feat: 接入 DeepSeek AI 分析公告是否为中国电信可承接项目

- 新增 ai_enabled/ai_api_key/ai_base_url/ai_model 等配置项，通过 .env 管理
- 新增 app/services/ai_analyzer.py — DeepSeek API 调用 + 详情页正文提取
- 新增 extract_page_content() 从详情页抓取正文供 AI 分析
- Announcement 模型新增 ai_relevant / ai_analysis 字段
- 流水线集成 AI 分析步骤（关键词匹配后、通知前）
- AI 标记为可承接的项目额外发送 markdown 着重通知
- 创建 alembic 迁移版本 6e8f4c2d1b0a

.env.example

@@ -29,3 +29,14 @@ SCHEDULER_CRON=0 8-21 * * *
 LOGHIVE_ENDPOINT=http://10.10.10.14:8000
 LOGHIVE_PROJECT=gx-gp-notify
 LOGHIVE_API_KEY=
+
+# AI 分析 (DeepSeek)
+AI_ENABLED=false
+AI_API_KEY=sk-your-deepseek-api-key
+AI_BASE_URL=https://api.deepseek.com/v1
+AI_MODEL=deepseek-chat
+AI_TIMEOUT=30
+AI_ANALYSIS_TITLE=🔔 中国电信可承接项目
+# 自定义分析提示词（可选），不设置则用代码内置默认值
+# 使用 \n 表示换行，支持 {title} {purchase_name} {announcement_type} {content} 四个占位符
+# AI_PROMPT_TEMPLATE=你是一个政府采购项目分析师...\n\n--- 公告信息 ---\n标题：{title}\n...

.gitignore

@@ -14,3 +14,5 @@ gx_gp_monitor/config/config.yaml
 onu.md
 vendor/build/
 vendor/*.egg-info/
+# Added by code-review-graph
+.code-review-graph/

alembic/versions/6e8f4c2d1b0a_add_ai_analysis_columns.py

@@ -0,0 +1,30 @@
+"""add_ai_analysis_columns
+
+Revision ID: 6e8f4c2d1b0a
+Revises: 1f59799a5083
+Create Date: 2026-05-21 10:00:00.000000
+
+"""
+from typing import Sequence, Union
+
+from alembic import op
+import sqlalchemy as sa
+
+
+# revision identifiers, used by Alembic.
+revision: str = '6e8f4c2d1b0a'
+down_revision: Union[str, Sequence[str], None] = '1f59799a5083'
+branch_labels: Union[str, Sequence[str], None] = None
+depends_on: Union[str, Sequence[str], None] = None
+
+
+def upgrade() -> None:
+    """添加 AI 分析相关字段"""
+    op.add_column('announcements', sa.Column('ai_relevant', sa.Boolean(), nullable=True))
+    op.add_column('announcements', sa.Column('ai_analysis', sa.Text(), nullable=True))
+
+
+def downgrade() -> None:
+    """回滚"""
+    op.drop_column('announcements', 'ai_analysis')
+    op.drop_column('announcements', 'ai_relevant')

app/config.py

@@ -1,5 +1,6 @@
 
 from pydantic_settings import BaseSettings, SettingsConfigDict
+from pydantic import field_validator
 
 
 class Settings(BaseSettings):
@@ -38,6 +39,40 @@ class Settings(BaseSettings):
     loghive_project: str = "gx-gp-notify"
     loghive_api_key: str = ""
 
