x

README.md

@@ -170,7 +170,7 @@ Once the server is up, hit the OpenAI‑compatible endpoint:
 ```bash
 curl -X POST http://localhost:8000/v1/chat/completions \
   -H "Content-Type: application/json" \
-  -d '{"model":"Qwen/Qwen2.5-7B-Instruct","messages":[{"role":"user","content":"Hello! Test the performance."}]}'
+  -d '{"model":"Qwen3-Coder-30B-A3B-Instruct-GPTQ-4bit","messages":[{"role":"user","content":"Hello! Test the performance."}]}'
 ```
 
 You should receive a JSON response with a `choices[0].message.content` reply.
@@ -206,4 +206,3 @@ docker run -p 3000:3000 \
   -v chat-ui-data:/data \
   ghcr.io/huggingface/chat-ui-db
 ```
-

config.json

@@ -35,6 +35,9 @@
       "max_num_seqs": "32",
       "max_tokens": "65536",
       "gpu_util": "0.95",
+      "tool_call_parser": "qwen3_xml",
+      "enable_auto_tool_choice": false,
+      "served_model_name": "Qwen3-Coder-30B-A3B-Instruct-GPTQ-4bit",
       "hf_model_id": "cpatonn/Qwen3-Coder-30B-A3B-Instruct-GPTQ-4bit"
     },
     "Qwen3-Next-80B-A3B-Instruct-AWQ-4bit": {
@@ -47,6 +50,8 @@
       "enforce_eager": false,
       "env": {"VLLM_USE_TRITON_AWQ": "1"},
       "tool_call_parser": "qwen3_xml",
+      "enable_auto_tool_choice": false,
+      "served_model_name": "Qwen3-Next-80B-A3B-Instruct-AWQ-4bit",
       "hf_model_id": "cpatonn/Qwen3-Next-80B-A3B-Instruct-AWQ-4bit"
     },
     "gemma-3-27b-it-FP8-dynamic": {

scripts/start_vllm.py

@@ -128,12 +128,14 @@ def launch_model(model_id, config, model_path, gpu_count):
     ctx = int(config.get("ctx", 8192))
     max_seqs = int(config.get("max_num_seqs", 64))
     gpu_util = float(config.get("gpu_util", 0.98))
+    served_model_name = config.get("served_model_name", model_id)
     
     log(f"Config: TP={tp_size}, Ctx={ctx}, Seqs={max_seqs}, Util={gpu_util}")
     
     # Build command
     cmd = [
         "vllm", "serve", model_path,
+        "--served-model-name", served_model_name,
         "--host", HOST,
         "--port", PORT,
         "--tensor-parallel-size", str(tp_size),
@@ -159,9 +161,16 @@ def launch_model(model_id, config, model_path, gpu_count):
     tool_call_parser = config.get("tool_call_parser")
     if tool_call_parser:
         cmd.extend(["--tool-call-parser", tool_call_parser])
-        cmd.extend(["--enable-auto-tool-choice"])
+        openclaw_compat = os.getenv("OPENCLAW_COMPAT", "false").lower() == "true"
+        enable_auto_tool_choice = config.get("enable_auto_tool_choice")
+        if enable_auto_tool_choice is None:
+            enable_auto_tool_choice = not openclaw_compat
+        if enable_auto_tool_choice:
+            cmd.extend(["--enable-auto-tool-choice"])
+            log("Added auto tool choice enabled")
+        else:
+            log("Auto tool choice disabled")
         log(f"Added tool call parser: {tool_call_parser}")
-        log("Added auto tool choice enabled")
     
     log(f"Command: {' '.join(cmd)}")
     

test.py

@@ -7,7 +7,7 @@ client = OpenAI(
 
 # 聊天完成
 response = client.chat.completions.create(
-    model="/opt/model/Qwen3-Next-80B-A3B-Instruct-AWQ-4bit",
+    model="Qwen3-Next-80B-A3B-Instruct-AWQ-4bit",
     messages=[
         {"role": "system", "content": "You are a helpful assistant."},
         {"role": "user", "content": "你好，请介绍一下你自己"}
@@ -16,4 +16,4 @@ response = client.chat.completions.create(
     temperature=0.7
 )
 
-print(response.choices[0].message.content)
+print(response.choices[0].message.content)