Skip to content

armgit / peach

Public Imported from GitHub
0

Peach: Local-first video metadata scraper and media manager.

8.6 KB
Raw Blame
#!/usr/bin/env python3
# -*- coding: utf-8 -*-
r"""
批处理进度同步 —— 从账本和产物反推当前状态,写进 peach-data/state/job-status.md。

数字从当前数据计算;产物落在仓库外的 `peach-data/state/`,不进入 Git。

用法:
    python scripts/job_status.py            # 打印,不改文件
    python scripts/job_status.py --write    # 写回 peach-data/state/job-status.md
"""
from __future__ import annotations

import argparse
import json
import os
import sqlite3
import sys
from datetime import datetime, timezone
from pathlib import Path

PROJECT_ROOT = Path(__file__).resolve().parent.parent
SRC_DIR = PROJECT_ROOT / "src"
if str(SRC_DIR) not in sys.path:
    sys.path.insert(0, str(SRC_DIR))

from peach.config import DATABASE_PATH, GENERATED_DIR, STATE_DIR
from peach.media_sources import SCANNABLE_LOCATIONS
from peach.review_csv import read_rows

# 路径由配置解析,不依赖开发机器的盘符与目录布局。
DOC = STATE_DIR / "job-status.md"
STATE = STATE_DIR / "agent-handoff.json"
START = "<!-- job-status:start -->"
END = "<!-- job-status:end -->"

# 产物 -> 说明。清单类产物必须能一眼看出"有没有过期",所以连带报出生成时间。
ARTIFACTS = {
    "code-scrape.csv": "番号刮削结果",
    "name-clean.csv": "文件名净化清单",
    "ad-candidates.csv": "广告候选(confidence 分级)",
    "disposal-candidates.csv": "待处置候选(含「保留」判定)",
    "creator-tags-review.csv": "创作者标签待审",
}


def _readonly(path: Path) -> sqlite3.Connection:
    return sqlite3.connect("file:" + path.resolve().as_posix() + "?mode=ro", uri=True)


def _local(text: str) -> str:
    """将 UTC 时间戳按本机时区显示;无法解析时保留原值。"""
    try:
        moment = datetime.fromisoformat(text.replace("Z", "+00:00"))
    except (AttributeError, TypeError, ValueError):
        return text
    if moment.tzinfo is None:
        moment = moment.replace(tzinfo=timezone.utc)
    return moment.astimezone().isoformat(timespec="seconds")


def collect(db_path: Path, generated: Path, state_path: Path | None = None) -> list[str]:
    conn = _readonly(db_path)
    one = lambda sql, *a: conn.execute(sql, a).fetchone()[0]

    lines: list[str] = [START, "", "<!-- 由 scripts/job_status.py 生成,勿手改;数字现算于账本与产物 -->",
                        f"<!-- generated {datetime.now().astimezone().isoformat(timespec='minutes')} -->", ""]

    if state_path and state_path.is_file():
        try:
            event = json.loads(state_path.read_text(encoding="utf-8"))
            lines.append(
                f"- 最近自动交接:`{event.get('agent', 'unknown')}` / "
                f"`{event.get('event', 'unknown')}` / `{event.get('outcome', 'unknown')}`,"
                f"{_local(str(event.get('at', 'unknown')))}。"
            )
        except (OSError, ValueError, TypeError):
            lines.append("- 最近自动交接:状态文件不可读;本次账本数字仍已重新计算。")

    videos = one("SELECT count(*) FROM asset WHERE medium='video'")
    lines.append(f"- 资产 {one('SELECT count(*) FROM asset')} 条,其中视频 {videos} 条。")

    # 待抽帧必须报两个数。只报一个必然引起口径之争:`4740` 和 `10196` 看着互相
    # 矛盾,其实一个是可执行量、一个是原始量,都对。
    # 没有时长就算不出 seek 点,抽帧脚本处理不了,那些要先补 probe。
    lines.append("- 待抽帧(可抽 / 缺时长待 probe / 合计):")
    for loc in SCANNABLE_LOCATIONS:
        ready = one("SELECT count(*) FROM asset WHERE medium='video' AND location=? "
                    "AND (snapshot_path IS NULL OR snapshot_path='') AND duration>0", loc)
        blocked = one("SELECT count(*) FROM asset WHERE medium='video' AND location=? "
                      "AND (snapshot_path IS NULL OR snapshot_path='') "
                      "AND (duration IS NULL OR duration<=0)", loc)
        lines.append(f"  - `{loc}`:{ready} / {blocked} / {ready + blocked}")
    # 抽帧的瓶颈是流量不是时间;下面是带日期与条件的实测参考,不是当前状态。
    lines.append("  抽帧流量参考(2026-08-15 实测):PikPak 走代理 9 帧 163 MB / 13.7 秒,"
                 "直连 30.5 MB / 64.2 秒——慢约 4.7 倍但流量少约 5 倍且不占代理预算;"
                 "全量抽帧约 773 GB(代理口径),按创作者采样 88 板直连约 2.7 GB;"
                 "115 直连一张九宫格约 285 MB。")

