From 62266dde4d7fb2238855a105bef1fd1706e671f6 Mon Sep 17 00:00:00 2001
From: Peter Zhang <18501667167@qq.com>
Date: Sun, 31 May 2026 20:20:54 +0800
Subject: [PATCH] =?UTF-8?q?fix:=20=E4=BF=AE=E5=A4=8D=20trigger.operator=20?=
 =?UTF-8?q?null=20+=20=E6=B7=BB=E5=8A=A0=E8=A6=86=E7=9B=96=E5=8F=8D?=
 =?UTF-8?q?=E9=A6=88=E9=87=8D=E8=AF=95=20-=20Closes=20#22,=20Closes=20#21?=
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

#22: _normalize_rule 补充 trigger 级别 operator (AND/OR) 默认值
#21: step1 验证失败时自动生成覆盖反馈并重试一轮
#22: step2 过滤空规则片段，避免污染下游

Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
---
 .../step1_semantic_index.py                   | 37 +++++++++++++++++++
 .../step2_ir_extraction.py                    |  7 ++++
 .../step3_merge_and_audit.py                  |  7 +++-
 3 files changed, 49 insertions(+), 2 deletions(-)

diff --git a/skills/ir_generation_skill/step1_semantic_index.py b/skills/ir_generation_skill/step1_semantic_index.py
index 86d5657..650e0b5 100644
--- a/skills/ir_generation_skill/step1_semantic_index.py
+++ b/skills/ir_generation_skill/step1_semantic_index.py
@@ -535,6 +535,20 @@ def _quick_validate(
     return passed, gaps
 
 
+def _build_coverage_feedback(gaps: dict) -> str:
+    """Generate feedback text for re-prompting when coverage is below threshold."""
+    lines = []
+    for item in gaps.get("missing_paths", []):
+        lines.append(f"- {item}")
+    if lines:
+        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)
+            + "\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。"
+        )
+    return ""
+
+
 def _collect_logic_tree_nodes(doc: dict) -> dict[str, dict[str, str]]:
     """Return {image_id: {node_id: node_type}} for all logic trees."""
     result = {}
@@ -750,6 +764,29 @@ def run_ensemble_semantic_index(doc: dict) -> dict:
             if v:
                 print(f"    {k}: {len(v)} 个问题")
 
+        # 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...")
+            try:
+                retry_prompt = build_prompt(doc, feedback, all_paths)
+                retry_result = call_llm(retry_prompt, max_retries=1, temperature=0.0)
+                n_retry_units = len(retry_result.get("function_units", []))
+                print(f"  重试返回: {n_retry_units} 功能单元")
+                if n_retry_units > 0:
+                    # Merge retry into results and re-validate
+                    semantic_indices.append(retry_result)
+                    merged = ensemble_merge(semantic_indices)
+                    merged["ensemble_temperatures"] = list(temperatures) + ["feedback_retry"]
+                    passed, gaps = _quick_validate(merged, doc, all_paths)
+                    merged["validation_passed"] = passed
+                    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'}")
+            except Exception as e:
+                print(f"  覆盖反馈重试失败: {e}")
+
     return merged
 
 
diff --git a/skills/ir_generation_skill/step2_ir_extraction.py b/skills/ir_generation_skill/step2_ir_extraction.py
index 688f441..7678d1e 100644
--- a/skills/ir_generation_skill/step2_ir_extraction.py
+++ b/skills/ir_generation_skill/step2_ir_extraction.py
@@ -497,6 +497,13 @@ def main():
     print(f"\n[2/3] 逐单元提取 IR 规则...")
     fragments = extract_all_rules(semantic_index, doc)
 
+    # Filter out fragments with empty rules (LLM extraction failures)
+    empty_units = [f["unit_id"] for f in fragments
+                   if not f.get("rules") and not f.get("error")]
+    if empty_units:
+        print(f"  [WARN] {len(empty_units)} 个单元规则为空，已过滤: {empty_units}")
+    fragments = [f for f in fragments if f.get("rules") or f.get("error")]
+
     # 3. Save
     print(f"\n[3/3] 保存 IR 片段...")
     config.save_json(fragments, config.IR_FRAGMENTS_JSON)
diff --git a/skills/ir_generation_skill/step3_merge_and_audit.py b/skills/ir_generation_skill/step3_merge_and_audit.py
index 7bc3d07..4401d16 100644
--- a/skills/ir_generation_skill/step3_merge_and_audit.py
+++ b/skills/ir_generation_skill/step3_merge_and_audit.py
@@ -139,6 +139,10 @@ def _normalize_rule(rule: dict) -> dict:
 
     trigger = rule["trigger"]
 
+    # Ensure trigger-level combining operator (AND/OR) for multi-condition triggers
+    if not trigger.get("operator"):
+        trigger["operator"] = "AND"
+
     # If trigger has an event, it's event-based (no conditions needed)
     if trigger.get("event") is not None:
         return rule
@@ -147,7 +151,7 @@ def _normalize_rule(rule: dict) -> dict:
     if "conditions" not in trigger:
         trigger["conditions"] = []
 
-    # Fix null operators in conditions
+    # Fix null operators in individual conditions
     for cond in trigger["conditions"]:
         if not cond.get("operator"):
             cond["operator"] = "=="
@@ -158,7 +162,6 @@ def _normalize_rule(rule: dict) -> dict:
 
     # If still no conditions, add a default one
     if not trigger["conditions"]:
-        trigger["operator"] = "AND"
         trigger["conditions"] = [{
             "signal": "system_state",
             "operator": "==",