daily-paper/app/services/crawler.py

"""爬虫服务 — 从 HuggingFace Daily Papers API 抓取论文元数据。"""

import logging
from datetime import date as date_type, datetime, timezone

import httpx
from sqlalchemy import select, text
from sqlalchemy.orm import Session

from app.config import settings
from app.models import (
    CrawlLog,
    Paper,
    PaperAuthor,
    PaperTag,
    SummaryState,
    SummaryStatus,
)
from app.utils import make_http_client, utc_now

logger = logging.getLogger(__name__)


async def fetch_daily(target_date: str, top_n: int | None = None) -> list[dict]:
    """从 HF Daily Papers API 获取指定日期的论文列表。

    Args:
        target_date: YYYY-MM-DD 格式
        top_n: 取前 N 篇，默认使用 settings.TOP_N

    Returns:
        论文元数据列表
    """
    top_n = top_n or settings.TOP_N
    url = f"{settings.HF_API_BASE}/daily_papers"
    params = {"date": target_date}

    async with make_http_client() as client:
        for attempt in range(1, settings.HTTP_MAX_RETRIES + 1):
            try:
                logger.info(
                    "Fetching HF Daily Papers: date=%s attempt=%d", target_date, attempt
                )
                resp = await client.get(url, params=params)
                resp.raise_for_status()
                data = resp.json()
                break
            except (httpx.HTTPError, httpx.HTTPStatusError) as exc:
                logger.warning(
                    "Fetch failed (attempt %d/%d): %s",
                    attempt,
                    settings.HTTP_MAX_RETRIES,
                    exc,
                )
                if attempt == settings.HTTP_MAX_RETRIES:
                    raise
        else:
            data = []

    papers = data[:top_n]
    logger.info(
        "Fetched %d papers for %s (raw=%d)", len(papers), target_date, len(data)
    )
    return papers


def _parse_paper(item: dict) -> dict:
    """从 HF API 响应中提取论文元数据。"""
    paper_info = item.get("paper", item)
    arxiv_id = paper_info.get("id", "")
    published_raw = paper_info.get("publishedAt", "")
    published_at = None
    if published_raw:
        try:
            published_at = date_type.fromisoformat(published_raw[:10])
        except ValueError:
            pass
    return {
        "arxiv_id": arxiv_id,
        "title_en": paper_info.get("title", ""),
        "abstract": paper_info.get("abstract", ""),
        "published_at": published_at,
        "upvotes": item.get("paper", {}).get("upvotes", 0) or item.get("upvotes", 0),
        "hf_url": f"https://huggingface.co/papers/{arxiv_id}" if arxiv_id else "",
        "arxiv_url": f"https://arxiv.org/abs/{arxiv_id}" if arxiv_id else "",
        "pdf_url": f"https://arxiv.org/pdf/{arxiv_id}.pdf" if arxiv_id else "",
        "authors": [
            a.get("name", a) if isinstance(a, dict) else a
            for a in paper_info.get("authors", [])
        ],
        "tags": [
            t.get("name", t) if isinstance(t, dict) else t
            for t in (paper_info.get("tags") or [])
        ],
    }


def upsert_papers(db: Session, papers_raw: list[dict], paper_date: str) -> list[Paper]:
    """将论文元数据写入数据库。已有论文仅更新可变字段（upvotes 等），不重复插入。"""
    now = datetime.now(timezone.utc)
    paper_date_obj = date_type.fromisoformat(paper_date)
    new_papers: list[Paper] = []

    for item in papers_raw:
        meta = _parse_paper(item)
        arxiv_id = meta["arxiv_id"]
        if not arxiv_id:
            continue

        existing = db.execute(
            select(Paper).where(Paper.arxiv_id == arxiv_id)
        ).scalar_one_or_none()

        if existing:
            existing.upvotes = meta["upvotes"]
            existing.crawled_at = now
            logger.debug("Updated existing paper: %s", arxiv_id)
        else:
            paper = Paper(
                arxiv_id=arxiv_id,
                title_en=meta["title_en"],
                abstract=meta["abstract"],
                published_at=meta["published_at"],
                paper_date=paper_date_obj,
                crawled_at=now,
                upvotes=meta["upvotes"],
                hf_url=meta["hf_url"],
                arxiv_url=meta["arxiv_url"],
                pdf_url=meta["pdf_url"],
            )
            db.add(paper)
            db.flush()

            seen_authors: set[str] = set()
            for idx, name in enumerate(meta["authors"]):
                if name and name not in seen_authors:
                    seen_authors.add(name)
                    db.add(PaperAuthor(paper_id=paper.id, name=name, position=idx))

            for tag_name in meta["tags"]:
                if tag_name:
                    db.add(PaperTag(paper_id=paper.id, tag=tag_name, source="hf"))

            db.add(SummaryStatus(paper_id=paper.id, status=SummaryState.PENDING))

            authors_text = ", ".join(meta["authors"])
            tags_text = ", ".join(meta["tags"])
            db.execute(
                text(
                    "INSERT INTO papers_fts(rowid, title_en, abstract, authors, tags) "
                    "VALUES (:id, :title, :abstract, :authors, :tags)"
                ),
                {
                    "id": paper.id,
                    "title": meta["title_en"],
                    "abstract": meta["abstract"] or "",
                    "authors": authors_text,
                    "tags": tags_text,
                },
            )

            new_papers.append(paper)
            logger.debug("Inserted new paper: %s", arxiv_id)

    db.commit()
    logger.info(
        "Upserted %d papers (%d new) for %s",
        len(papers_raw),
        len(new_papers),
        paper_date,
    )
    return new_papers


async def crawl_daily(db: Session, target_date: str, top_n: int | None = None) -> dict:
    """完整的抓取流程：获取 + 入库 + 写日志。"""
    now = utc_now()
    log_entry = CrawlLog(
        task="crawl",
        status="running",
        date=date_type.fromisoformat(target_date),
        started_at=now,
    )
    db.add(log_entry)
    db.commit()

    try:
        raw_papers = await fetch_daily(target_date, top_n)
        new_papers = upsert_papers(db, raw_papers, target_date)
        log_entry.status = "success"
        log_entry.papers_found = len(raw_papers)
        log_entry.papers_new = len(new_papers)
        log_entry.completed_at = utc_now()
        db.commit()
        return {
            "found": len(raw_papers),
            "new": len(new_papers),
            "status": "success",
            "error": None,
        }
    except Exception as exc:
        logger.exception("Crawl failed for %s", target_date)
        log_entry.status = "failed"
        log_entry.error = str(exc)
        log_entry.completed_at = utc_now()
        db.commit()
        return {"found": 0, "new": 0, "status": "failed", "error": str(exc)}