feat: glm-4.6v不支持网络搜索。更换为glm-4.5-air

server/utils/glm_adapter.py

@@ -34,7 +34,7 @@ def get_client():
             from zai import ZhipuAiClient
         except ImportError:
             raise ImportError("GLM 模式需要安装 zai-sdk：.venv/bin/pip install zai-sdk")
-        api_key = os.getenv("ZHIPU_API_KEY") or os.getenv("GLM_API_KEY")
+        api_key = os.getenv("ZHIPU_API_KEY").strip()  or os.getenv("GLM_API_KEY").strip() 
         if not api_key:
             raise ValueError("GLM 模式需要设置环境变量 ZHIPU_API_KEY")
         _client = ZhipuAiClient(api_key=api_key)
@@ -43,15 +43,15 @@ def get_client():
 
 
 # ── 模型映射 ──────────────────────────────────────────────────────────
-DEFAULT_TEXT_MODEL   = "glm-4.6v"   # glm-4.6v 文本+视觉统一模型
+DEFAULT_TEXT_MODEL   = "glm-4.5-Air"   # glm-4.6 文本统一模型
-DEFAULT_VISION_MODEL = "glm-4.6v"
+DEFAULT_VISION_MODEL = "glm-4.5-Air"
 
 MODEL_MAP = {
-    "qwen-max":      "glm-4.6v",
+    "qwen-max":      "glm-4.5-Air",
-    "qwen-plus":     "glm-4.6v",
+    "qwen-plus":     "glm-4.5-Air",
-    "qwen-turbo":    "glm-4.6v",
+    "qwen-turbo":    "glm-4.5-Air",
-    "qwen-vl-max":   "glm-4.6v",
+    "qwen-vl-max":   "glm-4.5-Air",
-    "qwen-vl-plus":  "glm-4.6v",
+    "qwen-vl-plus":  "glm-4.5-Air",
 }
 
 
@@ -195,7 +195,7 @@ def build_glm_messages(messages: list, files: list | None = None) -> tuple[list,
 # ── 哨兵对象 ─────────────────────────────────────────────────────────
 _SENTINEL = object()
 
-
+# ── 流式调用 ────────────────────────────────────────────────────────
 async def glm_stream_generator(
     messages: list,
     model: str,
@@ -221,11 +221,11 @@ async def glm_stream_generator(
     extra_kwargs: dict = {}
     if web_search:
         extra_kwargs["tools"] = [
-            {"type": "web_search", "web_search": {"search_result": True}}
+            {"type": "web_search", "web_search": {"enable":True,"search_result": True}}
         ]
-    if deep_thinking:
+    if not deep_thinking:
-        extra_kwargs["thinking"] = {"type": "enabled"}
+        # 智普默认开启思考模式，所以要用非门（不知道“非门”描述是否准确。前端选择开启思考模式，这里不做变动。前端选择关闭思考模式，这里关闭。）
-
+        extra_kwargs["thinking"] = {"type": "disabled"}
     print(f"[GLM] 流式请求：model={actual_model} vision={has_vision} "
           f"web_search={web_search} thinking={deep_thinking}")
 
@@ -334,9 +334,13 @@ def glm_chat_sync(
 
     extra_kwargs: dict = {}
     if web_search:
-        extra_kwargs["tools"] = [
+        extra_kwargs["tools"] = [{
-            {"type": "web_search", "web_search": {"search_result": True}}
+    "type": "web_search",
-        ]
+    "web_search": {
+        "enable": True,
+        "search_result": True
+    }
+}]
     if deep_thinking:
         extra_kwargs["thinking"] = {"type": "enabled"}
 

server/utils/test_glm_search.py

@@ -0,0 +1,31 @@
+import os
+import sys
+import asyncio
+from pathlib import Path
+
+# Add project root to sys.path
+root_dir = Path(__file__).parent
+sys.path.insert(0, str(root_dir))
+
+from utils.glm_adapter import glm_stream_generator, _ensure_venv, glm_chat_sync
+
+# Set API key from .env if needed
+from dotenv import load_dotenv
+load_dotenv()
+
+async def test_stream():
+    msgs = [{"role": "user", "content": "今天北京天气怎样？"}]
+    print("Testing stream...")
+    async for chunk in glm_stream_generator(msgs, "glm-4.5-air", 0.7, 1024, web_search=True):
+        print(chunk, end="")
+
+def test_sync():
+    msgs = [{"role": "user", "content": "今天几号？武汉天气怎样？"}]
+    print("Testing sync...")
+    res = glm_chat_sync(msgs, "glm-4.5-air", 0.7, 1024, web_search=True)
+    print(res)
+
+if __name__ == "__main__":
+    _ensure_venv()
+    # test_sync()
+    asyncio.run(test_stream())