o2o-negosium-original/lps/worker/handlers.py

"""잡 핸들러 — job_type 별 처리. 현재는 SEARCH(검색)만.

검색 핸들러 = 한정된 재정제 루프 + 명시적 outcome:
  0. 네거티브 캐시 확인(최근 not_found면 즉시 반환)
  각 라운드(원본 → 정밀(LLM) → 광역, 최대 max_rounds):
     소스 동시 검색 → 필터 → 이상치 → AI 같은상품 판정
     ├ 매칭 있음 → DONE(outcome=found), 조기 종료
     ├ 0매칭 + 기술적 실패(차단/예외) 있음 → raise → 큐가 잡 전체 백오프 재시도(→소진 시 DEAD)
     └ 0매칭 + 소스 정상 → 다음 라운드
  라운드 소진 → DONE(outcome=not_found) + 네거티브 캐시 기록

두 재시도 축을 분리한다: 기술적(큐 attempts/백오프) ≠ 검색어(refine 라운드, 유한).
'못 찾음'은 정상 종료(DONE)지 dead-letter 가 아니다.
"""

import asyncio
import time

from common.enums import JobType
from common.logger import LOG
from services.metrics import SearchMetrics
from services.search.contract import SearchAdapter, NormalizedProduct
from services.search.card_parser import canonical_mall, MALL_BY_SOURCE
from services.search.util import parse_price
from services.pipeline.core import apply_filters, rank_result, summarize_by_mall


def _price_snapshot(matched: list[NormalizedProduct]) -> dict:
    """매칭 목록에서 소스별 최저가 + 전체 최저가 스냅샷을 만든다(price_history 기록용)."""
    def lowest(src):
        items = [p for p in matched if p.source == src]
        return min(items, key=lambda p: p.price) if items else None

    n, c = lowest("naver"), lowest("coupang")
    # 최종 최저가는 소스 무관 전체 매칭 중 최저(G마켓·옥션·11번가 등 폴백 포함).
    f = min(matched, key=lambda p: p.price) if matched else None
    return {
        "matched_count": len(matched),
        "naver_lowest": n.price if n else None, "naver_name": n.name if n else None, "naver_url": n.detail_url if n else None,
        "coupang_lowest": c.price if c else None, "coupang_name": c.name if c else None, "coupang_url": c.detail_url if c else None,
        "final_lowest": f.price if f else None, "final_source": f.source if f else None,
        "by_mall": summarize_by_mall(matched),  # 몰별 최저가 스냅샷(열린 스키마)
    }


def build_search_handler(
    adapters: dict[str, SearchAdapter],
    sources: list[str] | None = None,
    limit: int = 40,
    top_n: int = 5,
    judge=None,
    keyword_gen=None,
    max_rounds: int = 3,
    neg_cache=None,
    history=None,
    fallback_adapters: dict[str, SearchAdapter] | None = None,
    ai_model: str = "",
):
    """검색 핸들러 생성.
    judge: SimilarityJudge(같은 상품 판정) / keyword_gen: KeywordGenerator(정밀·광역 재검색어) /
    neg_cache: NegativeCache(TTL not_found 캐시) / history: PriceHistory(최저가 스냅샷) /
    fallback_adapters: 오픈마켓 크롤(gmarket/auction/st11) — 네이버가 그 몰을 커버 못 했을 때만 크롤(폴백).
    모두 선택 — 없으면 해당 단계 생략."""
    use = list(sources) if sources else list(adapters.keys())
    fallbacks = fallback_adapters or {}

    async def _record_history(product_code: str, job_id, outcome: str, matched: list):
        if history is None:
            return
        event = {"product_code": product_code, "job_id": job_id, "outcome": outcome, **_price_snapshot(matched)}
        try:
            await history.record(event)
        except Exception as ex:
            LOG.e_no_callstack(f"[history] 스냅샷 기록 실패(무시): {ex}")

    async def _timed_search(adapter, query: str, source: str, metrics: SearchMetrics, crawl: bool):
        """어댑터 검색 1건을 타이밍+바이트 계측하며 실행. 예외는 그대로 전파(호출부가 처리)."""
        t0 = time.monotonic()
        try:
            res = await adapter.search(query, limit=limit)
            metrics.add_fetch(source, getattr(adapter, "last_bytes", 0), int((time.monotonic() - t0) * 1000), crawl=crawl)
            return res
        except Exception:
            metrics.add_fetch(source, 0, int((time.monotonic() - t0) * 1000), crawl=crawl)
            raise

    async def _search_round(query: str, metrics: SearchMetrics):
        """한 라운드: 모든 소스 동시 검색 → (products, per_source, tech_failed). 소스별 시간/바이트 계측."""
        results = await asyncio.gather(*[_timed_search(adapters[s], query, s, metrics, False) for s in use],
                                       return_exceptions=True)
        products, per_source, tech_failed = [], {}, False
        for src, res in zip(use, results):
            if isinstance(res, Exception):
                tech_failed = True
                per_source[src] = {"error": f"{type(res).__name__}: {res}"}
                LOG.w(f"[{src}] 검색 실패: {type(res).__name__}: {res}")
            else:
                products.extend(res)
                per_source[src] = {"count": len(res)}
        return products, per_source, tech_failed

