first commit

config.yaml

@@ -0,0 +1,111 @@
+# 定时任务配置
+scheduler:
+  enabled: true
+  time: "09:00"
+  days:
+    - mon
+    - tue
+    - wed
+    - thu
+    - fri
+  max_instances: 3
+  coalesce: true
+
+# 目标网站配置
+targets:
+  - name: "国家税务总局"
+    url: "https://www.chinatax.gov.cn/"
+    list_paths:
+      - "/npsite/chinatax/zcwj/"
+      - "/npsite/chinatax/tzgg/"
+    keywords:
+      - "最新"
+      - "通知"
+      - "公告"
+      - "政策"
+      - "法规"
+    enabled: true
+
+  - name: "财政部"
+    url: "https://www.mof.gov.cn/"
+    list_paths:
+      - "/zhengwugongkai/zhengceku/zhengcefagui/"
+    keywords:
+      - "最新"
+      - "通知"
+      - "公告"
+      - "政策"
+      - "法规"
+    enabled: false
+
+  - name: "国家税务局"
+    url: "http://www.chinatax.gov.cn/"
+    list_paths:
+      - "/cloudfw/zcwj/"
+    keywords:
+      - "最新"
+      - "通知"
+      - "公告"
+      - "政策"
+      - "法规"
+    enabled: false
+
+# 下载配置
+download:
+  path: "./downloads"
+  formats:
+    - pdf
+    - doc
+    - docx
+    - txt
+    - xlsx
+  max_size: 52428800
+  timeout: 60
+  retry: 3
+  user_agent: "Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36"
+
+# 去重配置
+deduplication:
+  title_similarity: 0.8
+  content_similarity: 0.9
+  hash_algorithm: "simhash"
+
+# 分类配置
+categories:
+  - name: "税收政策"
+    keywords:
+      - "税收"
+      - "税务"
+      - "纳税"
+      - "税费"
+      - "增值税"
+      - "所得税"
+    priority: 1
+
+  - name: "通知公告"
+    keywords:
+      - "通知"
+      - "公告"
+      - "通告"
+    priority: 2
+
+  - name: "法规文件"
+    keywords:
+      - "法规"
+      - "条例"
+      - "规章"
+      - "办法"
+      - "细则"
+    priority: 3
+
+  - name: "其他政策"
+    keywords: []
+    priority: 99
+
+# 日志配置
+logging:
+  level: "INFO"
+  format: "%(asctime)s - %(name)s - %(levelname)s - %(message)s"
+  file: "./logs/policy_retrieval.log"
+  max_bytes: 10485760
+  backup_count: 5