Jupyter Kernel Gateway

3.0.1 · active · verified Sat Apr 11

Jupyter Kernel Gateway is a web server for spawning and communicating with Jupyter kernels. It provides a REST API that allows execution of code in Jupyter Kernels and manages their lifecycle, often used to expose notebooks as RESTful APIs. The current version is 3.0.1, and the project maintains an active release cadence with several updates per year, now managed under the `jupyter-server` organization.

Warnings

breaking Starting from v2.5.2, `jupyter-kernel-gateway` removed its internal dependency on `ipython_genutils`. Custom extensions or deployment scripts that relied on this module might break upon upgrade.
Fix: Review custom code and remove any dependencies on `ipython_genutils` modules. Migrate to equivalent functionality in `traitlets` or standard Python libraries.
gotcha Versions of `jupyter-kernel-gateway` prior to 3.0.0 were pinned to `jupyter-client < 8`. Users with a newer `jupyter-client` version in their environment might encounter dependency conflicts or unexpected behavior.
Fix: Upgrade `jupyter-kernel-gateway` to version 3.0.0 or later, which has updated dependency requirements. Alternatively, ensure your `jupyter-client` version is compatible (e.g., `<8`) when using older `jupyter-kernel-gateway` versions.
gotcha When deploying `jupyter-kernel-gateway` as a public-facing service, critical security configuration like CORS origins (`--KernelGatewayApp.allow_origin`) and authentication must be explicitly set. Default settings are not secure for production environments.
Fix: Always review and configure `KernelGatewayApp` settings, especially `ip`, `port`, `allow_origin`, and `auth_token` or other authentication mechanisms suitable for your deployment scenario. Consult the official documentation for comprehensive security guidelines.

Install

pip install jupyter-kernel-gateway Install stable version

Quickstart

This quickstart demonstrates how to programmatically start the Jupyter Kernel Gateway server and make a simple API request to verify its operation. In typical use, Kernel Gateway is run directly from the command line using `jupyter kernelgateway`.

import subprocess
import sys
import time
import requests

# Start Kernel Gateway in a separate process
print("Starting Jupyter Kernel Gateway on port 9999...")
process = subprocess.Popen(
    [sys.executable, "-m", "jupyter_kernel_gateway.app",
     "--KernelGatewayApp.port=9999",
     "--KernelGatewayApp.ip=127.0.0.1",
     "--KernelGatewayApp.allow_origin='*'"],
    stdout=subprocess.PIPE, stderr=subprocess.PIPE, text=True
)

# Give it a moment to start up
time.sleep(5)

try:
    # Verify it's running by hitting an endpoint
    print("Attempting to connect to Kernel Gateway...")
    response = requests.get("http://127.0.0.1:9999/api/kernelspecs")
    response.raise_for_status()
    print("Kernel Gateway is running and accessible. Available kernelspecs:")
    print(response.json())
except Exception as e:
    print(f"Error connecting to Kernel Gateway: {e}")
    if process.poll() is not None:
        stdout, stderr = process.communicate()
        print("Server stdout:\n", stdout)
        print("Server stderr:\n", stderr)
finally:
    print("Stopping Kernel Gateway...")
    if process.poll() is None: # Check if still running
        process.terminate()
        process.wait(timeout=5)
    print("Kernel Gateway stopped.")

view raw JSON →