A runtime for python-based Knowledge Objects
KGrid Python Runtime
A KGrid runtime for Knowledge Objects in a native python environment that connects to an activator using the proxy adapter.
Installation from the python package:
python -m pip install kgrid-python-runtimeto download the latest package
python -m kgrid_python_runtimeto start the runtime
Installation from an image:
Download the latest image from docker hub:
docker pull kgrid/pythonruntime:#.#.#where
#.#.#is the latest version
Use the following command to run the image on Linux:
sudo docker run --network host kgrid/pythonruntime
- Use the following command to run the image on Windows:
docker run -it -p :5000:5000 -e KGRID_PROXY_ADAPTER_URL=http://host.docker.internal:8080 kgrid/pythonruntime
This starts the runtime pointed to an activator running on the same system at localhost:8080
The runtime exposes two endpoints which can be used to see the details of the runtime and what has been activated
Displays details about the runtime such as the running version and status.
Displays a list of the activated endpoints in the engine.
Set these environment variables to customize your runtime's settings.
- The address of this environment that the activator will use to communicate with it.
- Default value:
- The port this environment is available on.
- Default value:
The url of the adapter this runtime will communicate with
By default, the python runtime will tell the Kgrid Activator that it is started at
- The caching strategy of this runtime. It can take three values:
never- existing objects from the activator are overwritten on every activation call.
always- existing objects stored in the runtime will never be re-downloaded from the activator and the local pyshelf and context.json files must be deleted and the runtime restarted for the objects to be replaced.
use_checksum- objects will look for a checksum in the deployment descriptor sent over during activation and only re-download the object if that checksum has changed.
- Default value:
- The frequency (in seconds) at which the runtime will ping the activator and attempt to reconnect if the connection has been broken. Can be set to any value above 5 seconds or 0 to disable the heartbeat.
- Default value:
- Changes the logging level to debug, takes a boolean
Creating a python Knowledge Object:
Just like other knowledge objects, python objects have 4 basic parts: service.yaml, deployment.yaml, metadata.txt, and a payload that can be any number of python files.
The packaging spec for knowledge objects can be found here.
An example KO with naan of
hello, a name of
neighbor, api version of
1.0, and endpoint
a Deployment Specification might look like this:
/welcome: post: artifact: - "src/hello.js" engine: "python" function: "main" entry: "src/hello.js"
You would then execute this endpoint to see the code work:
POST <activator url>/<naan>/<name>/<api version>/<endpoint>
In this example:
POST <activator url>/hello/neighbor/1.0/welcome
The Service Specification for this object would likewise then be
openapi: 3.0.2 info: version: '1.0' title: 'Hello neighbor' description: An example of simple Knowledge Object license: name: GNU General Public License v3 (GPL-3) url: >- https://tldrlegal.com/license/gnu-general-public-license-v3-(gpl-3)#fulltext contact: name: KGrid Team email: email@example.com url: 'http://kgrid.org' servers: - url: /js/neighbor description: Hello world tags: - name: KO Endpoints description: Hello world Endpoints paths: /welcome: post: ...
In the Service Specification the servers.url must match the naan and name of the object (
/js/neighbor) and the path must match the path in Deployment Specification (
The service spec conforms to the swagger OpenAPI spec.
If your python package requires other python packages,
simply specify them in a file called
at the root of your object:
package-name=0.1.5 other-package-name=1.3.5 third-package-name=1.5.4
That's it! as long as the payload is written in valid python, and the object is built to the spec, you're ready to go. An example python object can be found in the example collection: python/simple/v1.0
To run the app:
Clone this project and set the environment variable:
PYTHONPATH to the project root.
python kgrid_python_runtime/app.py runserver from the top level of the project.
Build the image of kgrid-python-runtime
Use the following command to build the image:
sudo docker build -t kgrid/pythonruntime .
Push to DockerHub
Use the following command to push the image:
sudo docker push kgrid/pythonruntime:#.#.#
Pushing an image directly to Heroku
Log in to Heroku by
Log in to Heroku Container Registry by
Tag the image with (change #.#.# to version)
docker tag <image> registry.heroku.com/<app>/web
Push to Heroku:
docker push registry.heroku.com/<app>/web
Release the image so Heroku can start the deployment process
heroku container:release web -a <app>
Editing the cache directly from the runtime's shelf will not propagate changes to the endpoints in the runtime. New KOs must come from the activator.
The runtime will attempt to load any Knowledge Objects that were previously loaded onto its shelf before registering with the activator and acquiring its objects. The shelf directory can be deleted if there is a need to get all objects fresh from the activator.
Release history Release notifications | RSS feed
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.
|Filename, size||File type||Python version||Upload date||Hashes|
|Filename, size kgrid_python_runtime-0.0.20-py3-none-any.whl (22.1 kB)||File type Wheel||Python version py3||Upload date||Hashes View|
|Filename, size kgrid-python-runtime-0.0.20.tar.gz (24.3 kB)||File type Source||Python version None||Upload date||Hashes View|
Hashes for kgrid_python_runtime-0.0.20-py3-none-any.whl
Hashes for kgrid-python-runtime-0.0.20.tar.gz