docs: add consistent startup banners to SGLang and TRT-LLM launch scripts (#7252)

Signed-off-by: Keiven Chang <keivenchang@users.noreply.github.com>

examples/backends/sglang/launch/agg.sh

@@ -55,7 +55,7 @@ if [ "$ENABLE_OTEL" = true ]; then
 fi
 
 HTTP_PORT="${DYN_HTTP_PORT:-8000}"
-print_launch_banner "Launching Aggregated LLM Worker" "$MODEL" "$HTTP_PORT"
+print_launch_banner "Launching Aggregated Serving" "$MODEL" "$HTTP_PORT"
 
 # run ingress
 # dynamo.frontend accepts either --http-port flag or DYN_HTTP_PORT env var (defaults to 8000)

examples/backends/sglang/launch/agg_router.sh

@@ -53,7 +53,7 @@ fi
 
 MODEL="Qwen/Qwen3-0.6B"
 HTTP_PORT="${DYN_HTTP_PORT:-8000}"
-print_launch_banner "Launching Aggregated Router (2 workers)" "$MODEL" "$HTTP_PORT"
+print_launch_banner "Launching Aggregated + KV Routing (2 GPUs)" "$MODEL" "$HTTP_PORT"
 
 # run ingress
 # dynamo.frontend accepts either --http-port flag or DYN_HTTP_PORT env var (defaults to 8000)

examples/backends/sglang/launch/agg_vision.sh

@@ -61,7 +61,7 @@ if [ "$ENABLE_OTEL" = true ]; then
 fi
 
 HTTP_PORT="${DYN_HTTP_PORT:-8000}"
-print_launch_banner --multimodal "Launching Aggregated Vision Worker" "$MODEL" "$HTTP_PORT"
+print_launch_banner --multimodal "Launching Aggregated Multimodal Serving" "$MODEL" "$HTTP_PORT"
 
 # run ingress
 # dynamo.frontend accepts either --http-port flag or DYN_HTTP_PORT env var (defaults to 8000)

examples/backends/sglang/launch/disagg.sh

@@ -47,7 +47,7 @@ fi
 
 MODEL="Qwen/Qwen3-0.6B"
 HTTP_PORT="${DYN_HTTP_PORT:-8000}"
-print_launch_banner "Launching Disaggregated Workers (P/D)" "$MODEL" "$HTTP_PORT"
+print_launch_banner "Launching Disaggregated Serving (2 GPUs)" "$MODEL" "$HTTP_PORT"
 
 # run ingress
 # dynamo.frontend accepts either --http-port flag or DYN_HTTP_PORT env var (defaults to 8000)

examples/backends/sglang/launch/disagg_router.sh

@@ -48,7 +48,7 @@ fi
 
 MODEL="Qwen/Qwen3-0.6B"
 HTTP_PORT="${DYN_HTTP_PORT:-8000}"
-print_launch_banner "Launching Disaggregated Router (2P + 2D)" "$MODEL" "$HTTP_PORT"
+print_launch_banner "Launching Disaggregated + KV Routing (4 GPUs)" "$MODEL" "$HTTP_PORT"
 
 # Start frontend with KV routing
 # The frontend will automatically detect prefill workers and activate an internal prefill router

examples/backends/sglang/launch/disagg_same_gpu.sh

@@ -41,7 +41,7 @@ fi
 source "$SCRIPT_DIR/../../../common/launch_utils.sh"
 
