Describe the Bug

Frontend fails to register workers for the multimodal models. After digging through the code, I found that three multimodal init functions are missing the register_llm_with_readiness_gate() call that all the text workers have.

File: components/src/dynamo/sglang/main.py

Missing registration in:

init_multimodal_worker() (line 401)
init_multimodal_encode_worker() (line 360)
init_multimodal_prefill_worker() (line 441)

Have registration (working fine):

init_worker() (line 167)
init_prefill() (line 221)
init_embedding() (line 289)
Even init_multimodal_processor() has it (line 344)

Steps to Reproduce

`# Start frontend
python -m dynamo.frontend --http-port 8000

Start multimodal worker

python -m dynamo.sglang --model-path Qwen/Qwen2.5-VL-7B-Instruct --multimodal-worker

Check models

curl http://localhost:8000/v1/models

Returns: {"data": []} ← model not registered

Try to use it

curl -X POST http://localhost:8000/v1/chat/completions -H "Content-Type: application/json" -d '{
"model": "Qwen/Qwen2.5-VL-7B-Instruct",
"messages": [{"role": "user", "content": "Hello!"}]
}'

Expected Behavior

• Worker should register with frontend via etcd/NATS discovery
• Frontend logs should show: added model model_name="Qwen/Qwen2.5-VL-7B-Instruct"
• curl /v1/models should return the model

Actual Behavior

• Worker starts but never registers
• Frontend logs show no model addition
• curl /v1/models returns empty list
• Requests return 503 Service Unavailable

Environment

Dynamo: v0.6.1
Backend: SGLang
Hardware: 8x H100

Additional Context

No response

Screenshots

No response