platform-assistant/scripts/direv-publish/batch_publish_v4.py

#!/usr/bin/env python3
"""
批量发布招标公告到 direv.com.cn（v4 — 备注 + PDF 上传）
基于 v3，新增：
  - 从招标公告 PDF/HTML 提取关键商业要求，填入"备注"字段
  - 电建公告：上传公告 PDF 到备注区
  - 新矿公告：仅填备注文本，不上传 PDF
"""

import asyncio
import json
import os
import re
import sys
from datetime import datetime, timedelta

from playwright.async_api import async_playwright

sys.path.insert(0, str(__file__).rsplit("/", 1)[0])
from publish import fill_form, submit_form, AUTH_FILE
from extract_notes import get_notes

# ============ 公告定义 ============
# 新增字段:
#   source: "电建" | "新矿"
#   notes_path: PDF/HTML 文件路径（可选，留空则自动匹配）

ANNOUNCEMENTS = [
    # ==== 昨天 (5/18) 电建公告 ====
    # 1. 莱索托水泵
    {
        "title": "莱索托高地引水项目II期水泵采购",
        "deadline": "2026-05-26",
        "region": "天津或上海港等国内指定港口仓库",
        "category": "水泵",
        "unit": "台",
        "quantity": "10",
        "source": "电建",
        "notes_path": "",
    },
    # 2. 九峰山钢材
    {
        "title": "九峰山抽水蓄能电站工字钢槽钢镀锌扁铁采购",
        "deadline": "2026-05-26",
        "region": "河南省新乡市辉县市九峰山抽水蓄能电站",
        "category": "钢材",
        "unit": "吨",
        "quantity": "335",
        "source": "电建",
        "notes_path": "",
    },
    # 3. 武铁砌块
    {
        "title": "武铁信阳生产基地蒸压加气混凝土砌块采购",
        "deadline": "2026-05-22",
        "region": "河南省信阳市武铁项目施工现场",
        "category": "蒸压加气混凝土砌块",
        "unit": "m³",
        "quantity": "4500",
        "source": "电建",
        "notes_path": "",
    },
    # 4. 九峰山气体
    {
        "title": "九峰山抽水蓄能电站工业气体采购",
        "deadline": "2026-05-26",
        "region": "河南省新乡市辉县市九峰山抽水蓄能电站",
        "category": "工业气体",
        "unit": "瓶",
        "quantity": "2070",
        "source": "电建",
        "notes_path": "",
    },
    # 5. 庙湾水库打包箱房
    {
        "title": "郑州市庙湾水库打包箱房采购",
        "deadline": "2026-05-25",
        "region": "河南省郑州市高新区庙湾水库工程现场",
        "category": "打包箱房",
        "unit": "间",
        "quantity": "105",
        "source": "电建",
        "notes_path": "",
    },
    # 6. 红棉大道混凝土制品（30项合并为一条，用规格区分）
    {
        "title": "红棉大道工程混凝土制品采购",
        "deadline": "2026-05-25",
        "region": "广东省广州市花都区红棉大道工程田心立交至风神立交段",
        "category": "混凝土制品",
        "unit": "批",
        "quantity": "1",
        "source": "电建",
        "notes_path": "",
    },
    # 7. 巩义人防封堵板
    {
        "title": "巩义人民医院东区医院人防封堵板采购",
        "deadline": "2026-05-25",
        "region": "河南省巩义市巩义市人民医院东区医院二期续建工程",
        "category": "人防封堵板",
        "unit": "樘",
        "quantity": "23",
        "source": "电建",
        "notes_path": "",
    },
    # 8. 紫坪铺装修材料
    {
        "title": "紫坪铺项目装修材料采购",
        "deadline": "2026-05-25",
        "region": "四川省成都市都江堰市龙池镇紫坪铺项目",
        "category": "装修材料",
        "unit": "批",
        "quantity": "1",
        "source": "电建",
        "notes_path": "",
    },
]

COVER_DIR = "/tmp/direv_covers"