 HTTP_PORT="${DYN_HTTP_PORT:-8000}"
-print_launch_banner "Launching Disaggregated (same GPU)" "$MODEL" "$HTTP_PORT" \
+print_launch_banner "Launching Disaggregated on Same GPU" "$MODEL" "$HTTP_PORT" \
     "Context len: $CONTEXT_LENGTH" \
     "GPU Mem:     ${GPU_MEM_FRACTION} per worker (~${_EW_TOTAL_GIB} GiB each)" \
     "  estimate:  weights=${_EW_WEIGHTS_GIB} + kv=${_EW_KV_GIB} + overhead=${_EW_OVERHEAD_GIB} GiB"

examples/backends/sglang/launch/multimodal_disagg.sh

@@ -60,7 +60,7 @@ if [[ -n "$SERVED_MODEL_NAME" ]]; then
 fi
 
 HTTP_PORT="${DYN_HTTP_PORT:-8000}"
-print_launch_banner --multimodal "Launching Multimodal E/P/D Workers" "$MODEL_NAME" "$HTTP_PORT"
+print_launch_banner --multimodal "Launching Disaggregated Multimodal E/P/D (3 GPUs)" "$MODEL_NAME" "$HTTP_PORT"
 
 # run ingress
 # dynamo.frontend accepts either --http-port flag or DYN_HTTP_PORT env var (defaults to 8000)

examples/backends/sglang/launch/multimodal_epd.sh

@@ -60,7 +60,7 @@ if [[ -n "$SERVED_MODEL_NAME" ]]; then
 fi
 
 HTTP_PORT="${DYN_HTTP_PORT:-8000}"
-print_launch_banner --multimodal "Launching Multimodal E/PD Workers" "$MODEL_NAME" "$HTTP_PORT"
+print_launch_banner --multimodal "Launching Multimodal E/PD (2 GPUs)" "$MODEL_NAME" "$HTTP_PORT"
 
 # run ingress
 # dynamo.frontend accepts either --http-port flag or DYN_HTTP_PORT env var (defaults to 8000)

examples/backends/trtllm/launch/agg.sh

@@ -50,6 +50,8 @@ if [ "$ENABLE_OTEL" = true ]; then
     TRACE_ARGS+=(--override-engine-args "{\"return_perf_metrics\": true, \"otlp_traces_endpoint\": \"${OTEL_EXPORTER_OTLP_TRACES_ENDPOINT}\" }")
 fi
 
+HTTP_PORT="${DYN_HTTP_PORT:-8000}"
+print_launch_banner "Launching Aggregated Serving" "$MODEL_PATH" "$HTTP_PORT"
 
 # run frontend
 # dynamo.frontend accepts either --http-port flag or DYN_HTTP_PORT env var (defaults to 8000)

examples/backends/trtllm/launch/agg_metrics.sh

@@ -15,6 +15,9 @@ export SERVED_MODEL_NAME=${SERVED_MODEL_NAME:-"Qwen/Qwen3-0.6B"}
 export AGG_ENGINE_ARGS=${AGG_ENGINE_ARGS:-"$DYNAMO_HOME/examples/backends/trtllm/engine_configs/qwen3/agg.yaml"}
 export MODALITY=${MODALITY:-"text"}
 
+HTTP_PORT="${DYN_HTTP_PORT:-8000}"
+print_launch_banner "Launching Aggregated Serving + Metrics" "$MODEL_PATH" "$HTTP_PORT"
+
 # Run frontend
 # dynamo.frontend accepts either --http-port flag or DYN_HTTP_PORT env var (defaults to 8000)
 python3 -m dynamo.frontend &

examples/backends/trtllm/launch/agg_multimodal.sh

@@ -15,6 +15,9 @@ export SERVED_MODEL_NAME=${SERVED_MODEL_NAME:-"Qwen/Qwen2-VL-7B-Instruct"}
 export AGG_ENGINE_ARGS=${AGG_ENGINE_ARGS:-"$DYNAMO_HOME/examples/backends/trtllm/engine_configs/qwen2-vl-7b-instruct/agg.yaml"}
 export MODALITY=${MODALITY:-"multimodal"}
 
