Files
stock_chart_site/backend/app/pipelines/refresh_one.py
tkrmagid edda01adbf feat(phase-2): KR-FinBERT 감성 스코어링 + 일별 집계 뷰
- backend/app/nlp/finbert.py: snunlp/KR-FinBert-SC 어댑터.
  - score = P(pos) - P(neg) ∈ [-1, +1], label = argmax (neg/neu/pos)
  - 768d mean-pooled last hidden state → news.embedding (VECTOR) 저장
  - settings.huggingface_token 인증, lazy singleton, cuda/cpu auto
- backend/app/nlp/score_news.py: news 테이블에서 sentiment_score IS NULL
  행을 배치 스코어 → UPDATE (... embedding=(:e)::vector). 종목 필터 + limit 옵션.
- backend/app/db/migrations/002_sentiment_view.sql: v_sentiment_daily 뷰.
  종목·KST 일별 n_articles, mean_score, pos/neg/neu_ratio, weighted_score
  (naver_finance 1.0 / google_rss 0.7 / dart 0.5).
- backend/app/db/migrate.py: 이미 실행 중인 DB 에 새 SQL 마이그레이션 적용용
  CLI. 모든 SQL 파일은 idempotent.
- refresh_one.py: refresh 끝에 종목당 200건까지 finbert 스코어, finbert
  SourceStatus 를 RefreshReport 에 추가.
- daily_batch.py: 모든 종목 처리 후 score_pending_news(limit=2000) 로 mop-up.

모델 캐시는 docker-compose hf_cache 볼륨(/root/.cache/huggingface).

Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
2026-05-20 15:57:34 +09:00

170 lines
5.7 KiB
Python

"""한 종목에 대해 모든 소스를 갱신 + 구조화된 status 리턴.
POST /api/refresh/{code} 와 daily_batch 둘 다 이 함수를 호출.
"""
from __future__ import annotations
import logging
from dataclasses import asdict, dataclass, field
from datetime import date, timedelta
from typing import Any
from app.fetch import dart as dart_mod
from app.fetch import kis as kis_mod
from app.fetch import news as news_mod
from app.fetch import pykrx_helper
logger = logging.getLogger(__name__)
@dataclass
class SourceStatus:
status: str # 'ok' / 'skipped_missing_key' / 'failed'
inserted: int = 0
updated: int = 0
skipped: int = 0
extra: dict[str, Any] = field(default_factory=dict)
error: str | None = None
@dataclass
class RefreshReport:
code: str
pykrx_ohlcv: SourceStatus
pykrx_trading_value: SourceStatus
kis_daily: SourceStatus
dart: SourceStatus
naver_news: SourceStatus
google_rss: SourceStatus
finbert: SourceStatus
def to_dict(self) -> dict[str, Any]:
out: dict[str, Any] = {"code": self.code}
for f in (
"pykrx_ohlcv",
"pykrx_trading_value",
"kis_daily",
"dart",
"naver_news",
"google_rss",
"finbert",
):
v: SourceStatus = getattr(self, f)
out[f] = asdict(v)
return out
def _pykrx_ohlcv(code: str, start: date, end: date) -> SourceStatus:
try:
res = pykrx_helper.fetch_ohlcv_daily(code, start, end)
return SourceStatus(
status=res.status(),
inserted=res.inserted,
updated=res.updated,
error=res.error,
)
except Exception as exc: # noqa: BLE001
return SourceStatus(status="failed", error=str(exc))
def _pykrx_trading(code: str, start: date, end: date) -> SourceStatus:
try:
res = pykrx_helper.fetch_trading_value(code, start, end)
return SourceStatus(
status=res.status(),
inserted=res.inserted,
updated=res.updated,
error=res.error,
)
except Exception as exc: # noqa: BLE001
return SourceStatus(status="failed", error=str(exc))
def _kis(code: str, start: date, end: date) -> SourceStatus:
"""KIS read-only EOD. 실제 DB 적재는 하지 않고 sanity 호출 + sample row 수만 리포트.
pykrx 와 중복 데이터이므로 KIS 는 백업/실시간 용도이고, 일별 적재는 pykrx 가 1차.
"""
try:
rows = kis_mod.fetch_daily_price(code, start, end)
return SourceStatus(status="ok", extra={"sample_rows": len(rows)})
except kis_mod.SkippedMissingKey:
return SourceStatus(status="skipped_missing_key")
except Exception as exc: # noqa: BLE001
return SourceStatus(status="failed", error=str(exc))
def _dart(code: str, start: date, end: date) -> SourceStatus:
try:
items = dart_mod.fetch_disclosures(code, start, end)
# 공시는 news 테이블에 upsert
news_items = [
news_mod.NewsItem(
code=d.code,
source="dart",
title=d.title,
url=d.url,
published_at=d.published_at,
)
for d in items
]
ins, skip = news_mod.upsert_news(news_items)
return SourceStatus(status="ok", inserted=ins, skipped=skip, extra={"fetched": len(items)})
except dart_mod.SkippedMissingKey:
return SourceStatus(status="skipped_missing_key")
except Exception as exc: # noqa: BLE001
return SourceStatus(status="failed", error=str(exc))
def _naver_news(code: str) -> SourceStatus:
try:
items = news_mod.fetch_naver_finance_news(code, max_pages=1)
ins, skip = news_mod.upsert_news(items)
return SourceStatus(status="ok", inserted=ins, skipped=skip, extra={"fetched": len(items)})
except Exception as exc: # noqa: BLE001
return SourceStatus(status="failed", error=str(exc))
def _google_rss(code: str, name: str) -> SourceStatus:
try:
query = name or code
items = news_mod.fetch_google_news_rss(query, code=code)
ins, skip = news_mod.upsert_news(items)
return SourceStatus(status="ok", inserted=ins, skipped=skip, extra={"fetched": len(items)})
except Exception as exc: # noqa: BLE001
return SourceStatus(status="failed", error=str(exc))
def _finbert(code: str) -> SourceStatus:
"""방금 upsert 된 뉴스 중 sentiment_score 가 비어있는 행을 KR-FinBERT 로 스코어."""
try:
from app.nlp.score_news import score_pending_news
# 한 종목에 대해 신규 뉴스가 매우 많아도 200건으로 컷.
# daily_batch 끝에서 잔여분을 별도로 mop-up 한다.
res = score_pending_news(code=code, limit=200)
return SourceStatus(
status="ok" if res.error is None else "failed",
inserted=res.scored,
skipped=res.failed,
extra={"fetched": res.fetched},
error=res.error,
)
except Exception as exc: # noqa: BLE001
return SourceStatus(status="failed", error=str(exc))
def refresh_code(code: str, name: str, *, lookback_days: int = 7) -> RefreshReport:
"""단기 갱신 (daily_batch 용). 최근 lookback_days 만 가져온다."""
end = date.today()
start = end - timedelta(days=lookback_days)
return RefreshReport(
code=code,
pykrx_ohlcv=_pykrx_ohlcv(code, start, end),
pykrx_trading_value=_pykrx_trading(code, start, end),
kis_daily=_kis(code, start, end),
dart=_dart(code, start, end),
naver_news=_naver_news(code),
google_rss=_google_rss(code, name),
finbert=_finbert(code),
)