    untagged = one("SELECT count(*) FROM asset WHERE medium='video' "
                   "AND id NOT IN (SELECT asset_id FROM asset_tag)")
    lines.append(f"- 无内容标签视频 {untagged} 条(占视频 {untagged/max(videos,1)*100:.0f}%)。")

    sources = conn.execute(
        "SELECT source,count(*) FROM asset_tag GROUP BY source ORDER BY 2 DESC").fetchall()
    lines.append("- `asset_tag` 来源分布:" +
                 "、".join(f"`{s}` {n}" for s, n in sources) + "。")

    codes = one("SELECT count(DISTINCT code) FROM asset WHERE code>''")
    with_studio = one("SELECT count(DISTINCT code) FROM asset WHERE code>'' AND studio>''")
    lines.append(f"- 番号 {codes} 个,其中 {with_studio} 个有厂牌"
                 f"({with_studio/max(codes,1)*100:.0f}%)。")
    conn.close()

    lines.append("")
    lines.append("| 产物 | 行数 | 生成时间(本地) | 说明 |")
    lines.append("| --- | ---: | --- | --- |")
    for name, desc in ARTIFACTS.items():
        path = generated / name
        if not path.is_file():
            lines.append(f"| `{name}` | — | 未生成 | {desc} |")
            continue
        rows = len(read_rows(path))
        # mtime 本来就是本地时间,显式 astimezone 是为了和上面两处同一个时钟,
        # 列宽放不下偏移量,所以时区写在表头里。
        stamp = datetime.fromtimestamp(path.stat().st_mtime).astimezone().strftime("%m-%d %H:%M")
        lines.append(f"| `{name}` | {rows} | {stamp} | {desc} |")

    lines += ["", END]
    return lines


def _atomic_write(path: Path, text: str) -> None:
    path.parent.mkdir(parents=True, exist_ok=True)
    temp = path.with_name(f".{path.name}.{os.getpid()}.tmp")
    temp.write_text(text, encoding="utf-8", newline="\n")
    os.replace(temp, path)


def write_back(doc: Path, block: list[str]) -> str:
    text = doc.read_text(encoding="utf-8") if doc.is_file() else ""
    body = "\n".join(block)
    if START in text and END in text:
        head, _, rest = text.partition(START)
        _, _, tail = rest.partition(END)
        _atomic_write(doc, head + body + tail)
        return "替换了已有区块"
    head = text.rstrip() + "\n\n" if text.strip() else "# 批处理进度(自动生成)\n\n"
    _atomic_write(doc, head + body + "\n")
    return "新建了区块"


def record_hook_event(state_path: Path, payload: dict[str, object]) -> None:
    event = str(payload.get("hook_event_name") or payload.get("event") or "unknown")
    outcome = "failed" if event == "StopFailure" else (
        str(payload.get("reason") or "ended")[:80] if event == "SessionEnd" else "completed"
    )
    safe = {
        "agent": "claude",
        "event": event[:80],
        "outcome": outcome,
        "at": datetime.now(timezone.utc).isoformat(timespec="seconds"),
    }
    _atomic_write(state_path, json.dumps(safe, ensure_ascii=False, indent=2) + "\n")


def main() -> int:
    parser = argparse.ArgumentParser(description="把批处理进度同步进 peach-data/state/job-status.md")
    parser.add_argument("--write", action="store_true")
    parser.add_argument("--db", type=Path, default=DATABASE_PATH)
    parser.add_argument("--generated", type=Path, default=GENERATED_DIR)
    parser.add_argument("--doc", type=Path, default=DOC)
    parser.add_argument("--state", type=Path, default=STATE)
    parser.add_argument("--hook-event", action="store_true", help="从 stdin 接收 Claude hook JSON")
    args = parser.parse_args()

    if args.hook_event:
        try:
            payload = json.load(sys.stdin)
            if not isinstance(payload, dict):
                raise ValueError("hook payload must be an object")
            record_hook_event(args.state, payload)
        except Exception as exc:
            print(f"自动交接事件记录失败:{exc}", file=sys.stderr)

    block = collect(args.db, args.generated, args.state)
    print("\n".join(block))
    if args.write:
        print("\n" + write_back(args.doc, block) + f" → {args.doc}")
    else:
        print("\n未加 --write,只打印未改文档。")
    return 0


if __name__ == "__main__":
    raise SystemExit(main())