# ============ 品类→文件名关键词映射 ============
# 用于精确匹配品类名与公告 PDF 文件名（品类名和文件名词不同时使用）
CATEGORY_FILE_KEYWORDS = {
    "工业气体": ["液态氧气", "液态二氧化碳", "丙烷", "乙炔"],
    "混凝土制品": ["钢筋混凝土管", "混凝土检查井", "透水砖", "电缆沟盖板", "仿花岗岩", "预制砼"],
    "装修材料": ["木地板", "踢脚线", "防盗门", "玻璃隔断", "石膏板吊顶"],
    "打包箱房": ["打包箱房"],
    "水泵": ["水泵"],
    "人防封堵板": ["人防封堵板", "封堵板"],
    "蒸压加气混凝土砌块": ["蒸压加气混凝土砌块", "混凝土砌块"],
}

# ============ 线索→文件匹配 ============


def find_notes_file(lead: dict) -> str | None:
    """
    为一条线索自动匹配 PDF（电建）或 HTML（新矿）文件。
    从最近 7 天的下载目录搜索，按标题/品类关键词匹配。
    """
    source = lead.get("source", "电建")
    ext = "pdf" if source == "电建" else "html"

    # 提取关键词
    category = lead.get("category", "")
    title = lead.get("title", "")
    proj_keywords = []
    for m in re.finditer(r"[\u4e00-\u9fff]{2,}(?:工程|项目|电站|水库|风电)", title):
        proj_keywords.append(m.group())

    today = datetime.now()
    all_candidates = []

    for days_back in range(7):
        date_dir = (today - timedelta(days=days_back)).strftime("%Y-%m-%d")
        search_dir = os.path.expanduser(f"~/desktop/招标公告/{date_dir}/{ext}")
        if not os.path.isdir(search_dir):
            continue

        for fname in os.listdir(search_dir):
            if not fname.endswith(f".{ext}"):
                continue
            full_path = os.path.join(search_dir, fname)

            score = 0
            if category in fname:
                score += 10
            # 品类→文件名关键词映射（补充匹配）
            extra_kws = CATEGORY_FILE_KEYWORDS.get(category, [])
            for kw in extra_kws:
                if kw in fname:
                    score += 10
                    break  # 命中一个关键词即可
            for kw in proj_keywords:
                if kw in fname:
                    score += 5

            if score > 0:
                all_candidates.append((score, full_path, fname))

        # 如果当前目录已有候选且来自最近2天，优先使用
        if all_candidates and days_back < 2:
            all_candidates.sort(key=lambda x: x[0], reverse=True)
            best = all_candidates[0]
            print(f"    自动匹配: {best[2]}")
            return best[1]

    if not all_candidates:
        return None

    all_candidates.sort(key=lambda x: x[0], reverse=True)
    best = all_candidates[0]
    print(f"    自动匹配: {best[2]}")
    return best[1]


def compute_delivery(deadline_str: str) -> str:
    if not deadline_str:
        # 默认：今天 + 30 天
        d = datetime.now() + timedelta(days=30)
    else:
        try:
            d = datetime.strptime(deadline_str, "%Y-%m-%d")
        except (ValueError, TypeError):
            d = datetime.now() + timedelta(days=30)
    return d.strftime("%Y-%m-%d")


# ============ 备注填充 ============


async def fill_notes(page, notes_text: str):
    """填入备注 textarea"""
    await page.evaluate("""(text) => {
        const ta = document.querySelector('.beizhu textarea.el-textarea__inner');
        if (!ta) return false;
        Object.getOwnPropertyDescriptor(
            HTMLTextAreaElement.prototype, 'value'
        ).set.call(ta, text);
        ta.dispatchEvent(new Event('input', {bubbles: true}));
        ta.dispatchEvent(new Event('change', {bubbles: true}));
        return true;
    }""", notes_text)
    print(f"    ✓ 备注已填入 ({len(notes_text)} 字)")


async def upload_pdf(page, pdf_path: str):
    """上传 PDF 到备注区"""
    try:
        await page.wait_for_timeout(500)

        # 方式1：直接 setInputFiles（更可靠，不需要 file_chooser 事件）
        uploaded = await page.evaluate("""() => {
            const input = document.querySelector('.beizhu .el-upload__input');
            return !!input;
        }""")

        if uploaded:
            # 用 Playwright locator 定位 file input 并直接 set input files
            file_input = page.locator('.beizhu .el-upload__input')
            await file_input.set_input_files(pdf_path)
            await page.wait_for_timeout(3000)
            print(f"    ✓ PDF 已上传: {os.path.basename(pdf_path)}")
            return True

