Langgraph api dockerfile with healthcheck

LangGraph API Dockerfile with Stdlib Healthcheck

When to use

You're deploying a LangGraph-based agent to a self-hosted Railway/EC2/ECS environment (not LangSmith Deployments). You need a slim Dockerfile that uses the official LangGraph API server, installs your agent's deps under the API's pinned constraints, and supports a Docker HEALTHCHECK without curl in the base image.

How it works

• FROM langchain/langgraph-api:3.11 provides the API server, port 2024, and /ok health endpoint.
• pip install --no-cache-dir -c /api/constraints.txt /deps/agent pins all transitive deps to versions tested by the LangGraph maintainers.
• LANGSERVE_GRAPHS env wires your agent factory by file:symbol path.
• HEALTHCHECK uses python -c "import urllib.request; urllib.request.urlopen('http://localhost:2024/ok', timeout=5)" — no curl/wget required in the image.

Example

FROM langchain/langgraph-api:3.11

ADD . /deps/agent

RUN PYTHONDONTWRITEBYTECODE=1 \
    pip install --no-cache-dir -c /api/constraints.txt /deps/agent

# Intentionally omit custom auth here; self-hosted LangGraph does not
# support enterprise-only custom auth.
ENV LANGSERVE_GRAPHS='{"agent":"/deps/agent/app/graph_pipeline.py:build_graph"}'

WORKDIR /deps/agent
EXPOSE 2024

HEALTHCHECK --interval=30s --timeout=10s --start-period=60s --retries=3 \
    CMD python -c "import urllib.request; urllib.request.urlopen('http://localhost:2024/ok', timeout=5)" || exit 1

Gotchas

• --start-period=60s is critical — LangGraph's first-time DB migration can take 30+ seconds and the container would otherwise be marked unhealthy and recycled.
• Pin against /api/constraints.txt so a newer pydantic/langchain release doesn't break the API server in your image.
• PYTHONDONTWRITEBYTECODE=1 shaves layer size by ~30% — no __pycache__ directories.
• Bind your LANGSERVE_GRAPHS path to a stable shim module (see langgraph-compat-shim-import-path-preserved) so internal refactors don't require rebuilding the image config.