Increase model wait time for vLLM

workers/openai/README.md

@@ -34,30 +34,12 @@ uv pip install -r requirements.txt
 
 Several examples have been provided in the client to help you get started with your own implementation.
 
-First, set your API key as an environment variable:
+### Completions
+
+Call to `/v1/completions` with json response
 
 ```bash
-export VAST_API_KEY=<your_api_key>
-```
-
-The `--model` and `--endpoint` flags are optional. If not provided, they default to `Qwen/Qwen3-8B` and `my-vllm-endpoint` respectively.
-
-### Chat Completion (streaming)
-
-Call to `/v1/chat/completions` with streaming response
-
-```bash
-python -m workers.openai.client --chat-stream --endpoint <ENDPOINT_NAME> --model <MODEL_NAME>
-```
-
-### Interactive Chat (streaming)
-
-Interactive session with calls to `/v1/chat/completions`.
-
-Type `clear` to clear the chat history or `quit` to exit.
-
-```bash
-python -m workers.openai.client --interactive --endpoint <ENDPOINT_NAME> --model <MODEL_NAME>
+python -m workers.openai.client -k <API_KEY> -e <ENDPOINT_NAME> --completion --model <MODEL_NAME>
 ```
 
 ### Chat Completion (json)
@@ -65,7 +47,15 @@ python -m workers.openai.client --interactive --endpoint <ENDPOINT_NAME> --model
 Call to `/v1/chat/completions` with json response
 
 ```bash
-python -m workers.openai.client --chat --endpoint <ENDPOINT_NAME> --model <MODEL_NAME>
+python -m workers.openai.client -k <API_KEY> -e <ENDPOINT_NAME> --chat --model <MODEL_NAME>
+```
+
+### Chat Completion (streaming)
+
+Call to `/v1/chat/completions` with streaming response
+
+```bash
+python -m workers.openai.client -k <API_KEY> -e <ENDPOINT_NAME> --chat-stream --model <MODEL_NAME>
 ```
 
 ### Tool Use (json)
@@ -75,14 +65,16 @@ Call to `/v1/chat/completions` with tool and json response.
 This test defines a simple tool which will list the contents of the local pyworker directory.  The output is then analysed by the model.
 
 ```bash
-python -m workers.openai.client --tools --endpoint <ENDPOINT_NAME> --model <MODEL_NAME>
+python -m workers.openai.client -k <API_KEY> -e <ENDPOINT_NAME> --tools --model <MODEL_NAME>
 ```
 
-### Completions
+### Interactive Chat (streaming)
 
-Call to `/v1/completions` with json response
+Interactive session with calls to `/v1/chat/completions`.
+
+Type `clear` to clear the chat history or `quit` to exit.
 
 ```bash
-python -m workers.openai.client --completion --endpoint <ENDPOINT_NAME> --model <MODEL_NAME>
+python -m workers.openai.client -k <API_KEY> -e <ENDPOINT_NAME> --interactive --model <MODEL_NAME>
 ```
 

workers/openai/client.py

@@ -18,7 +18,7 @@ logging.basicConfig(
 log = logging.getLogger(__file__)
 
 # ---------------------- Prompts ----------------------
-COMPLETIONS_PROMPT = "Zebras are primarily grazers and can subsist on lower-quality vegetation. They are preyed on mainly by"
+COMPLETIONS_PROMPT = "the capital of USA is"
 CHAT_PROMPT = "Think step by step: Tell me about the Python programming language."
 TOOLS_PROMPT = (
     "Can you list the files in the current working directory and tell me what you see? "
@@ -97,9 +97,9 @@ def _tool_state_to_message_tool_calls(state: Dict[int, Dict[str, Any]]) -> List[
 
 
 # ---- OpenAI-compatible calls (non-streaming) ----
-async def call_completions(client: Serverless, *, model: str, prompt: str, endpoint_name: str, **kwargs) -> Dict[str, Any]:
+async def call_completions(client: Serverless, *, model: str, prompt: str, **kwargs) -> Dict[str, Any]:
 
