feat(models): support reasoning_content streaming

api/app/controllers/ontology_controller.py

@@ -163,6 +163,7 @@ def _get_ontology_service(
             api_key=api_key_config.api_key,
             base_url=api_key_config.api_base,
             is_omni=api_key_config.is_omni,
+            support_thinking="thinking" in (api_key_config.capability or []),
             max_retries=3,
             timeout=60.0
         )

api/app/controllers/public_share_controller.py

@@ -453,6 +453,9 @@ async def chat(
         # 流式返回
         agent_config = agent_config_4_app_release(release)
 
+        if not (agent_config.model_parameters.get("deep_thinking", False) and payload.thinking):
+            agent_config.model_parameters["deep_thinking"] = False
+
         if payload.stream:
             async def event_generator():
                 async for event in app_chat_service.agnet_chat_stream(
@@ -634,7 +637,8 @@ async def config_query(
             "app_type": release.app.type,
             "variables": release.config.get("variables"),
             "memory": release.config.get("memory", {}).get("enabled"),
-            "features": release.config.get("features")
+            "features": release.config.get("features"),
+            "model_parameters": release.config.get("model_parameters")
         }
     elif release.app.type == AppType.MULTI_AGENT:
         content = {

api/app/controllers/service/app_api_controller.py

@@ -144,6 +144,11 @@ async def chat(
         # print(app.current_release.default_model_config_id)
         agent_config = agent_config_4_app_release(app.current_release)
         # print(agent_config.default_model_config_id)
+
+        # thinking 开关：仅当 agent 配置了 deep_thinking 且请求 thinking=True 时才启用
+        if not (agent_config.model_parameters.get("deep_thinking", False) and payload.thinking):
+            agent_config.model_parameters["deep_thinking"] = False
+
         # 流式返回
         if payload.stream:
             async def event_generator():