agg_vision.sh

#!/bin/bash
# SPDX-FileCopyrightText: Copyright (c) 2025-2026 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
# SPDX-License-Identifier: Apache-2.0
#
# Aggregated multimodal (vision + LLM) serving.
# GPUs: 1

set -e
trap 'echo Cleaning up...; kill 0' EXIT

SCRIPT_DIR="$(dirname "$(readlink -f "$0")")"
source "$SCRIPT_DIR/../../../common/launch_utils.sh"

# Default values
MODEL="Qwen/Qwen3-VL-8B-Instruct"
CHAT_TEMPLATE=""
ENABLE_OTEL=false

# Parse command line arguments
EXTRA_ARGS=()
while [[ $# -gt 0 ]]; do
    case $1 in
        --model-path)
            MODEL="$2"
            shift 2
            ;;
        --chat-template)
            CHAT_TEMPLATE="$2"
            shift 2
            ;;
        --enable-otel)
            ENABLE_OTEL=true
            shift
            ;;
        -h|--help)
            echo "Usage: $0 [OPTIONS]"
            echo "Options:"
            echo "  --model-path <name>      Specify model (default: $MODEL)"
            echo "  --chat-template <name>   Specify SGLang chat template (default: $CHAT_TEMPLATE)"
            echo "  --enable-otel            Enable OpenTelemetry tracing"
            echo "  -h, --help               Show this help message"
            echo ""
            echo "Additional SGLang/Dynamo flags can be passed and will be forwarded"
            echo "Note: System metrics are enabled by default on port 8081 (worker)"
            exit 0
            ;;
        *)
            EXTRA_ARGS+=("$1")
            shift
            ;;
    esac
done

# Enable tracing if requested
TRACE_ARGS=()
if [ "$ENABLE_OTEL" = true ]; then
    export DYN_LOGGING_JSONL=true
    export OTEL_EXPORT_ENABLED=1
    export OTEL_EXPORTER_OTLP_TRACES_ENDPOINT=${OTEL_EXPORTER_OTLP_TRACES_ENDPOINT:-http://localhost:4317}
    TRACE_ARGS+=(--enable-trace --otlp-traces-endpoint localhost:4317)
fi

HTTP_PORT="${DYN_HTTP_PORT:-8000}"
print_launch_banner --multimodal "Launching Aggregated Vision Worker" "$MODEL" "$HTTP_PORT"

# run ingress
# dynamo.frontend accepts either --http-port flag or DYN_HTTP_PORT env var (defaults to 8000)
OTEL_SERVICE_NAME=dynamo-frontend \
python3 -m dynamo.frontend &

# Build chat template args (only if explicitly set)
TEMPLATE_ARGS=()
if [ -n "$CHAT_TEMPLATE" ]; then
    TEMPLATE_ARGS+=(--chat-template "$CHAT_TEMPLATE")
fi

# run worker with vision model (SGLang auto-detects chat template from HF tokenizer)
OTEL_SERVICE_NAME=dynamo-worker DYN_SYSTEM_PORT=${DYN_SYSTEM_PORT:-8081} \
python3 -m dynamo.sglang \
  --model-path "$MODEL" \
  --served-model-name "$MODEL" \
  "${TEMPLATE_ARGS[@]}" \
  --page-size 16 \
  --tp 1 \
  --trust-remote-code \
  --skip-tokenizer-init \
  --enable-metrics \
  "${TRACE_ARGS[@]}" \
  "${EXTRA_ARGS[@]}" &

# Exit on first worker failure; kill 0 in the EXIT trap tears down the rest
wait_any_exit