skills/ir_generation_skill/step1_semantic_index.py

@@ -536,17 +536,24 @@ def _quick_validate(
 
 
 def _build_coverage_feedback(gaps: dict) -> str:
-    """Generate feedback text for re-prompting when coverage is below threshold."""
+    """Generate targeted feedback text for re-prompting when coverage is below threshold."""
-    lines = []
+    parts = []
     for item in gaps.get("missing_paths", []):
-        lines.append(f"- {item}")
+        parts.append(f"- {item}")
-    if lines:
+    if not parts:
-        return (
+        return ""
-            "\n## 覆盖反馈（上一次运行未满足覆盖要求，请重新生成并修复以下缺口）\n\n"
+
-            + "\n".join(lines)
+    return (
-            + "\n\n请重新审视文档，为缺失的章节和表格行创建对应的 function_unit。"
+        "\n## 关键覆盖反馈（上一轮 LLM 输出了以下缺口，请重新处理）\n\n"
-        )
+        + "\n".join(parts)
-    return ""
+        + "\n\n"
+        "### 修复动作（必须执行）\n\n"
+        "1. **重新扫描上述每个缺失章节**，从文字和表格中提取所有可被测试的功能行为\n"
+        "2. **为每个缺失的表格行创建独立的 function_unit**，不得合并不同行的规则\n"
+        "3. **每个 function_unit 必须引用具体的 section 号和 row 号**作为 source\n"
+        "4. **非功能章节可以跳过**（如背景、术语、变更日志），但行为规则章节必须覆盖\n"
+        "5. 输出中必须包含针对上述缺口的新 function_unit\n"
+    )
 
 
 def _collect_logic_tree_nodes(doc: dict) -> dict[str, dict[str, str]]:
@@ -767,13 +774,22 @@ def run_ensemble_semantic_index(doc: dict) -> dict:
         # Feedback retry: re-run with coverage feedback (one retry)
         feedback = _build_coverage_feedback(gaps)
         if feedback:
-            print(f"\n  覆盖反馈重试...")
+            print(f"\n  覆盖反馈重试 (feedback长度={len(feedback)}字符)...", flush=True)
             try:
                 retry_prompt = build_prompt(doc, feedback, all_paths)
-                retry_result = call_llm(retry_prompt, max_retries=1, temperature=0.0)
+                print(f"  重试 prompt 长度: {len(retry_prompt)} 字符", flush=True)
+                retry_result = call_llm(retry_prompt, max_retries=1, temperature=0.3)
                 n_retry_units = len(retry_result.get("function_units", []))
-                print(f"  重试返回: {n_retry_units} 功能单元")
+                n_retry_concepts = len(retry_result.get("concepts", []))
+                print(f"  重试返回: {n_retry_concepts} 概念, {n_retry_units} 功能单元", flush=True)
                 if n_retry_units > 0:
+                    # Check which new sections were covered
+                    retry_sections = set()
+                    for fu in retry_result.get("function_units", []):
+                        for src in fu.get("sources", []):
+                            if src.get("section"):
+                                retry_sections.add(src["section"])
+                    print(f"  重试新增 sections: {sorted(retry_sections)}", flush=True)
                     # Merge retry into results and re-validate
                     semantic_indices.append(retry_result)
                     merged = ensemble_merge(semantic_indices)
@@ -783,9 +799,11 @@ def run_ensemble_semantic_index(doc: dict) -> dict:
                     merged["validation_gaps"] = {
                         k: v for k, v in gaps.items() if v
                     }
-                    print(f"  重试后验证: {'PASS' if passed else 'GAPS FOUND'}")
+                    print(f"  重试后验证: {'PASS' if passed else 'GAPS FOUND'}", flush=True)
             except Exception as e:
-                print(f"  覆盖反馈重试失败: {e}")
+                print(f"  覆盖反馈重试失败: {e}", flush=True)
+                import traceback
+                traceback.print_exc()
 
     return merged