    async def _match(target: dict, products: list, base_price, metrics: SearchMetrics):
        """필터 → (있으면) AI 같은상품 판정 → 매칭 후보. AI 토큰은 metrics 에 누적."""
        candidates, _ = apply_filters(products, base_price=base_price)
        if judge is not None and candidates:
            verdicts = await judge.judge(target, candidates)
            metrics.add_ai(judge.last_usage)
            candidates = [c for c, v in zip(candidates, verdicts) if v.is_match]
        return candidates

    async def _enrich_with_fallback(target: dict, query: str, matched: list, base_price, metrics: SearchMetrics):
        """네이버가 커버 못 한 오픈마켓만 직접 크롤(폴백) → 같은상품 판정 후 병합.
        사용자 규칙: '네이버로 그 몰 값 확보 성공 → 그 값, 실패(몰 없음) → 실사이트 크롤'."""
        if not fallbacks:
            return matched
        covered = {canonical_mall(p) for p in matched}
        for src, adapter in fallbacks.items():
            mall = MALL_BY_SOURCE.get(src, src)
            if mall in covered:  # 네이버가 이미 그 몰 최저가 확보 → 크롤 생략
                continue
            try:
                crawled = await _timed_search(adapter, query, src, metrics, crawl=True)
            except Exception as ex:
                LOG.w(f"[fallback:{src}] 크롤 실패(무시): {type(ex).__name__}: {ex}")
                continue
            hits = await _match(target, crawled, base_price, metrics)
            if hits:
                LOG.d(f"[fallback:{src}] 크롤 {len(crawled)}건 중 같은상품 {len(hits)}건 병합")
                matched = matched + hits
        return matched

    async def _round_queries(base_query: str, target: dict, metrics: SearchMetrics):
        """라운드 쿼리 지연 생성: 원본 → (0매칭 시에만 LLM 호출로) 정밀 → 광역."""
        yield ("original", base_query)
        if keyword_gen is not None:
            kw = await keyword_gen.generate(target)  # 원본이 실패해 여기까지 온 경우에만 호출됨
            metrics.add_ai(keyword_gen.last_usage)
            seen = {base_query}
            for label, q in (("precise", kw.precise), ("broad", kw.broad)):
                q = (q or "").strip()
                if q and q not in seen:
                    seen.add(q)
                    yield (label, q)

    async def handler(job: dict) -> dict:
        if job["job_type"] != JobType.SEARCH.value:
            raise ValueError(f"unsupported job_type: {job['job_type']}")

        payload = job.get("payload") or {}
        base_query = (payload.get("product_name") or "").strip()
        if not base_query:
            raise ValueError("empty product_name")
        target = {k: payload.get(k, "") for k in ("product_name", "model", "specification", "company")}
        base_price = parse_price(payload.get("price"))
        cache_key = payload.get("product_code") or base_query
        metrics = SearchMetrics(ai_model)  # 검색 1건의 리소스/비용/시간 계측

        # 0) 네거티브 캐시 — 최근 not_found면 재검색 생략
        if neg_cache is not None and await neg_cache.is_negative(cache_key):
            return {"outcome": "not_found", "cached": True, "query": base_query,
                    "rounds_tried": 0, "lowest": None, "top": [], "stages": [], "sources": {},
                    "metrics": metrics.snapshot()}

        rounds_done = 0
        last_stages, last_sources = [], {}
        async for label, query in _round_queries(base_query, target, metrics):
            if rounds_done >= max_rounds:
                break
            rounds_done += 1

            products, per_source, tech_failed = await _search_round(query, metrics)
            candidates, stages = apply_filters(products, base_price=base_price)
            if judge is not None and candidates:
                verdicts = await judge.judge(target, candidates)
                metrics.add_ai(judge.last_usage)
                matched = [c for c, v in zip(candidates, verdicts) if v.is_match]
                stages.append({"stage": "ai_match", "in": len(candidates), "out": len(matched)})
                candidates = matched
            last_stages, last_sources = stages, per_source

            if candidates:  # 찾음 → 오픈마켓 폴백 보강 후 종료
                before = len(candidates)
                candidates = await _enrich_with_fallback(target, query, candidates, base_price, metrics)
                if len(candidates) > before:
                    stages.append({"stage": "fallback_crawl", "in": before, "out": len(candidates)})
                result = rank_result(candidates, len(products), stages, top_n)
                result.update(outcome="found", query=query, round=label, rounds_tried=rounds_done,
                              sources=per_source, metrics=metrics.snapshot())
                await _record_history(cache_key, job.get("job_id"), "found", candidates)
                return result

            if tech_failed:  # 0매칭인데 소스가 죽어 있었음 → '없음'이라 단정 불가 → 기술 재시도
                raise RuntimeError(f"기술적 실패로 0매칭(round={label}) — 잡 재시도: {per_source}")

        # 모든 라운드 클린 0매칭 → 정상 not_found 종료
        if neg_cache is not None:
            await neg_cache.put(cache_key, reason=f"not_found after {rounds_done} rounds")
        result = rank_result([], 0, last_stages, top_n)
        result.update(outcome="not_found", query=base_query, rounds_tried=rounds_done,
                      sources=last_sources, metrics=metrics.snapshot())
        await _record_history(cache_key, job.get("job_id"), "not_found", [])
        return result

    return handler