Read model configurations dynamically at runtime

[marcobebway]

Description

Currently we use the config.yml file to read the model configurations. We need to come up with another approach to read these configurations at runtime.

Reason(s)

• Remove configuration hardcoding.
• Support for loading dynamic configurations at runtime.
• Support for the containerization of the Companion backend.
• Support for the secret rotation at runtime (even if it is achieved by restarting the Companion deployment).

Acceptance Criteria
TBD

Results from research and team agreements

[friedrichwilken]

Considerations

gen-ai-hub needs either
- A: env vars
- B: a json file, that is located at ~/.aicore/config.json, which can be changed via the env. var AICORE_HOME

Team agreement

• Have one file for running the app that contains all the configurations.
• Load config from file, set env vars for gen-ai-hub doc from config file via code.
• When deploying to k8s, mount secret with all settings as config file.
• Have the config in the json format for easy reading.
• Watch changes in that file and restart app if accordingly.

[grischperl]

Desired final solution:

• all variables moved from .env to config.json
• default location of the config.json
• if desire to change that location, do so by overriding location with an env-var / .env file