+    # AI 分析 (DeepSeek)
+    ai_enabled: bool = False
+    ai_api_key: str = ""
+    ai_base_url: str = "https://api.deepseek.com/v1"
+    ai_model: str = "deepseek-chat"
+    ai_timeout: int = 30
+    ai_analysis_title: str = "🔔 中国电信可承接项目"
+    ai_prompt_template: str = (
+        "你是一个政府采购项目分析师，专门帮助中国电信识别可以承接的项目。\n\n"
+        "可承接范围包括但不限于：\n"
+        "- 通信工程、光缆建设、基站建设\n"
+        "- 信息化系统建设、系统集成\n"
+        "- 云计算、大数据、政务云\n"
+        "- 物联网、智慧城市、智慧园区\n"
+        "- 安防监控、视频会议、应急通信\n"
+        "- 网络运维、网络优化、IDC 服务\n"
+        "- 5G 应用、专线服务\n\n"
+        "--- 公告信息 ---\n"
+        "标题：{title}\n"
+        "采购人：{purchase_name}\n"
+        "公告类型：{announcement_type}\n\n"
+        "--- 公告正文 ---\n"
+        "{content}\n\n"
+        "请用 JSON 格式回答：\n"
+        '{{"is_relevant": true/false, "reason": "简要判断理由", "business_type": "业务分类"}}'
+    )
+
+    @field_validator("ai_prompt_template", mode="before")
+    @classmethod
+    def convert_newlines(cls, v: str) -> str:
+        """将 .env 中字面 \\n 转换为真实换行"""
+        if isinstance(v, str) and "\\n" in v:
+            return v.replace("\\n", "\n")
+        return v
 
     # 公告来源（JSON 字符串，从环境变量读取）
     announcement_sources: str = '{"ZcyAnnouncement1":{"category_id":66485,"name":"采购公告","type":"purchase"},"ZcyAnnouncement2":{"category_id":66485,"name":"结果公告","type":"result"},"ZcyAnnouncement3":{"category_id":66485,"name":"合同公告","type":"contract"},"ZcyAnnouncement4":{"category_id":66485,"name":"更正公告","type":"correction"},"ZcyAnnouncement5":{"category_id":66485,"name":"招标文件预公示","type":"pre_announcement"},"ZcyAnnouncement6":{"category_id":66485,"name":"单一来源公示","type":"single_source"},"ZcyAnnouncement7":{"category_id":66485,"name":"电子卖场公示","type":"electronic_market"},"ZcyAnnouncement10":{"category_id":66485,"name":"履约验收公示","type":"acceptance"},"ZcyAnnouncement11":{"category_id":66485,"name":"工程类公告","type":"engineering"},"ZcyAnnouncement20":{"category_id":66485,"name":"框架协议征集公告","type":"framework_agreement"},"ZcyAnnouncement21":{"category_id":66485,"name":"框架协议入围结果公告","type":"framework_result"},"ZcyAnnouncement23":{"category_id":66485,"name":"框架协议成交结果汇总公告","type":"framework_summary"},"61-266648":{"category_id":66485,"name":"采购意向公开","type":"intention"}}'  # noqa: E501

app/crawler/parsers.py

@@ -149,6 +149,57 @@ def parse_dahuagov_detail_pubdate(html: str) -> datetime | None:
     return None
 
 
+async def extract_page_content(url: str, timeout: int = 30) -> str | None:
+    """抓取详情页并用 BeautifulSoup 提取正文纯文本"""
+    import httpx
+
+    try:
+        headers = {
+            "User-Agent": "Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36",
+            "Accept": "text/html,application/xhtml+xml",
+            "Accept-Language": "zh-CN,zh;q=0.9",
+        }
+        async with httpx.AsyncClient(timeout=timeout, follow_redirects=True) as client:
+            resp = await client.get(url, headers=headers)
+            if resp.status_code != 200:
+                return None
+
+            soup = BeautifulSoup(resp.text, "html.parser")
+
+            # 尝试多种常见正文容器选择器
+            selectors = [
+                "div.article-content", "div.content", "div.TRS_Editor",
+                "div.Custom_UnionStyle", "div.pages_content", "div#content",
+                "div.main-content", "article", ".article", ".detail-content",
+                ".text-content", ".news-content", ".detail-article",
+            ]
+            for selector in selectors:
+                container = soup.select_one(selector)
+                if container:
+                    # 移除脚本和样式
+                    for tag in container.find_all(["script", "style"]):
+                        tag.decompose()
+                    text = container.get_text(separator="\n", strip=True)
+                    if len(text) > 50:  # 至少50字才算有效正文
+                        return text
+
+            # 兜底：取 body 内所有文本
+            body = soup.find("body")
+            if body:
+                for tag in body.find_all(["script", "style", "nav", "footer", "header"]):
+                    tag.decompose()
+                text = body.get_text(separator="\n", strip=True)
+                # 移除过长的空白行
+                lines = [l.strip() for l in text.split("\n") if l.strip()]
+                text = "\n".join(lines[:200])  # 最多取前200行
+                if len(text) > 50:
+                    return text
+
+            return None
+    except Exception:
+        return None
+
+
 def _generate_hash(ann: dict[str, Any]) -> str:
     content = (
         f"{ann['title']}|{ann['publish_date'].strftime('%Y-%m-%d')}"

