feat: 更新安装调试审查;

core/config.py

@@ -12,24 +12,18 @@ use_docker = False
 #     model: str = "deepseek-v4-pro"
 
 
-# @dataclass
-# class LLMConfig:
-#     base_url: str = "http://172.21.107.80:9002/v1"
-#     api_key: str = "none"
-#     model: str = "Qwen2-72B-Instruct"
+@dataclass
+class LLMConfig:
+    base_url: str = "http://172.21.107.80:9002/v1"
+    api_key: str = "none"
+    model: str = "Qwen2-72B-Instruct"
 
 # @dataclass
 # class LLMConfig:
-#     base_url: str = "http://172.21.107.80:9002/v1"
+#     base_url: str = "http://192.168.252.71:9002/v1"
 #     api_key: str = "none"
 #     model: str = "Qwen2-72B-Instruct"
 
-@dataclass
-class LLMConfig:
-    base_url: str = "http://192.168.252.71:9002/v1"
-    api_key: str = "none"
-    model: str = "Qwen2-72B-Instruct"
-
 # 最大分片数量
 min_single_chunk_size = 2000
 max_single_chunk_size = 100000
@@ -63,8 +57,8 @@ MAX_WORKERS = 10
 FILE_SUFFIX = "-审核批注"
 
 ## 关键参数**
-use_non_fastgpt_llm = True
-use_lufa = True
+use_non_fastgpt_llm = False
+use_lufa = False
 use_jp_machine = True
 debug_mode = False
 

core/tools/rule_filter.py

@@ -7,50 +7,35 @@ from core.tool import tool, tool_func
 
 @tool("rule_filter", "规则过滤")
 class RuleFilterTool:
-    @tool_func(
-        {
-            "type": "object",
-            "properties": {
-                "payload": {"type": "object"},
-            },
-            "required": ["payload"],
-        }
-    )
-    def run(self, payload: Dict[str, Any]) -> Dict[str, Any]:
-        raise NotImplementedError("Subclasses must implement run")
+    def pre_run(self, payload: Dict[str, Any]) -> Dict[str, Any]:
+        return payload
+    def post_run(self, payload: Dict[str, Any]) -> Dict[str, Any]:
+        return payload
 
 
 @tool("lufa_party_rule_filter_tool", "LUFA 当事人与支付主体规则过滤")
 class LufaPartyRuleFilterTool(RuleFilterTool):
-    @tool_func(
-        {
-            "type": "object",
-            "properties": {
-                "payload": {"type": "object"},
-            },
-            "required": ["payload"],
-        }
-    )
-    def run(self, payload: Dict[str, Any]) -> Dict[str, Any]:
-        # rules = payload.get("rules") or []
-        # segment_idx = int(payload.get("segment_idx", 0))
-        # total_segments = int(payload.get("total_segments", 0))
-
-        # if not rules or total_segments <= 0:
-        #     payload["rules"] = rules
-        #     return payload
-
-        # # 奇数分段时将中间段归入前半段
-        # first_half_count = (total_segments + 1) // 2
-
-        # if segment_idx < first_half_count:
-        #     filtered_rules: List[Dict[str, Any]] = [
-        #         r for r in rules if "支付主体审查" not in str(r.get("title", ""))
-        #     ]
-        # else:
-        #     filtered_rules = [
-        #         r for r in rules if "当事人审查" not in str(r.get("title", ""))
-        #     ]
-
-        # payload["rules"] = filtered_rules
+    def pre_run(self, payload: Dict[str, Any]) -> Dict[str, Any]:
+        return payload
+
+    def post_run(self, payload: Dict[str, Any]) -> Dict[str, Any]:
+        return payload
+
+@tool("jp_party_rule_filter_tool", "JP 通用审查过滤")
+class JpRuleFilterTool(RuleFilterTool):
+    # 过滤通用审查
+    def pre_run(self, payload: Dict[str, Any]) -> Dict[str, Any]:
+        rules = payload.get("rules") or []
+        filtered_rules: List[Dict[str, Any]] = [
+            r for r in rules if "通用审查" not in str(r.get("title", ""))
+        ]
+        payload["rules"] = filtered_rules
         return payload
+    
+    # 增加通用审查
+    def post_run(self, payload: Dict[str, Any]) -> Dict[str, Any]:
+        if 'routed_rule_titles' in payload:
+            routed_rule_titles = payload['routed_rule_titles']
+            routed_rule_titles.append('通用审查')
+            payload['routed_rule_titles'] = routed_rule_titles
+        return payload
\ No newline at end of file

demo/spire_doc_remove_comments.py

@@ -4,7 +4,7 @@ from spire.doc.common import *
 # 创建一个 Document 类对象并加载一个 Word 文档
 doc = Document()
 doc.LoadFromFile(
-    "/home/ccran/lufa-contract/demo/湖南麓谷发展集团“主数据管理系统与合同管理系统开发”项目合同协议书-审核批注20260511_153215.docx"
+    r"/Users/chenran/Desktop/test.doc"
 )
 
