fix: Handle model not found error for multimodal example (#1545)

examples/multimodal/components/frontend.py

@@ -18,8 +18,9 @@
 
 from components.processor import Processor
 from fastapi import FastAPI
-from fastapi.responses import StreamingResponse
+from fastapi.responses import JSONResponse, StreamingResponse
 from utils.protocol import MultiModalRequest
+from utils.vllm import parse_vllm_args
 
 from dynamo.sdk import DYNAMO_IMAGE, api, depends, service
 
@@ -38,8 +39,18 @@
 class Frontend:
     processor = depends(Processor)
 
+    def __init__(self):
+        class_name = self.__class__.__name__
+        self.engine_args = parse_vllm_args(class_name, "")
+
     @api(name="v1/chat/completions")
     async def generate(self, request: MultiModalRequest):
+        if self.engine_args.model != request.model:
+            return JSONResponse(
+                {"error": f"Model '{request.model}' not found"},
+                status_code=404,
+            )
+
         async def content_generator():
             async for response in self.processor.generate(request.model_dump_json()):
                 try:

examples/multimodal/components/video_frontend.py

@@ -18,8 +18,9 @@
 
 from components.video_processor import Processor
 from fastapi import FastAPI
-from fastapi.responses import StreamingResponse
+from fastapi.responses import JSONResponse, StreamingResponse
 from utils.protocol import MultiModalRequest
+from utils.vllm import parse_vllm_args
 
 from dynamo.sdk import DYNAMO_IMAGE, api, depends, service
 
@@ -38,8 +39,18 @@
 class Frontend:
     processor = depends(Processor)
 
+    def __init__(self):
+        class_name = self.__class__.__name__
+        self.engine_args = parse_vllm_args(class_name, "")
+
     @api(name="v1/chat/completions")
     async def generate(self, request: MultiModalRequest):
+        if self.engine_args.model != request.model:
+            return JSONResponse(
+                {"error": f"Model '{request.model}' not found"},
+                status_code=404,
+            )
+
         async def content_generator():
             async for response in self.processor.generate(request.model_dump_json()):
                 try:

examples/multimodal/configs/agg-llava.yaml

@@ -17,6 +17,9 @@ Common:
   block-size: 64
   max-model-len: 4096
 
+Frontend:
+  common-configs: [model]
+
 Processor:
   router: round-robin
   prompt-template: "USER: <image>\n<prompt> ASSISTANT:"

examples/multimodal/configs/agg-phi3v.yaml

@@ -18,6 +18,9 @@ Common:
   max-model-len: 4096
   trust-remote-code: true
 
+Frontend:
+  common-configs: [model]
+
 Processor:
   router: round-robin
   prompt-template: "<|user|>\n<|image_1|>\n<prompt><|end|>\n<|assistant|>\n"

examples/multimodal/configs/agg-qwen.yaml

@@ -17,6 +17,9 @@ Common:
   block-size: 64
   max-model-len: 4096
 
+Frontend:
+  common-configs: [model]
+
 Processor:
   router: round-robin
   prompt-template: "<|im_start|>system\nYou are a helpful assistant.<|im_end|>\n<|im_start|>user\n<|vision_start|><|image_pad|><|vision_end|><prompt><|im_end|>\n<|im_start|>assistant\n"

examples/multimodal/configs/agg_video.yaml

@@ -24,6 +24,9 @@ Common:
   video-token-id: 32000
   dummy-tokens-per-frame: 144
 
+Frontend:
+  common-configs: [model]
+
 Processor:
   router: round-robin
   common-configs: [model, block-size, max-model-len]

examples/multimodal/configs/disagg.yaml

@@ -20,6 +20,9 @@ Common:
   num-patches: 576
   kv-transfer-config: '{"kv_connector":"DynamoNixlConnector"}'
 
+Frontend:
+  common-configs: [model]
+
 Processor:
   router: round-robin
   prompt-template: "USER: <image>\n<prompt> ASSISTANT:"

examples/multimodal/configs/disagg_video.yaml

@@ -25,6 +25,9 @@ Common:
   video-token-id: 32000
   dummy-tokens-per-frame: 144
 
+Frontend:
+  common-configs: [model]
+
 Processor:
   router: round-robin
   common-configs: [model, block-size]