Quickstart

$
# Install uv (recommended Python package manager)
$
curl -LsSf https://astral.sh/uv/install.sh | sh
$
$
# Create virtual environment and install Dynamo
$
uv venv venv
$
source venv/bin/activate
$
# Use prerelease flag to install RC versions of flashinfer and/or other dependencies
$
uv pip install --prerelease=allow "ai-dynamo[sglang]"  # or [vllm], [trtllm]

$
# Fetch and start etcd and NATS using Docker Compose
$
VERSION=$(uv pip show ai-dynamo | grep Version | cut -d' ' -f2)
$
curl -fsSL -o docker-compose.yml https://raw.githubusercontent.com/ai-dynamo/dynamo/refs/tags/v${VERSION}/deploy/docker-compose.yml
$
docker compose -f docker-compose.yml up -d

$
# Start the OpenAI compatible frontend (default port is 8000)
$
python -m dynamo.frontend
$
$
# In another terminal, start an SGLang worker
$
python -m dynamo.sglang --model-path Qwen/Qwen3-0.6B

$
curl localhost:8000/v1/chat/completions \
>
  -H "Content-Type: application/json" \
>
  -d '{"model": "Qwen/Qwen3-0.6B",
>
       "messages": [{"role": "user", "content": "Hello!"}],
>
       "max_tokens": 50}'