+HTTP_PORT="${DYN_HTTP_PORT:-8000}"
+print_launch_banner --multimodal "Launching Aggregated Multimodal Serving" "$MODEL_PATH" "$HTTP_PORT"
+
 # run frontend
 # dynamo.frontend accepts either --http-port flag or DYN_HTTP_PORT env var (defaults to 8000)
 python3 -m dynamo.frontend --router-mode kv &

examples/backends/trtllm/launch/agg_router.sh

@@ -14,6 +14,9 @@ export MODEL_PATH=${MODEL_PATH:-"Qwen/Qwen3-0.6B"}
 export SERVED_MODEL_NAME=${SERVED_MODEL_NAME:-"Qwen/Qwen3-0.6B"}
 export AGG_ENGINE_ARGS=${AGG_ENGINE_ARGS:-"$DYNAMO_HOME/examples/backends/trtllm/engine_configs/qwen3/agg.yaml"}
 
+HTTP_PORT="${DYN_HTTP_PORT:-8000}"
+print_launch_banner "Launching Aggregated + KV Routing" "$MODEL_PATH" "$HTTP_PORT"
+
 # run frontend
 # dynamo.frontend accepts either --http-port flag or DYN_HTTP_PORT env var (defaults to 8000)
 python3 -m dynamo.frontend --router-mode kv &

examples/backends/trtllm/launch/agg_router_approx.sh

@@ -14,6 +14,9 @@ export MODEL_PATH=${MODEL_PATH:-"Qwen/Qwen3-0.6B"}
 export SERVED_MODEL_NAME=${SERVED_MODEL_NAME:-"Qwen/Qwen3-0.6B"}
 export AGG_ENGINE_ARGS=${AGG_ENGINE_ARGS:-"$DYNAMO_HOME/examples/backends/trtllm/engine_configs/qwen3/agg.yaml"}
 
+HTTP_PORT="${DYN_HTTP_PORT:-8000}"
+print_launch_banner "Launching Aggregated + Approximate KV Routing" "$MODEL_PATH" "$HTTP_PORT"
+
 # run frontend with KV router in approximate mode (i.e. no KV events)
 python3 -m dynamo.frontend --router-mode kv --no-kv-events &
 

examples/backends/trtllm/launch/disagg.sh

@@ -52,6 +52,9 @@ if [ "$ENABLE_OTEL" = true ]; then
     TRACE_ARGS+=(--override-engine-args "{\"return_perf_metrics\": true, \"otlp_traces_endpoint\": \"${OTEL_EXPORTER_OTLP_TRACES_ENDPOINT}\" }")
 fi
 
+HTTP_PORT="${DYN_HTTP_PORT:-8000}"
+print_launch_banner "Launching Disaggregated Serving (2 GPUs)" "$MODEL_PATH" "$HTTP_PORT"
+
 # run frontend
 # dynamo.frontend accepts either --http-port flag or DYN_HTTP_PORT env var (defaults to 8000)
 OTEL_SERVICE_NAME=dynamo-frontend \

examples/backends/trtllm/launch/disagg_e_pd.sh

@@ -27,6 +27,9 @@ export MAX_FILE_SIZE_MB=${MAX_FILE_SIZE_MB:-50}
 # Extra arguments forwarded to the PD worker (e.g. --multimodal-embedding-cache-capacity-gb 10)
 EXTRA_PD_ARGS=("$@")
 
+HTTP_PORT="${DYN_HTTP_PORT:-8000}"
+print_launch_banner --multimodal "Launching Multimodal E/PD" "$MODEL_PATH" "$HTTP_PORT"
+
 # run frontend
 # dynamo.frontend accepts either --http-port flag or DYN_HTTP_PORT env var (defaults to 8000)
 python3 -m dynamo.frontend &

examples/backends/trtllm/launch/disagg_multimodal.sh

@@ -18,6 +18,9 @@ export PREFILL_CUDA_VISIBLE_DEVICES=${PREFILL_CUDA_VISIBLE_DEVICES:-"0"}
 export DECODE_CUDA_VISIBLE_DEVICES=${DECODE_CUDA_VISIBLE_DEVICES:-"1"}
 export MODALITY=${MODALITY:-"multimodal"}
 