-    endpoint = await client.get_endpoint(name=endpoint_name)
+    endpoint = await client.get_endpoint(name=ENDPOINT_NAME)
 
     payload = {
         "input": {
@@ -113,9 +113,9 @@ async def call_completions(client: Serverless, *, model: str, prompt: str, endpo
     resp = await endpoint.request("/v1/completions", payload, cost=payload["input"]["max_tokens"])
     return resp["response"]
 
-async def call_chat_completions(client: Serverless, *, model: str, messages: List[Dict[str, Any]], endpoint_name: str, **kwargs) -> Dict[str, Any]:
+async def call_chat_completions(client: Serverless, *, model: str, messages: List[Dict[str, Any]], **kwargs) -> Dict[str, Any]:
 
-    endpoint = await client.get_endpoint(name=endpoint_name)
+    endpoint = await client.get_endpoint(name=ENDPOINT_NAME)
 
     payload = {
         "input": {
@@ -132,9 +132,9 @@ async def call_chat_completions(client: Serverless, *, model: str, messages: Lis
     return resp["response"]
 
 # ---- Streaming variants ----
-async def stream_completions(client: Serverless, *, model: str, prompt: str, endpoint_name: str, **kwargs):
+async def stream_completions(client: Serverless, *, model: str, prompt: str, **kwargs):
 
-    endpoint = await client.get_endpoint(name=endpoint_name)
+    endpoint = await client.get_endpoint(name=ENDPOINT_NAME)
 
     payload = {
         "input": {
@@ -150,9 +150,9 @@ async def stream_completions(client: Serverless, *, model: str, prompt: str, end
     resp = await endpoint.request("/v1/completions", payload, cost=payload["input"]["max_tokens"], stream=True)
     return resp["response"]  # async generator
 
-async def stream_chat_completions(client: Serverless, *, model: str, messages: List[Dict[str, Any]], endpoint_name: str, **kwargs):
+async def stream_chat_completions(client: Serverless, *, model: str, messages: List[Dict[str, Any]], **kwargs):
 
-    endpoint = await client.get_endpoint(name=endpoint_name)
+    endpoint = await client.get_endpoint(name=ENDPOINT_NAME)
 
     payload = {
         "input": {
@@ -174,10 +174,9 @@ async def stream_chat_completions(client: Serverless, *, model: str, messages: L
 class APIDemo:
     """Demo and testing functionality for the API client"""
 
-    def __init__(self, client: Serverless, model: str, endpoint_name: str, tool_manager: Optional[ToolManager] = None):
+    def __init__(self, client: Serverless, model: str, tool_manager: Optional[ToolManager] = None):
         self.client = client
         self.model = model
-        self.endpoint_name = endpoint_name
         self.tool_manager = tool_manager or ToolManager()
 
     # ----- Streaming handler -----
@@ -186,15 +185,10 @@ class APIDemo:
         reasoning_content = ""
         printed_reasoning = False
         printed_answer = False
-        finish_reason = None
 
         async for chunk in stream:
             choice = (chunk.get("choices") or [{}])[0]
             delta = choice.get("delta", {})
-            
-            # Track finish reason
-            if choice.get("finish_reason"):
-                finish_reason = choice.get("finish_reason")
 
             # reasoning tokens
             rc = delta.get("reasoning_content")
@@ -225,8 +219,6 @@ class APIDemo:
                 print(f"Reasoning tokens: {len(reasoning_content.split())}")
             if printed_answer:
                 print(f"Response tokens: {len(full_response.split())}")
-            if finish_reason:
-                print(f"Finish reason: {finish_reason}")
 
