feat(agent): 增加 LLM 路由与诊断日志

review_agent/file_summary/workflow.py

@@ -1,5 +1,6 @@
 from __future__ import annotations
 
+import logging
 from threading import Thread
 from uuid import uuid4
 
@@ -36,6 +37,9 @@ NODE_DEFINITIONS = [
 ]
 
 
+logger = logging.getLogger("review_agent.file_summary.workflow")
+
+
 def default_skill_registry() -> SkillRegistry:
     registry = SkillRegistry()
     registry.register(ArchiveExtractSkill())
@@ -65,6 +69,14 @@ def create_file_summary_batch(
     )
     if not active_attachments:
         raise ValueError("当前对话没有可用附件。")
+    logger.info(
+        "File summary batch creation started",
+        extra={
+            "conversation_id": conversation.pk,
+            "user_id": user.pk,
+            "attachment_ids": [attachment.pk for attachment in active_attachments],
+        },
+    )
 
     batch = FileSummaryBatch.objects.create(
         conversation=conversation,
@@ -82,6 +94,10 @@ def create_file_summary_batch(
         WorkflowNodeRun.objects.create(batch=batch, node_code=code, node_name=name)
 
     record_event(batch, "workflow_created", {"batch_id": batch.pk, "batch_no": batch.batch_no})
+    logger.info(
+        "File summary batch created",
+        extra={"batch_id": batch.pk, "batch_no": batch.batch_no},
+    )
     return batch
 
 
@@ -91,6 +107,7 @@ class WorkflowExecutor:
         self.registry = registry or default_skill_registry()
 
     def run(self) -> None:
+        logger.info("Workflow run started", extra={"batch_id": self.batch.pk})
         self.batch.status = FileSummaryBatch.Status.RUNNING
         self.batch.started_at = timezone.now()
         self.batch.save(update_fields=["status", "started_at"])
@@ -100,6 +117,10 @@ class WorkflowExecutor:
             for node in self.batch.node_runs.order_by("id"):
                 self._run_node(node)
         except Exception as exc:
+            logger.exception(
+                "Workflow run failed",
+                extra={"batch_id": self.batch.pk, "error": str(exc)},
+            )
             self.batch.status = FileSummaryBatch.Status.FAILED
             self.batch.error_message = str(exc)
             self.batch.finished_at = timezone.now()
@@ -111,8 +132,17 @@ class WorkflowExecutor:
         self.batch.finished_at = timezone.now()
         self.batch.save(update_fields=["status", "finished_at"])
         record_event(self.batch, "workflow_completed", {"batch_id": self.batch.pk})
+        logger.info("Workflow run completed", extra={"batch_id": self.batch.pk})
 
     def _run_node(self, node: WorkflowNodeRun) -> None:
+        logger.info(
+            "Workflow node started",
+            extra={
+                "batch_id": self.batch.pk,
+                "node_code": node.node_code,
+                "node_name": node.node_name,
+            },
+        )
         now = timezone.now()
         node.status = WorkflowNodeRun.Status.RUNNING
         node.progress = 10
@@ -132,6 +162,15 @@ class WorkflowExecutor:
         if skill_name:
             result = self.registry.execute(skill_name, WorkflowContext(batch=self.batch))
             if not result.success:
+                logger.warning(
+                    "Workflow node skill failed",
+                    extra={
+                        "batch_id": self.batch.pk,
+                        "node_code": node.node_code,
+                        "skill_name": skill_name,
+                        "result_message": result.message,
+                    },
+                )
                 raise RuntimeError(result.message or f"{node.node_name}执行失败")
 
         node.status = WorkflowNodeRun.Status.SUCCESS
@@ -144,11 +183,17 @@ class WorkflowExecutor:
             "node_progress",
             {"node_code": node.node_code, "status": node.status, "progress": node.progress},
         )
+        logger.info(
+            "Workflow node finished",
+            extra={"batch_id": self.batch.pk, "node_code": node.node_code},
+        )
 
 
 def start_file_summary_workflow(batch: FileSummaryBatch, *, async_run: bool = True) -> None:
     executor = WorkflowExecutor(batch)
     if not async_run:
+        logger.info("Workflow starting synchronously", extra={"batch_id": batch.pk})
         executor.run()
         return
+    logger.info("Workflow starting asynchronously", extra={"batch_id": batch.pk})
     Thread(target=executor.run, daemon=True).start()