        # 方式2：fallback — 点击上传区域触发 file_chooser
        async with page.expect_file_chooser(timeout=5000) as fc_info:
            # 点击 up_con 内的任意可点击区域
            await page.evaluate("""() => {
                const upCon = document.querySelector('.beizhu .up_con');
                if (!upCon) return;
                // 尝试点击内部元素
                const clickable = upCon.querySelector('*') || upCon;
                clickable.click();
            }""")
            file_chooser = await fc_info.value
            await file_chooser.set_files(pdf_path)

        await page.wait_for_timeout(3000)
        print(f"    ✓ PDF 已上传: {os.path.basename(pdf_path)}")
        return True

    except Exception as e:
        print(f"    ⚠️ PDF 上传异常: {e}")
        return False


# ============ 封面图 ============

async def ai_generate_cover(category: str, filepath: str) -> bool:
    """用 MiniMax image-01 生成商品封面图"""
    auth_path = os.path.expanduser("~/.openclaw/agents/main/agent/auth-profiles.json")
    api_key = None
    try:
        with open(auth_path) as f:
            profiles = json.load(f)
        container = profiles.get("profiles", profiles)
        if isinstance(container, dict):
            for key, val in container.items():
                if isinstance(val, dict) and val.get("provider") == "minimax":
                    api_key = val.get("key")
                    break
    except Exception:
        pass

    if not api_key:
        return False

    prompt_map = {
        "石灰": "石灰，工业用生石灰，白色块状或粉末状，建筑材料，白色背景，高清商品展示，表面无任何文字",
        "水泥": "水泥，建筑用普通硅酸盐水泥，灰色粉末，袋装水泥堆放，建筑材料，白色背景，高清摄影，包装袋表面无任何文字",
        "钢筋": "钢筋，建筑用螺纹钢，热轧带肋钢筋HRB400，金属质感，白色背景，高清商品摄影，表面无任何文字",
        "钢材": "钢材，工字钢H型钢槽钢，建筑钢结构材料，金属质感，工业材料，白色背景，高清商品展示，表面无任何文字",
        "选矿药剂": "选矿药剂，工业浮选用化学药剂，化工原料桶装，矿业用品，白色背景，商品展示，高清摄影，桶面无任何文字",
        "水泵": "工业水泵，离心泵潜水泵，金属铸造泵体，工业设备，白色背景，高清商品摄影，表面无任何文字",
        "蒸压加气混凝土砌块": "蒸压加气混凝土砌块，灰白色轻质砌块，建筑材料堆场，白色背景，高清摄影，表面无任何文字",
        "工业气体": "工业气体，氧气乙炔丙烷钢瓶，金属气瓶排列，工业用品，白色背景，高清摄影，瓶身无任何文字",
        "打包箱房": "打包箱房，工地临时活动板房，钢结构模块化箱房，建筑临建，白色背景，高清展示，表面无任何文字",
        "混凝土制品": "混凝土制品，钢筋混凝土管检查井盖板，预制构件堆场，建筑材料，白色背景，高清摄影，表面无任何文字",
        "人防封堵板": "人防封堵板，人防工程防护设备，钢制防护门封堵板，建筑安全设备，白色背景，高清商品展示，表面无任何文字",
        "装修材料": "装修材料，木地板防盗门玻璃隔断，室内装饰建材，白色背景，高清商品展示，表面无任何文字",
    }
    default_prompt = f"{category}，工业用{category}产品，建筑材料或工业原料，白色背景，商品展示，高清摄影，表面无任何文字"
    prompt = prompt_map.get(category, default_prompt)

    headers = {"Authorization": f"Bearer {api_key}", "Content-Type": "application/json"}
    payload = {"model": "image-01", "prompt": prompt, "n": 1, "aspect_ratio": "1:1"}

    import io as _io
    import requests as _requests
    from PIL import Image as _Image

