feat: integrate govdoc module into leaudit platform

fastapi_modules/fastapi_leaudit/govdoc_engine/pipeline.py

@@ -130,12 +130,12 @@ def _merge_llm_into_entities(
 # ── 实体构建 (同步，供 sync 入口使用) ──────────────────
 
 def _build_entities(
-    doc, ruleset: RuleSet, llm: LlmClient,
+    doc, ruleset: RuleSet, llm: LlmClient | None,
 ) -> dict[str, SemanticEntity | None]:
     """构建实体 + 差量 LLM 抽取（同步）。"""
     entities = EntityBuilder().build(doc)
     spec = _compute_missing_spec(entities, ruleset.extract.entities)
-    if spec:
+    if spec and llm is not None:
         llm_vals = FieldExtractor(llm).extract_missing(doc, spec)
         _merge_llm_into_entities(entities, llm_vals)
     return entities
@@ -144,12 +144,12 @@ def _build_entities(
 # ── 实体构建 (异步，供 async 入口使用) ──────────────────
 
 async def _build_entities_async(
-    doc, ruleset: RuleSet, llm: LlmClient,
+    doc, ruleset: RuleSet, llm: LlmClient | None,
 ) -> dict[str, SemanticEntity | None]:
     """构建实体 + 差量 LLM 抽取（异步）。"""
     entities = EntityBuilder().build(doc)
     spec = _compute_missing_spec(entities, ruleset.extract.entities)
-    if spec:
+    if spec and llm is not None:
         llm_vals = await FieldExtractor(llm).extract_missing_async(doc, spec)
         _merge_llm_into_entities(entities, llm_vals)
     return entities
@@ -174,7 +174,7 @@ def audit_file(
     """
     docx_path = Path(docx_path)
     rules_path = Path(rules_path)
-    llm = llm_client or LlmClient()
+    llm = llm_client
 
     doc = parse_docx(docx_path)
     RoleTagger(llm_client=llm).tag(doc)
@@ -210,7 +210,7 @@ async def run(
     """
     file_path = Path(file_path)
     rules_path = Path(rules_path)
-    llm = llm_client or LlmClient()
+    llm = llm_client
 
     _log.info("Govdoc pipeline start: %s", file_path.name)
 
@@ -219,18 +219,21 @@ async def run(
     _log.info("  parsed: %d paragraphs", len(doc.paragraphs))
 
     # 2. 段落角色标注
-    RoleTagger(llm_client=llm).tag(doc)
+    if llm is not None:
+        await RoleTagger(llm_client=llm).tag_async(doc)
+    else:
+        RoleTagger(llm_client=None).tag(doc)
 
     # 3. 加载规则
     ruleset = load_rules(rules_path)
-    _log.info("  rules: %d groups, %d rules", len(ruleset.groups), len(ruleset.all_rules()))
+    _log.info("  rules: %d groups, %d rules", len(ruleset.rules), len(ruleset.all_rules()))
 
     # 4. 实体抽取 (含差量 LLM)
     entities = await _build_entities_async(doc, ruleset, llm)
     _log.info("  entities: %d/%d resolved", sum(1 for v in entities.values() if v), len(entities))
 
     # 5. 规则评估
-    findings, outcomes = RuleRunner(llm_client=llm).evaluate(
+    findings, outcomes = await RuleRunner(llm_client=llm).evaluate_async(
         ruleset.all_rules(), doc, entities
     )
     _log.info("  evaluated: %d findings from %d rules", len(findings), len(outcomes))