A Cloud Foundry buildpack for pushing prompt text that result in Python based apps.
This buildpack builds upon the official Python buildpack.
To push an app you simply need a prompt and an access token to the STACKIT Model Serving Service or another OpenAI-compatible API of you choice.
You have the following options to push an app to Cloud Foundry this way:
-
Via the following
ai.envfile:CF_APP_SPEC='A web page that allows to do basic math with two numbers.' STACKIT_MODEL_SERVING_AUTH_TOKEN='ey...'
Pushed with:
cf push mybasicmath -b 'https://github.com/funcf/ai-buildpack.git' -
Via the following
Promptfile:A web page that allows to do basic math with two numbers.
and the following
manifest.yml:--- applications: - name: mybasicmath buildpacks: - https://github.com/funcf/ai-buildpack.git env: STACKIT_MODEL_SERVING_AUTH_TOKEN: ey...
Pushed with:
cf push -
Or via the
Promptfile:A web page that allows to do basic math with two numbers.
and some more commands:
cf create-app mybasicmath -b 'https://github.com/funcf/ai-buildpack.git' cf set-env mybasicmath STACKIT_MODEL_SERVING_AUTH_TOKEN 'ey...' cf push
A) Why not pushing just a manifest.yml with all needed environment variables or just a cf push with -b 'https://github.com/funcf/ai-buildpack.git' and the prompt injected as start command with -c 'I need an app to play tic tac toe in the browser.' (the initial idea btw.)?
The current Cloud Foundry implementation will fail if you push no files other than a manifest.yml (because magic AI buildpacks were not there yet ;-)!
Via app environment variable or a separate ai.env file you can change the following default values:
STACKIT_MODEL_SERVING_BASE_URL='https://api.openai-compat.model-serving.eu01.onstackit.cloud/v1'
STACKIT_MODEL_SERVING_MODEL='google/gemma-3-27b-it'The API Call will be a POST request to: ${STACKIT_MODEL_SERVING_BASE_URL}/chat/completions.
Open a GitHub issue on this project here. Or directly reach out to the maintainer via CF Slack.