app/models/announcement.py

@@ -25,6 +25,8 @@ class Announcement(Base):
     is_new: Mapped[bool] = mapped_column(Boolean, default=True)
     is_sent: Mapped[bool] = mapped_column(Boolean, default=False)
     keyword_matched: Mapped[bool] = mapped_column(Boolean, default=False)
+    ai_relevant: Mapped[bool | None] = mapped_column(Boolean, nullable=True, default=None)
+    ai_analysis: Mapped[str | None] = mapped_column(Text, nullable=True, default=None)
     created_at: Mapped[datetime] = mapped_column(DateTime, server_default=func.now())
     updated_at: Mapped[datetime] = mapped_column(DateTime, server_default=func.now(), onupdate=func.now())  # noqa: E501
 

app/services/ai_analyzer.py

@@ -0,0 +1,147 @@
+import json
+import logging
+from dataclasses import dataclass
+from typing import Any
+
+import httpx
+
+from app.config import settings
+from app.crawler.parsers import extract_page_content
+
+logger = logging.getLogger(__name__)
+
+
+@dataclass
+class AiResult:
+    """DeepSeek 分析结果"""
+    is_relevant: bool = False
+    reason: str = ""
+    business_type: str = ""
+    error: str | None = None
+    content_snippet: str | None = None  # 提取到的正文前 200 字，供入库参考
+
+
+class AiAnalyzer:
+    """AI 分析器 — 调用 DeepSeek 判断公告是否为中国电信可承接项目"""
+
+    def __init__(self):
+        self.api_key = settings.ai_api_key
+        self.base_url = settings.ai_base_url.rstrip("/")
+        self.model = settings.ai_model
+        self.timeout = settings.ai_timeout
+        self.prompt_template = settings.ai_prompt_template
+
+    async def analyze(self, announcement: dict[str, Any]) -> AiResult:
+        """分析单条公告"""
+        if not self.api_key:
+            return AiResult(error="AI_API_KEY 未配置")
+
+        # 1. 获取公告正文
+        content_url = announcement.get("content_url", "")
+        content = None
+        content_snippet = None
+        if content_url:
+            content = await extract_page_content(content_url, self.timeout)
+
+        if content:
+            content_snippet = content[:200]
+        else:
+            logger.warning("无法获取公告正文: %s", content_url)
+
+        # 2. 构建 prompt
+        prompt = self.prompt_template.format(
+            title=announcement.get("title", ""),
+            purchase_name=announcement.get("purchase_name", ""),
+            announcement_type=announcement.get("announcement_type", ""),
+            content=content or "（无法获取正文，请仅根据标题和采购人信息判断）",
+        )
+
+        # 3. 调用 DeepSeek API
+        try:
+            result = await self._call_deepseek(prompt)
+            if result.error:
+                return AiResult(error=result.error, content_snippet=content_snippet)
+            return AiResult(
+                is_relevant=result.is_relevant,
+                reason=result.reason,
+                business_type=result.business_type,
+                content_snippet=content_snippet,
+            )
+        except Exception as e:
+            logger.exception("AI 分析异常")
+            return AiResult(error=str(e), content_snippet=content_snippet)
+
+    async def analyze_batch(
+        self, announcements: list[dict[str, Any]], max_concurrent: int = 3
+    ) -> list[AiResult]:
+        """批量分析，控制并发数"""
+        import asyncio
+
+        sem = asyncio.Semaphore(max_concurrent)
+
+        async def _limited(ann: dict[str, Any]) -> AiResult:
+            async with sem:
+                return await self.analyze(ann)
+
+        tasks = [_limited(ann) for ann in announcements]
+        return await asyncio.gather(*tasks)
+
+    async def _call_deepseek(self, prompt: str) -> AiResult:
+        """调用 DeepSeek Chat API"""
+        url = f"{self.base_url}/chat/completions"
+        headers = {
+            "Authorization": f"Bearer {self.api_key}",
+            "Content-Type": "application/json",
+        }
+        payload = {
+            "model": self.model,
+            "messages": [
+                {
+                    "role": "system",
+                    "content": "你是一个专业的政府采购项目分析师。请根据公告信息判断是否为中国电信可以承接的项目，并用 JSON 格式回答。",
+                },
+                {"role": "user", "content": prompt},
+            ],
+            "temperature": 0.3,  # 低温度，提高判断一致性
+            "max_tokens": 512,
+        }
+
+        async with httpx.AsyncClient(timeout=self.timeout) as client:
+            response = await client.post(url, headers=headers, json=payload)
+            if response.status_code != 200:
+                return AiResult(
+                    error=f"API 请求失败 (HTTP {response.status_code}): {response.text[:200]}"
+                )
+
+            data = response.json()
+            choices = data.get("choices", [])
+            if not choices:
+                return AiResult(error="API 返回空 choices")
+
+            content = choices[0].get("message", {}).get("content", "")
+            return self._parse_response(content)
+
+    @staticmethod
+    def _parse_response(content: str) -> AiResult:
+        """从 LLM 回复中提取 JSON 结果"""
+        # 清理可能的 markdown 代码块标记
+        content = content.strip()
+        if content.startswith("```"):
+            # 移除 ```json 或 ``` 包裹
+            lines = content.split("\n")
+            if lines[0].strip().startswith("```"):
+                lines = lines[1:]
+            if lines and lines[-1].strip() == "```":
+                lines = lines[:-1]
+            content = "\n".join(lines).strip()
+
+        try:
+            result = json.loads(content)
+            return AiResult(
+                is_relevant=bool(result.get("is_relevant", False)),
+                reason=str(result.get("reason", "")),
+                business_type=str(result.get("business_type", "")),
+            )
+        except (json.JSONDecodeError, ValueError) as e:
+            logger.warning("JSON 解析失败: %s\n原始内容: %s", e, content[:200])
+            return AiResult(error=f"JSON 解析失败: {e}")

