feat: `Frontend` component uses served_model_name instead of model (#302)

docs/guides/dynamo_serve.md

@@ -163,7 +163,7 @@ This will print out something like
 Service Configuration:
 {
   "Frontend": {
-    "model": "deepseek-ai/DeepSeek-R1-Distill-Llama-8B",
+    "served_model_name": "deepseek-ai/DeepSeek-R1-Distill-Llama-8B",
     "endpoint": "dynamo.Processor.chat/completions",
     "port": 8000
   },
@@ -189,7 +189,7 @@ Service Configuration:
 }
 
 Environment Variable that would be set:
-DYNAMO_SERVICE_CONFIG={"Frontend": {"model": "deepseek-ai/DeepSeek-R1-Distill-Llama-8B", "endpoint": "dynamo.Processor.chat/completions", "port": 8000}, "Processor": {"model": "deepseek-ai/DeepSeek-R1-Distill-Llama-8B", "block-size": 64,
+DYNAMO_SERVICE_CONFIG={"Frontend": {"served_model_name": "deepseek-ai/DeepSeek-R1-Distill-Llama-8B", "endpoint": "dynamo.Processor.chat/completions", "port": 8000}, "Processor": {"model": "deepseek-ai/DeepSeek-R1-Distill-Llama-8B", "block-size": 64,
 "max-model-len": 16384, "router": "round-robin"}, "VllmWorker": {"model": "deepseek-ai/DeepSeek-R1-Distill-Llama-8B", "enforce-eager": true, "block-size": 64, "max-model-len": 16384, "max-num-batched-tokens": 16384, "enable-prefix-caching":
 true, "router": "random", "tensor-parallel-size": 1, "ServiceArgs": {"workers": 1}}}
 ```

examples/llm/components/frontend.py

@@ -17,7 +17,6 @@ import subprocess
 from pathlib import Path
 
 from components.processor import Processor
-from components.routerless.worker import VllmWorkerRouterLess
 from components.worker import VllmWorker
 from pydantic import BaseModel
 
@@ -37,7 +36,7 @@ def get_http_binary_path():
 
 
 class FrontendConfig(BaseModel):
-    model: str
+    served_model_name: str
     endpoint: str
     port: int = 8080
 
@@ -50,7 +49,6 @@ class FrontendConfig(BaseModel):
 # todo this should be called ApiServer
 class Frontend:
     worker = depends(VllmWorker)
-    worker_routerless = depends(VllmWorkerRouterLess)
     processor = depends(Processor)
 
     def __init__(self):
@@ -58,7 +56,13 @@ class Frontend:
         frontend_config = FrontendConfig(**config.get("Frontend", {}))
 
         subprocess.run(
-            ["llmctl", "http", "remove", "chat-models", frontend_config.model]
+            [
+                "llmctl",
+                "http",
+                "remove",
+                "chat-models",
+                frontend_config.served_model_name,
+            ]
         )
         subprocess.run(
             [
@@ -66,7 +70,7 @@ class Frontend:
                 "http",
                 "add",
                 "chat-models",
-                frontend_config.model,
+                frontend_config.served_model_name,
                 frontend_config.endpoint,
             ]
         )

examples/llm/configs/agg.yaml

@@ -14,7 +14,7 @@
 # limitations under the License.
 
 Frontend:
-  model: deepseek-ai/DeepSeek-R1-Distill-Llama-8B
+  served_model_name: deepseek-ai/DeepSeek-R1-Distill-Llama-8B
   endpoint: dynamo.Processor.chat/completions
   port: 8000
 
@@ -35,3 +35,5 @@ VllmWorker:
   tensor-parallel-size: 1
   ServiceArgs:
     workers: 1
+    resources:
+      gpu: 1

examples/llm/configs/agg_router.yaml

@@ -14,7 +14,7 @@
 # limitations under the License.
 
 Frontend:
-  model: deepseek-ai/DeepSeek-R1-Distill-Llama-8B
+  served_model_name: deepseek-ai/DeepSeek-R1-Distill-Llama-8B
   endpoint: dynamo.Processor.chat/completions
   port: 8000
 
@@ -40,3 +40,5 @@ VllmWorker:
   tensor-parallel-size: 1
   ServiceArgs:
     workers: 1
+    resources:
+      gpu: 1

examples/llm/configs/disagg.yaml

@@ -15,7 +15,7 @@
 
 
 Frontend:
-  model: deepseek-ai/DeepSeek-R1-Distill-Llama-8B
+  served_model_name: deepseek-ai/DeepSeek-R1-Distill-Llama-8B
   endpoint: dynamo.Processor.chat/completions
   port: 8000
 
@@ -24,19 +24,23 @@ Processor:
   router: round-robin
 
 VllmWorker:
-  # vllm enging args
   model: deepseek-ai/DeepSeek-R1-Distill-Llama-8B
   kv-transfer-config: '{"kv_connector":"DynamoNixlConnector"}'
   max-model-len: 16384
-
-  # dynamo args
   remote-prefill: true
   conditional-disagg: true
   max-local-prefill-length: 10
+  ServiceArgs:
+    workers: 1
+    resources:
+      gpu: 1
 
 PrefillWorker:
-  # vllm enging args
   model: deepseek-ai/DeepSeek-R1-Distill-Llama-8B
   kv-transfer-config: '{"kv_connector":"DynamoNixlConnector"}'
   max-model-len: 16384
   max-num-batched-tokens: 16384
+  ServiceArgs:
+    workers: 1
+    resources:
+      gpu: 1

examples/llm/configs/disagg_router.yaml

@@ -14,7 +14,7 @@
 # limitations under the License.
 
 Frontend:
-  model: deepseek-ai/DeepSeek-R1-Distill-Llama-8B
+  served_model_name: deepseek-ai/DeepSeek-R1-Distill-Llama-8B
   endpoint: dynamo.Processor.chat/completions
   port: 8000
 
@@ -48,3 +48,7 @@ PrefillWorker:
   block-size: 64
   max-model-len: 16384
   max-num-batched-tokens: 16384
+  ServiceArgs:
+    workers: 1
+    resources:
+      gpu: 1