Repository navigation
Modify components to take a config file as input rather than multiple command line inputs #98
Description
This is more scalable for a large number of parameters and will also help us keep a static docker-compose.yml file for components, instead of a jinja-templated docker compose file and/or a docker compose file that reads multiple env variables. We will need to do this for each component.
This is similar to when we changed the experiment infrastructure to have a simple Hydra config file, instead of command line inputs.
We will need to define the input config file format.
Activity
milindsrivastava1997 commentedon Mar 4, 2026
PRO-343
milindsrivastava1997 commentedon Mar 4, 2026
For Python configs, we can use pydantic. We could also re-use the Pydantic type definitions between the specific component and the service in Utilities/experiments that orchestrates it.
For Rust, we can use figment.
milindsrivastava1997 commentedon Mar 4, 2026
I've pushed some code to the related branch but am unsure whether the ROI on this change is worth it or not. The issue is that any component can be orchestrated by either quickstart, asap-cli or Utilities experiment scripts.
Changing the command line interface of any component will require changes to all 3 orchestration methods. Maybe this needs to be broken down into smaller, more incremental issues.
milindsrivastava1997 commentedon May 4, 2026
Stale issue, closing.