feat(regulatory): 对齐附件4目录核查规则

review_agent/regulatory_review/workflow.py

@@ -1,5 +1,6 @@
 from __future__ import annotations
 
+import json
 import logging
 from pathlib import Path
 from threading import Thread
@@ -26,6 +27,7 @@ from review_agent.regulatory_review.services.structure_check import run_structur
 from review_agent.regulatory_review.services.text_extract import extract_text
 
 from .events import record_event
+from .storage import save_artifact
 
 
 NODE_DEFINITIONS = [
@@ -105,6 +107,7 @@ class RegulatoryWorkflowExecutor:
         self.rule_set: dict | None = None
         self.findings = []
         self.document_texts: dict[str, str] = {}
+        self.text_extract_status: dict[str, dict[str, object]] = {}
 
     def run(self) -> None:
         self.batch.status = RegulatoryReviewBatch.Status.RUNNING
@@ -176,6 +179,13 @@ class RegulatoryWorkflowExecutor:
             return
         if node_code == "text_extract":
             self.document_texts = self._extract_source_texts()
+            save_artifact(
+                self.batch,
+                name="text_extract_status.json",
+                artifact_type="json",
+                content=json.dumps(self.text_extract_status, ensure_ascii=False, indent=2),
+                metadata={"artifact": "text_extract_status"},
+            )
             return
         if node_code == "structure_check":
             self.findings.extend(run_structure_check(self.document_texts, self._rules()))
@@ -184,7 +194,29 @@ class RegulatoryWorkflowExecutor:
             self.findings.extend(run_consistency_check(self.document_texts))
             return
         if node_code == "risk_assess":
-            persist_findings(self.batch, self.findings)
+            issues = persist_findings(self.batch, self.findings)
+            save_artifact(
+                self.batch,
+                name="rag_result_json.json",
+                artifact_type="json",
+                content=json.dumps(
+                    {
+                        "batch_no": self.batch.batch_no,
+                        "text_extract_status": self.text_extract_status,
+                        "issues": [
+                            {
+                                "rule_code": issue.rule_code,
+                                "title": issue.title,
+                                "citations": issue.citations,
+                            }
+                            for issue in issues
+                        ],
+                    },
+                    ensure_ascii=False,
+                    indent=2,
+                ),
+                metadata={"artifact": "rag_result_json"},
+            )
             return
         if node_code == "report_export":
             exports = export_review_results(self.batch)
@@ -234,8 +266,25 @@ class RegulatoryWorkflowExecutor:
             if not path.is_absolute():
                 path = Path(settings.MEDIA_ROOT) / item.storage_path
             if not path.exists():
+                self.text_extract_status[item.file_name] = {
+                    "status": "missing",
+                    "path": str(path),
+                    "content_hash": "",
+                    "section_candidates": [],
+                    "field_candidates": {},
+                    "front_text": "",
+                }
                 continue
             result = extract_text(path)
+            self.text_extract_status[item.file_name] = {
+                "status": result.status,
+                "path": str(path),
+                "content_hash": result.content_hash,
+                "section_candidates": result.section_candidates,
+                "field_candidates": result.field_candidates,
+                "front_text": result.front_text,
+                "error_message": result.error_message,
+            }
             if result.status == "success" and result.text:
                 texts[item.file_name] = result.text
         return texts