app/services/notification_service.py

@@ -17,6 +17,7 @@ class NotificationService:
         sent = 0
         for ann in announcements:
             try:
+                # 发送普通 textcard
                 title = ann.get("title", "")
                 if len(title) > 128:
                     title = title[:125] + "..."
@@ -36,7 +37,42 @@ class NotificationService:
 
                 if await self.client.send_textcard(title, description, url):
                     sent += 1
+
+                # AI 标记为可承接的，额外发送着重通知
+                ai_result = ann.get("ai_result")
+                if ai_result and ai_result.get("is_relevant"):
+                    await self._send_ai_emphasis(ann, ai_result)
+
             except Exception:
                 continue
 
         return sent
+
+    async def _send_ai_emphasis(
+        self, ann: dict[str, Any], ai_result: dict[str, Any]
+    ) -> bool:
+        """发送 AI 分析的着重通知（markdown 格式）"""
+        title = ann.get("title", "")
+        purchase_name = ann.get("purchase_name", "")
+        pub_date = ann.get("publish_date")
+        time_str = pub_date.strftime("%Y-%m-%d %H:%M") if pub_date else "时间未知"
+        url = ann.get("content_url", "")
+
+        reason = ai_result.get("reason", "")
+        business_type = ai_result.get("business_type", "")
+
+        # 企业微信 markdown 格式
+        md = (
+            f"{settings.ai_analysis_title}\n"
+            f"---\n"
+            f"**标题：** [{title}]({url})\n"
+            f"> 采购人：{purchase_name}\n"
+            f"> 发布时间：{time_str}\n\n"
+            f"**🤖 AI 分析：**\n"
+            f"> {reason}\n\n"
+            f"**🏷 业务分类：** {business_type}\n"
+            f"---\n"
+            f"[📄 查看公告原文]({url})"
+        )
+
+        return await self.client.send_markdown(md)

app/services/pipeline.py

@@ -1,8 +1,12 @@
+import logging
 from typing import Any
 
+from app.config import settings
 from app.crawler.base import PipelineConfig, PipelineResult
 from app.services.filter_service import dedup_by_hash
 
