agg_embed.sh

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

# Setup cleanup trap
cleanup() {
    echo "Cleaning up background processes..."
    kill $DYNAMO_PID 2>/dev/null || true
    wait $DYNAMO_PID 2>/dev/null || true
    echo "Cleanup complete."
}
trap cleanup EXIT INT TERM

# Parse command line arguments
while [[ $# -gt 0 ]]; do
    case $1 in
        -h|--help)
            echo "Usage: $0 [OPTIONS]"
            echo "Options:"
            echo "  -h, --help           Show this help message"
            echo ""
            echo "Note: System metrics are enabled by default on port 8081 (worker)"
            echo "Note: OpenTelemetry tracing is not yet supported for embedding models"
            exit 0
            ;;
        *)
            echo "Unknown option: $1"
            echo "Use --help for usage information"
            exit 1
            ;;
    esac
done

MODEL="Qwen/Qwen3-Embedding-4B"
HTTP_PORT="${DYN_HTTP_PORT:-8000}"
echo "=========================================="
echo "Launching Embedding Worker"
echo "=========================================="
echo "Model:       $MODEL"
echo "Frontend:    http://localhost:$HTTP_PORT"
echo "=========================================="
echo ""
echo "Example test command:"
echo ""
echo "  curl http://localhost:${HTTP_PORT}/v1/embeddings \\"
echo "    -H 'Content-Type: application/json' \\"
echo "    -d '{"
echo "      \"model\": \"${MODEL}\","
echo "      \"input\": \"Explain why Roger Federer is considered one of the greatest tennis players of all time\""
echo "    }'"
echo ""
echo "=========================================="

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

# run worker
DYN_SYSTEM_PORT=${DYN_SYSTEM_PORT:-8081} \
python3 -m dynamo.sglang \
  --embedding-worker \
  --model-path Qwen/Qwen3-Embedding-4B \
  --served-model-name Qwen/Qwen3-Embedding-4B \
  --page-size 16 \
  --tp 1 \
  --trust-remote-code \
  --use-sglang-tokenizer \
  --enable-metrics