Prompt2Service

Prompt2Service is an experiment to let LLMs produce valid configuration files to achieve a working cloud service. In its initial version, it asks to generate a Kubernetes deployment in YAML
format for the Ningx web server. The generated files undergo check to proceed from stage 0 (invalid YAML) to stage 3 (deployment succeeded).

The experiment requires the native installation of ollama and microk8s. The file experiment.sh specifies which models need to be available in ollama.

This package contains the experiment files as well as reference data. This reference was produced on the following machine:
- 32 cores VM (à 2.6 GHz Broadwell CPU, virtualised on OpenStack)
- 64 GB RAM
- A30 GPU with 24 GB memory
- Linux 6.8.0
- ollama 0.9.6
- models as in statistics/ollama.csv

The experiment was supported with a grant by the Swiss State Secretariate for Education, Research and Innovation (SERI) in the CLOUDSTARS Cloud Open Source Research Mobility Network.