+HTTP_PORT="${DYN_HTTP_PORT:-8000}"
+print_launch_banner --multimodal "Launching Disaggregated Multimodal Serving (2 GPUs)" "$MODEL_PATH" "$HTTP_PORT"
+
 # run frontend
 # dynamo.frontend accepts either --http-port flag or DYN_HTTP_PORT env var (defaults to 8000)
 python3 -m dynamo.frontend &

examples/backends/trtllm/launch/disagg_router.sh

@@ -17,6 +17,9 @@ export DECODE_ENGINE_ARGS=${DECODE_ENGINE_ARGS:-"$DYNAMO_HOME/examples/backends/
 export PREFILL_CUDA_VISIBLE_DEVICES=${PREFILL_CUDA_VISIBLE_DEVICES:-"0"}
 export DECODE_CUDA_VISIBLE_DEVICES=${DECODE_CUDA_VISIBLE_DEVICES:-"1"}
 
+HTTP_PORT="${DYN_HTTP_PORT:-8000}"
+print_launch_banner "Launching Disaggregated + KV Routing (2 GPUs)" "$MODEL_PATH" "$HTTP_PORT"
+
 # run frontend with KV routing for cache-aware optimization
 # dynamo.frontend accepts either --http-port flag or DYN_HTTP_PORT env var (defaults to 8000)
 python3 -m dynamo.frontend --router-mode kv &

examples/backends/trtllm/launch/epd_disagg.sh

@@ -23,6 +23,9 @@ export MODALITY=${MODALITY:-"multimodal"}
 export ALLOWED_LOCAL_MEDIA_PATH=${ALLOWED_LOCAL_MEDIA_PATH:-"/tmp"}
 export MAX_FILE_SIZE_MB=${MAX_FILE_SIZE_MB:-50}
 
+HTTP_PORT="${DYN_HTTP_PORT:-8000}"
+print_launch_banner --multimodal "Launching Disaggregated Multimodal E/P/D (8 GPUs)" "$MODEL_PATH" "$HTTP_PORT"
+
 # run frontend
 # dynamo.frontend accepts either --http-port flag or DYN_HTTP_PORT env var (defaults to 8000)
 python3 -m dynamo.frontend &

examples/backends/trtllm/launch/epd_multimodal_image_and_embeddings.sh

@@ -23,6 +23,9 @@ export MODALITY=${MODALITY:-"multimodal"}
 export ALLOWED_LOCAL_MEDIA_PATH=${ALLOWED_LOCAL_MEDIA_PATH:-"/tmp"}
 export MAX_FILE_SIZE_MB=${MAX_FILE_SIZE_MB:-50}
 
+HTTP_PORT="${DYN_HTTP_PORT:-8000}"
+print_launch_banner --multimodal "Launching Multimodal E/P/D" "$MODEL_PATH" "$HTTP_PORT"
+
 # run frontend
 # dynamo.frontend accepts either --http-port flag or DYN_HTTP_PORT env var (defaults to 8000)
 python3 -m dynamo.frontend &

examples/backends/trtllm/launch/gpt_oss_disagg.sh

@@ -15,6 +15,9 @@ export SERVED_MODEL_NAME=${SERVED_MODEL_NAME:-"openai/gpt-oss-120b"}
 export PREFILL_ENGINE_ARGS=${PREFILL_ENGINE_ARGS:-"$DYNAMO_HOME/examples/backends/trtllm/engine_configs/gpt-oss-120b/prefill.yaml"}
 export DECODE_ENGINE_ARGS=${DECODE_ENGINE_ARGS:-"$DYNAMO_HOME/examples/backends/trtllm/engine_configs/gpt-oss-120b/decode.yaml"}
 
+HTTP_PORT="${DYN_HTTP_PORT:-8000}"
+print_launch_banner "Launching Disaggregated Serving (8 GPUs)" "$SERVED_MODEL_NAME" "$HTTP_PORT"
+
 # run frontend
 # dynamo.frontend accepts either --http-port flag or DYN_HTTP_PORT env var (defaults to 8000)
 python3 -m dynamo.frontend --router-mode round-robin &