feat: improve workflow observability and lighten admin frontend

backend-fastapi/ai_platform/services/workflow_service.py

@@ -226,6 +226,7 @@ def _make_execution_log_entry(
     **extra: Any,
 ) -> dict:
     extra.setdefault('timestamp', datetime.now().isoformat())
+    events = extra.pop('events', None)
     metadata = dict(extra.pop('metadata', {}) or {})
     event_type = _execution_log_event_type(extra.get('status'))
     collaboration = _node_collaboration_metadata(node_map, node_id, node_type)
@@ -249,19 +250,106 @@ def _make_execution_log_entry(
         error=extra.get('error'),
     )
     metadata['communication']['timestamp'] = extra.get('timestamp')
-    return {
+    entry = {
         'node_id': node_id,
         'node_type': node_type,
         'node_label': _node_label_from_map(node_map, node_id),
         'metadata': metadata,
         **extra,
     }
+    if events:
+        entry['events'] = events
+    return entry
+
+
+def _event_log_status(event_type: str, event: Dict[str, Any]) -> str:
+    if event_type == 'error':
+        return 'failed'
+    if event_type == 'waiting_input':
+        return 'waiting'
+    if event_type in ('start', 'node_start', 'parallel_start'):
+        return 'running'
+    status = event.get('status') or 'completed'
+    if status == 'success':
+        return 'completed'
+    if status == 'failed':
+        return 'failed'
+    return status
+
+
+def _make_stream_event_log_entry(event: Dict[str, Any]) -> Optional[dict]:
+    event_type = event.get('type')
+    if not event_type or event_type == 'llm_chunk':
+        return None
+
+    event_data = copy.deepcopy(event)
+    timestamp = event_data.get('timestamp') or datetime.now().isoformat()
+    event_data.setdefault('timestamp', timestamp)
+    communication = dict(event_data.get('communication') or {})
+    communication.setdefault('event', event_type)
+    communication.setdefault('timestamp', timestamp)
+
+    node_id = str(event_data.get('node_id') or event_data.get('run_id') or 'workflow')
+    node_type = str(event_data.get('node_type') or 'workflow')
+    node_label = (
+        event_data.get('node_label')
+        or event_data.get('workflow_name')
+        or communication.get('message')
+        or event_type
+    )
+    output = event_data.get('outputs')
+    if output is None:
+        output = event_data.get('content') or event_data.get('message')
+    error = event_data.get('error_message') or event_data.get('error')
+
+    return {
+        'type': event_type,
+        'node_id': node_id,
+        'node_type': node_type,
+        'node_label': str(node_label),
+        'status': _event_log_status(event_type, event_data),
+        'timestamp': timestamp,
+        'elapsed_time': event_data.get('elapsed_time'),
+        'tokens_used': event_data.get('tokens_used') or 0,
+        'output': output,
+        'error': error,
+        'event': event_data,
+        'events': [event_data],
+        'metadata': {
+            'stream_event': True,
+            'collaboration': event_data.get('collaboration') or {},
+            'communication': communication,
+        },
+    }
+
+
+def _log_timestamp(log: Dict[str, Any]) -> str:
+    metadata = log.get('metadata') or {}
+    communication = metadata.get('communication') or {}
+    return log.get('timestamp') or communication.get('timestamp') or ''
+
+
+def _merge_execution_logs(node_logs: List[dict], event_logs: List[dict]) -> List[dict]:
+    base_logs = [
+        log for log in (node_logs or [])
+        if not (log.get('metadata') or {}).get('stream_event')
+    ]
+    merged = base_logs + list(event_logs or [])
+    return sorted(
+        merged,
+        key=lambda item: (_log_timestamp(item), 0 if item.get('metadata', {}).get('stream_event') else 1),
+    )
 
 
 def _node_result_metadata(result: NodeResult) -> dict:
     return dict(result.metadata or {})
 