 # 移除第二个注释
@@ -15,6 +15,6 @@ doc.Comments.Clear()
 
 # 保存文档
 doc.SaveToFile(
-    "/home/ccran/lufa-contract/demo/湖南麓谷发展集团“主数据管理系统与合同管理系统开发”项目合同协议书-审核批注20260511_153215-无批注.docx"
+    r"/Users/chenran/Desktop/test_no_comments.doc"
 )
 doc.Close()

main.py

@@ -26,7 +26,7 @@ from core.config import (
 from core.tools.segment_summary import SegmentSummaryTool
 from core.tools.segment_review import SegmentReviewTool
 from core.tools.segment_rule_router import SegmentRuleRouterTool
-from core.tools.rule_filter import LufaPartyRuleFilterTool
+from core.tools.rule_filter import LufaPartyRuleFilterTool,JpRuleFilterTool,RuleFilterTool
 from core.tools.rules_retrieve_reference import RulesRetrieveReferenceTool
 from core.tools.reflect_retry import ReflectRetryTool
 from core.tools.segment_merger import SegmentMergerTool
@@ -43,7 +43,12 @@ TMP_DIR.mkdir(parents=True, exist_ok=True)
 summary_tool = SegmentSummaryTool()
 review_tool = SegmentReviewTool()
 rule_router_tool = SegmentRuleRouterTool()
-lufa_party_rule_filter_tool = LufaPartyRuleFilterTool()
+rule_filter_tool = RuleFilterTool()
+# if use_lufa:
+#     rule_filter_tool = LufaPartyRuleFilterTool()
+# else:
+#     rule_filter_tool = JpRuleFilterTool()
+
 rules_reference_tool = RulesRetrieveReferenceTool()
 reflect_tool = ReflectRetryTool()
 merger_tool = SegmentMergerTool()
@@ -431,20 +436,19 @@ def route_segment_rules(payload: SegmentReviewRequest) -> SegmentRuleRouterRespo
     ruleset_id = payload.ruleset_id or rules_reference_tool.default_ruleset_id
     rules = rules_reference_tool.run(ruleset_id=ruleset_id).get("rules", [])
 
-    if use_lufa and rules:
-        try:
-            total_segments = len(doc_obj.get_chunk_id_list() or [])
-        except Exception:
-            total_segments = 0
-
-        filtered_payload = lufa_party_rule_filter_tool.run(
-            {
+    try:
+        total_segments = len(doc_obj.get_chunk_id_list() or [])
+    except Exception:
+        total_segments = 0
+    # filter rules
+    filtered_payload = rule_filter_tool.pre_run(
+        {
                 "rules": rules,
                 "segment_idx": segment_idx,
                 "total_segments": total_segments,
-            }
-        )
-        rules = filtered_payload.get("rules", rules)
+        }
+    )
+    rules = filtered_payload.get("rules", rules)
 
     result = rule_router_tool.run(
         segment_id=segment_idx,
@@ -454,13 +458,19 @@ def route_segment_rules(payload: SegmentReviewRequest) -> SegmentRuleRouterRespo
         context_memories=payload.context_memories,
         route_by=payload.route_by,
     )
+    routed_rule_titles=result.get("routed_rule_titles", [])
+    routed_rule_titles = rule_filter_tool.post_run(
+        {
+            "routed_rule_titles": routed_rule_titles
+        }
+    ).get("routed_rule_titles", routed_rule_titles)
 
     return SegmentRuleRouterResponse(
         conversation_id=payload.conversation_id,
         segment_id=payload.segment_id,
         ruleset_id=ruleset_id,
         route_by=result.get("route_by", payload.route_by),
-        routed_rule_titles=result.get("routed_rule_titles", []),
+        routed_rule_titles=routed_rule_titles,
         routed_summary_names=result.get("routed_summary_names", []),
         routed_rules=result.get("routed_rules", []),
     )