    endpoints = ["https://api.minimax.chat/v1/image_generation", "https://api.minimaxi.com/v1/image_generation"]
    for url in endpoints:
        try:
            r = _requests.post(url, headers=headers, json=payload, timeout=60)
            if r.status_code != 200:
                continue
            data = r.json()
            img_urls = data.get("data", {}).get("image_urls", [])
            if not img_urls:
                continue
            img_resp = _requests.get(img_urls[0], timeout=30)
            if img_resp.status_code != 200 or len(img_resp.content) < 5000:
                continue
            img = _Image.open(_io.BytesIO(img_resp.content)).convert("RGB")
            w, h = img.size
            if w > h:
                left = (w - h) // 2
                img = img.crop((left, 0, left + h, h))
            elif h > w:
                top = (h - w) // 2
                img = img.crop((0, top, w, top + w))
            img = img.resize((400, 400), _Image.LANCZOS)
            buf = _io.BytesIO()
            img.save(buf, "JPEG", quality=85, optimize=True)
            processed = buf.getvalue()
            if len(processed) < 5000:
                continue
            with open(filepath, "wb") as f:
                f.write(processed)
            return True
        except Exception:
            continue
    return False


async def fetch_cover_image(category: str, max_retries: int = 3) -> str:
    """获取商品封面图（AI 生成）"""
    category = category.strip()
    os.makedirs(COVER_DIR, exist_ok=True)
    safe = re.sub(r"[^\w\u4e00-\u9fff-]", "_", category)[:20]
    filepath = os.path.join(COVER_DIR, f"{safe}_cover.jpg")

    if os.path.exists(filepath):
        return filepath

    for attempt in range(1, max_retries + 1):
        print(f"    MiniMax 生图: {category}（第{attempt}次）")
        ok = await ai_generate_cover(category, filepath)
        if ok:
            size = os.path.getsize(filepath)
            print(f"    ✓ AI 生成成功 ({size//1024}KB)")
            return filepath
        if attempt < max_retries:
            print(f"    ⚠️ 第{attempt}次失败，重试...")
            await asyncio.sleep(2)

    raise RuntimeError(f"AI 生图失败: {category}（重试{max_retries}次均失败）")


async def upload_cover_image(page, row_index: int, image_path: str) -> bool:
    """上传商品封面图到询价物料表格"""
    try:
        await page.wait_for_timeout(1000)
        async with page.expect_file_chooser() as fc_info:
            clicked = await page.evaluate("""([idx]) => {
                var rows = document.querySelectorAll('.el-table__body-wrapper tbody tr');
                if (!rows[idx]) return false;
                var els = rows[idx].querySelectorAll('*');
                for (var el of els) {
                    if (el.children.length === 0 && el.textContent.includes('上传商品封面')) {
                        el.click();
                        return true;
                    }
                }
                return false;
            }""", [row_index])

        if not clicked:
            print(f"    未找到第{row_index+1}行上传按钮")
            return False

        file_chooser = await fc_info.value
        await file_chooser.set_files(image_path)
        await page.wait_for_timeout(4000)
        print(f"    ✓ 封面图已上传")
        return True
    except Exception as e:
        print(f"    上传异常: {e}")
        return False


# ============ 发布单条 ============

async def publish_one(browser, lead: dict, idx: int, total: int, cover_path: str | None):
    ctx_options = {"viewport": {"width": 1440, "height": 900}}
    if os.path.exists(AUTH_FILE):
        ctx_options["storage_state"] = AUTH_FILE
    ctx = await browser.new_context(**ctx_options)
    page = await ctx.new_page()
    page.set_default_timeout(15000)

    lead = dict(lead)
    # deadline 默认兜底：今天 + 7 天
    if not lead.get("deadline"):
        from datetime import datetime, timedelta
        lead["deadline"] = (datetime.now() + timedelta(days=7)).strftime("%Y-%m-%d")
    lead["delivery"] = compute_delivery(lead["deadline"])

    print(f"\n[{idx+1}/{total}] {lead['title']}")
    print(f"    交货日期: {lead['delivery']}")