+logger = logging.getLogger(__name__)
+
 
 class PostCrawlPipeline:
     def __init__(self, db_session, notification_service):
@@ -41,6 +45,10 @@ class PostCrawlPipeline:
         if to_notify:
             to_notify = await self._exclude_sent(to_notify)
 
+        # 5.5 AI 分析（可选）
+        if settings.ai_enabled and to_notify:
+            await self._ai_analyze(to_notify)
+
         # 6. Notify
         if config.notify_mode == "all":
             result.notified = await self._send_notifications(to_notify)
@@ -129,3 +137,52 @@ class PostCrawlPipeline:
     def _match_keywords(announcement: dict[str, Any], keywords: list[str]) -> bool:
         text = f"{announcement.get('title', '')} {announcement.get('purchase_name', '')}"
         return any(kw in text for kw in keywords)
+
+    async def _ai_analyze(self, announcements: list[dict[str, Any]]) -> None:
+        """对公告列表执行 AI 分析，将结果附加到每条公告的 ai_result 字段，并更新数据库"""
+        from app.services.ai_analyzer import AiAnalyzer
+
+        analyzer = AiAnalyzer()
+        logger.info("AI 分析开始：共 %d 条公告", len(announcements))
+
+        results = await analyzer.analyze_batch(announcements)
+
+        ai_updates = []
+        for ann, ai_result in zip(announcements, results):
+            ann["ai_result"] = {
+                "is_relevant": ai_result.is_relevant,
+                "reason": ai_result.reason,
+                "business_type": ai_result.business_type,
+            }
+            if ai_result.error:
+                logger.warning("AI 分析失败 [%s]: %s", ann.get("title", "")[:30], ai_result.error)
+            else:
+                ai_updates.append({
+                    "content_hash": ann["content_hash"],
+                    "ai_relevant": ai_result.is_relevant,
+                    "ai_analysis": ai_result.reason,
+                })
+                if ai_result.is_relevant:
+                    logger.info("AI 标记可承接项目: %s (%s)", ann.get("title", "")[:40], ai_result.business_type)
+
+        # 批量更新数据库中的 AI 分析结果
+        if ai_updates:
+            await self._update_ai_results(ai_updates)
+
+    async def _update_ai_results(self, updates: list[dict[str, Any]]) -> None:
+        """批量更新公告的 AI 分析结果到数据库"""
+        from sqlalchemy import update
+
+        from app.models.announcement import Announcement
+
+        for u in updates:
+            stmt = (
+                update(Announcement)
+                .where(Announcement.content_hash == u["content_hash"])
+                .values(
+                    ai_relevant=u["ai_relevant"],
+                    ai_analysis=u["ai_analysis"],
+                )
+            )
+            await self.db.execute(stmt)
+        await self.db.commit()

docker/Dockerfile

@@ -1,5 +1,9 @@
 FROM python:3.12-slim
 
+LABEL maintainer="GX-gp-notify Dev Team" \
+      version="2.0.0" \
+      description="广西政府采购网公告监控系统"
+
 WORKDIR /app
 
 RUN apt-get update && apt-get install -y --no-install-recommends \

docker/docker-compose.yml

@@ -3,7 +3,10 @@ services:
     build:
       context: ..
       dockerfile: docker/Dockerfile
-    image: gx-gp-notify:latest
+      tags:
+        - gx-gp-notify:2.0.0
+        - gx-gp-notify:latest
+    image: gx-gp-notify:2.0.0
     container_name: gx-gp-notify
     ports:
       - "18001:8000"