         return full_response
     
@@ -239,7 +231,6 @@ class APIDemo:
             client=self.client,
             model=self.model,
             prompt=COMPLETIONS_PROMPT,
-            endpoint_name=self.endpoint_name,
             max_tokens=MAX_TOKENS,
             temperature=DEFAULT_TEMPERATURE,
         )
@@ -258,7 +249,6 @@ class APIDemo:
                 client=self.client,
                 model=self.model, 
                 messages=messages,
-                endpoint_name=self.endpoint_name,
                 max_tokens=MAX_TOKENS,
                 temperature=DEFAULT_TEMPERATURE
             )
@@ -271,7 +261,6 @@ class APIDemo:
                 client=self.client,
                 model=self.model, 
                 messages=messages,
-                endpoint_name=self.endpoint_name,
                 max_tokens=MAX_TOKENS,
                 temperature=DEFAULT_TEMPERATURE
             )
@@ -298,7 +287,6 @@ class APIDemo:
                 client=self.client,
                 model=self.model,
                 messages=messages,
-                endpoint_name=self.endpoint_name,
                 tools=minimal_tool,
                 tool_choice="none",
                 max_tokens=10
@@ -324,7 +312,6 @@ class APIDemo:
             client=self.client,
             model=self.model,
             messages=messages,
-            endpoint_name=self.endpoint_name,
             tools=self.tool_manager.get_ls_tool_definition(),
             tool_choice="auto",
             max_tokens=MAX_TOKENS,
@@ -402,7 +389,6 @@ class APIDemo:
             client=self.client,
             model=self.model,
             messages=messages,
-            endpoint_name=self.endpoint_name,
             max_tokens=MAX_TOKENS,
             temperature=DEFAULT_TEMPERATURE,
         )
@@ -441,6 +427,7 @@ class APIDemo:
         print("=" * 60)
         print("INTERACTIVE STREAMING CHAT")
         print("=" * 60)
+        print(f"Using model: {self.model}")
         print("Type 'quit' to exit, 'clear' to clear history")
         print()
 
@@ -466,8 +453,7 @@ class APIDemo:
                 stream = await stream_chat_completions(
                     client=self.client,
                     model=self.model, 
-                    messages=messages,
-                    endpoint_name=self.endpoint_name,
+                    messages=messages, 
                     max_tokens=MAX_TOKENS, 
                     temperature=0.7
                 )
@@ -487,8 +473,8 @@ class APIDemo:
 # ---------------------- CLI ----------------------
 def build_arg_parser() -> argparse.ArgumentParser:
     p = argparse.ArgumentParser(description="Vast vLLM Demo (Serverless SDK)")
-    p.add_argument("--model", default=DEFAULT_MODEL, help=f"Model to use for requests (default: {DEFAULT_MODEL})")
-    p.add_argument("--endpoint", default=ENDPOINT_NAME, help=f"Vast endpoint name (default: {ENDPOINT_NAME})")
+    p.add_argument("--model", required=True, help="Model to use for requests (required)")
+    p.add_argument("--endpoint", default="my-vllm-endpoint", help="Vast endpoint name (default: my-vllm-endpoint)")
 
     modes = p.add_mutually_exclusive_group(required=False)
     modes.add_argument("--completion", action="store_true", help="Test completions endpoint")
@@ -516,14 +502,12 @@ async def main_async():
         print("Please specify exactly one test mode")
         sys.exit(1)
 
-    print("=" * 60)
     print(f"Using model: {args.model}")
-    print(f"Using endpoint: {args.endpoint}")
-
+    print("=" * 60)
 
     try:
         async with Serverless() as client:
-            demo = APIDemo(client, args.model, args.endpoint, ToolManager())
+            demo = APIDemo(client, args.model, ToolManager())
 
             if args.completion:
                 await demo.demo_completions()

workers/openai/server.py

@@ -35,6 +35,7 @@ backend = Backend(
     model_server_url=os.environ["MODEL_SERVER_URL"],
     model_log_file=os.environ["MODEL_LOG"],
     allow_parallel_requests=True,
+    max_wait_time=600.0,
     benchmark_handler=CompletionsHandler(benchmark_runs=3, benchmark_words=256),
     log_actions=[
         *[(LogAction.ModelLoaded, info_msg) for info_msg in MODEL_SERVER_START_LOG_MSG],