fix(bridge): project_id透传补全——proxy_chat_completion委托chat_inference带token绑定的project_id;bug_flow的PM确认调用带project_id落流水(配套pipeline-llm 427cd7c,2026-09-10遗留)

pipeline_service/bug_flow.py

@@ -378,7 +378,7 @@ async def pm_bug_confirm_run(sor, project_id):
     try:
         raw = await llm_call_msgs(
             [{"role": "system", "content": system}, {"role": "user", "content": user}],
-            model=model, temperature=0.2, org_id=org_id)
+            model=model, temperature=0.2, org_id=org_id, project_id=project_id or '')
     except Exception as e:
         logger.warning(f"pm_bug_confirm_run LLM 失败 project={project_id}: {e}")
         return {"confirmed": 0, "rejected": 0, "skipped": len(bugs)}

pipeline_service/llm_proxy.py

@@ -282,7 +282,8 @@ async def proxy_chat_completion(token, payload):
         data = await chat_inference(
             org_id, info.get('created_by', '') or '', payload,
             model_name=model_name,
-            task_ref='proxy:%s' % (info.get('project_id') or ''))
+            task_ref='proxy:%s' % (info.get('project_id') or ''),
+            project_id=info.get('project_id') or '')
     except ImportError:
         return False, "模型治理模块（pipeline-llm）未安装，推理引擎不可用"
     except Exception as e: