))
|
|
if version is None:
|
|
raise ValueError(f'样本包#{identifier}缺少成功版本')
|
|
rows = await session.scalars(select(NewsItem).join(ArticleAnalysis, ArticleAnalysis.news_id == NewsItem.id)
|
|
.join(SummaryArticle, SummaryArticle.analysis_id == ArticleAnalysis.id).where(
|
|
SummaryArticle.summary_id == identifier, SummaryArticle.active.is_(True),
|
|
ArticleAnalysis.content_hash == NewsItem.content_hash))
|
|
source_ids = []
|
|
for row in rows:
|
|
if not start <= row.published_at < end:
|
|
raise ValueError(f'样本新闻#{row.id}超出时间范围')
|
|
source_ids.append(row.id)
|
|
news[row.id] = {key: getattr(row, key) for key in ('id', 'source_code', 'url', 'title', 'summary',
|
|
'content_text', 'content_hash', 'revision_no', 'published_at', 'source_updated_at')}
|
|
if not source_ids:
|
|
raise ValueError(f'样本包#{identifier}无有效原文')
|
|
records = list(await session.scalars(select(PackageEvidence).where(PackageEvidence.version_id == version.id)
|
|
.order_by(PackageEvidence.id)))
|
|
packages.append({'id': identifier, 'title': package.title, 'revision': version.revision,
|
|
'news_ids': sorted(set(source_ids)), 'input': version.input, 'result': version.result,
|
|
'rules': version.rules, 'model': version.model, 'usage': version.usage,
|
|
'started_at': version.started_at, 'finished_at': version.finished_at,
|
|
'research': [r.data for r in records]})
|
|
return {'exported_at': datetime.now(UTC), 'start': start, 'end': end,
|
|
'model_settings': {key: getattr(settings, key) for key in ('model', 'second_model', 'third_model',
|
|
'reasoning_effort', 'second_reasoning_effort', 'third_reasoning_effort', 'max_concurrency')},
|
|
'packages': packages, 'news': sorted(news.values(), key=lambda n: (n['published_at'], n['id']))}
|
|
finally:
|
|
await engine.dispose()
|
|
|
|
|
|
if __name__ == '__main__':
|
|
print(json.dumps(asyncio.run(export()), ensure_ascii=False, default=str))
|