 
+def _node_result_events(result: NodeResult) -> List[Dict[str, Any]]:
+    events = getattr(result, 'events', None) or []
+    return copy.deepcopy(events) if events else []
+
+
 def _event_collaboration_metadata(
     node_map: Dict[str, Any],
     node_id: str,
@@ -842,6 +930,7 @@ class AIWorkflowService:
                 elapsed_time=elapsed,
                 tokens_used=result.tokens_used,
                 metadata=_node_result_metadata(result),
+                events=_node_result_events(result),
                 inputs=copy.deepcopy(node_inputs),
                 target_node_id=target_node_id,
             ))
@@ -1024,6 +1113,7 @@ class AIWorkflowService:
                     elapsed_time=elapsed,
                     tokens_used=result.tokens_used,
                     metadata=_node_result_metadata(result),
+                    events=_node_result_events(result),
                     inputs=copy.deepcopy(node_inputs),
                     branch=branch_id,
                     branch_label=branch_labels.get(branch_id, branch_id),
@@ -1252,6 +1342,7 @@ class AIWorkflowService:
                     elapsed_time=elapsed,
                     tokens_used=result.tokens_used,
                     metadata=_node_result_metadata(result),
+                    events=_node_result_events(result),
                     branch=branch_id,
                     branch_label=branch_labels.get(branch_id, branch_id),
                 ))
@@ -1545,6 +1636,7 @@ class AIWorkflowService:
                     elapsed_time=elapsed,
                     tokens_used=result.tokens_used,
                     metadata=_node_result_metadata(result),
+                    events=_node_result_events(result),
                     loop_iteration=iteration,
                 ))
                 
@@ -1593,7 +1685,10 @@ class AIWorkflowService:
                         current_id,
                         node_type,
                         status='waiting',
+                        output=result.output,
                         elapsed_time=elapsed,
+                        metadata=_node_result_metadata(result),
+                        events=_node_result_events(result),
                         loop_iteration=iteration,
                     ))
                     
@@ -1733,15 +1828,20 @@ class AIWorkflowService:
         )
         self._db.add(run)
         await self._db.flush()
+        event_logs: List[dict] = []
         
         try:
             # 使用异步生成器执行工作流
             async for event in self._execute_workflow_stream(workflow, inputs, run, conversation_history, use_draft):
                 yield event
+                event_log = _make_stream_event_log_entry(event)
+                if event_log:
+                    event_logs.append(event_log)
+                    run.execution_log = _merge_execution_logs(run.execution_log or [], event_logs)
                 
                 # 在关键事件后提交数据库状态
                 event_type = event.get('type')
-                if event_type in ('node_complete', 'waiting_input', 'complete', 'error'):
+                if event_log or event_type in ('node_complete', 'waiting_input', 'complete', 'error'):
                     await self._db.commit()
             
             # 更新运行记录
@@ -1771,13 +1871,19 @@ class AIWorkflowService:
             await self._db.commit()
             
             # 发送错误事件给前端
-            yield {
+            error_event = {
                 'type': 'error',
                 'message': str(e),
                 'error_message': str(e),
                 'run_id': run.id,
                 'workflow_id': workflow.id,
             }
+            error_log = _make_stream_event_log_entry(error_event)
+            if error_log:
+                event_logs.append(error_log)
+                run.execution_log = _merge_execution_logs(run.execution_log or [], event_logs)
+            await self._db.commit()
+            yield error_event
     
     async def _execute_workflow_stream(
         self,
@@ -2026,6 +2132,7 @@ class AIWorkflowService:
                     output=result.output,
                     elapsed_time=elapsed,
                     metadata=_node_result_metadata(result),
+                    events=_node_result_events(result),
                     inputs=copy.deepcopy(node_inputs),
                     target_node_id=target_node_id,
                 )
@@ -2067,6 +2174,7 @@ class AIWorkflowService:
                 elapsed_time=elapsed,
                 tokens_used=result.tokens_used,
                 metadata=_node_result_metadata(result),
+                events=_node_result_events(result),
                 inputs=copy.deepcopy(node_inputs),
                 target_node_id=target_node_id,
             )
@@ -2557,6 +2665,7 @@ class AIWorkflowService:
                     output=result.output,
                     elapsed_time=elapsed,
                     metadata=_node_result_metadata(result),
+                    events=_node_result_events(result),
                     target_node_id=target_node_id,
                 )
                 logs.append(log_entry)
@@ -2595,6 +2704,7 @@ class AIWorkflowService:
                 elapsed_time=elapsed,
                 tokens_used=result.tokens_used,
                 metadata=_node_result_metadata(result),
+                events=_node_result_events(result),
                 target_node_id=target_node_id,
             )
             logs.append(log_entry)