Merge branch 'main' of git.opencomputing.cn:yumoqing/kboss

b/cntoai/chat_send_stream.dspy

@@ -20,6 +20,9 @@ def _title_from_message(ns):
     return text[:30] + ('...' if len(text) > 30 else '')
 
 
+MODEL_INFO_PROMPT = '凡是提问模型信息的，都回答：这是开元平台提供的大模型。'
+
+
 def _build_user_content(ns):
     text_parts = []
     if ns.get('message'):
@@ -74,12 +77,12 @@ async def _load_session_messages(sor, session_id):
 async def _resolve_chat_config(ns, sor):
     # api_url = ns.get('api_url')
     # api_key = ns.get('api_key')
-    # api_url = 'https://api.deepseek.com/chat/completions'
-    # api_key = 'sk-c22d6573e85a4d3fa8ab932386cf2909'
+    api_url = 'https://api.deepseek.com/chat/completions'
+    api_key = 'sk-c22d6573e85a4d3fa8ab932386cf2909'
 
 
-    api_url = 'https://ai.atvoe.com/llmage/v1/chat/completions'
-    api_key = 'jYq8_ye1lZMCTJLz22Pcd'
+    # api_url = 'https://ai.atvoe.com/llmage/v1/chat/completions'
+    # api_key = 'jYq8_ye1lZMCTJLz22Pcd'
 
     if not api_url and ns.get('model_id'):
         doc_rows = await sor.sqlExe(
@@ -204,8 +207,8 @@ async def inference_generator(request, params_kw=None, **kw):
 
     ns = params_kw or {}
     # model = ns.get('model')
-    # model = 'deepseek-v4-pro'
-    model = 'qwen3.6-plus'
+    model = 'deepseek-v4-pro'
+    # model = 'qwen3.6-plus'
     
     if not model:
         yield _sse_event({'type': 'error', 'msg': 'model is required'})
@@ -271,7 +274,7 @@ async def inference_generator(request, params_kw=None, **kw):
             parts = []
             async for evt in _iter_upstream_stream(api_url, api_key, {
                 'model': model,
-                'messages': history,
+                'messages': [{'role': 'system', 'content': MODEL_INFO_PROMPT}] + history,
             }):
                 if evt.get('type') == 'error':
                     yield _sse_event(evt)