    # --- 处理备注内容 ---
    source = lead.get("source", "电建")
    notes_path = lead.get("notes_path", "")

    # 自动匹配文件
    if not notes_path:
        notes_path = find_notes_file(lead)

    notes_text = ""
    if notes_path and os.path.exists(notes_path):
        try:
            ext = os.path.splitext(notes_path)[1].lower()
            source_type = "pdf" if ext == ".pdf" else "html"
            notes_text = get_notes(source_type, notes_path)
            print(f"    备注摘要: {notes_text[:80]}...")
        except Exception as e:
            print(f"    ⚠️ 备注提取失败: {e}")

    # --- 导航 ---
    await page.goto("https://www.direv.com.cn/#/index", wait_until="domcontentloaded")
    await page.wait_for_timeout(5000)

    # 点击发布询价（优先 button，精确匹配 → 模糊匹配）
    clicked = False
    # 1) 优先找 button 元素，文字精确为「发布询价」
    for el in await page.query_selector_all("button"):
        try:
            if (await el.text_content() or "").strip() == "发布询价":
                await el.click()
                clicked = True
                break
        except Exception:
            pass
    # 2) 兜底：模糊匹配所有元素（兼容旧文案「发布询价信息」）
    if not clicked:
        for el in await page.query_selector_all("span, div, button"):
            try:
                text = (await el.text_content() or "").strip()
                if text == "发布询价信息":
                    await el.click()
                    clicked = True
                    break
            except Exception:
                pass
    if not clicked:
        print(f"  ❌ 未找到发布按钮")
        await ctx.close()
        return False

    await page.wait_for_timeout(8000)

    try:
        await page.wait_for_selector(".c_title", timeout=15000)
    except Exception:
        print(f"  ❌ 表单未加载")
        await ctx.close()
        return False

    # --- 填表 ---
    await fill_form(page, lead)

    # --- 备注 + PDF 上传 ---
    if notes_text:
        await fill_notes(page, notes_text)

    if notes_path and source == "电建" and notes_path.endswith(".pdf"):
        await upload_pdf(page, notes_path)

    # --- 上传封面图 ---
    if cover_path and os.path.exists(cover_path):
        print(f"    上传封面: {os.path.basename(cover_path)} ({os.path.getsize(cover_path)//1024}KB)")
        await upload_cover_image(page, 0, cover_path)
    else:
        print(f"    ⚠️ 无封面图")

    # --- 提交 ---
    print(f"    提交...")
    ok = await submit_form(page)
    if not ok:
        print(f"    ❌ 验证错误")
        await page.screenshot(path=f"/tmp/direv_err_{lead['title'][:10]}.png", full_page=True)
        await ctx.close()
        return False

    await page.wait_for_timeout(3000)
    await page.screenshot(path=f"/tmp/direv_{idx+1}_{lead['title'][:10]}.png", full_page=True)
    print(f"    ✅ 发布成功")
    await ctx.close()
    return True


# ============ 主流程 ============

async def batch_publish(headless=True):
    print(f"模式: {'headless' if headless else 'headed'}")
    print(f"待发布: {len(ANNOUNCEMENTS)} 条\n")

    # 预下载封面图
    print("预下载封面图...")
    cover_paths = {}
    for i, lead in enumerate(ANNOUNCEMENTS):
        cat = lead["category"]
        print(f"  [{i+1}/{len(ANNOUNCEMENTS)}] 封面: {cat}")
        path = await fetch_cover_image(cat)
        cover_paths[i] = path
        if path:
            size = os.path.getsize(path)
            print(f"    ✅ {os.path.basename(path)} ({size//1024}KB)")
        else:
            print(f"    ❌ 生图失败")

    # 发布
    async with async_playwright() as p:
        browser = await p.chromium.launch(headless=headless)
        success = 0
        for i, lead in enumerate(ANNOUNCEMENTS):
            try:
                ok = await publish_one(browser, lead, i, len(ANNOUNCEMENTS), cover_paths.get(i))
                if ok:
                    success += 1
            except Exception as e:
                print(f"  ❌ 异常: {e}")
        await browser.close()

    print(f"\n========== 完成 ==========")
    print(f"成功: {success}/{len(ANNOUNCEMENTS)} 条")


if __name__ == "__main__":
    headless = "--headed" not in sys.argv
    asyncio.run(batch_publish(headless=headless))