The requirements to be installed in ventis build are currently being declared in the stub_generator.py file, which only works for the current workflows, and will fail any workflows that try to call a library not currently included.
Decoupling this would also open up the possibility of speeding up docker build as redundant installs won't have to happen, though a cache currently fixes that.
Example snippet of the code:
requirements = (
"grpcio\ngrpcio-tools\nredis\npyyaml\nflask\nboto3\nyfinance\npsutil\nipdb\nipython\n"
"sqlalchemy\npsycopg[binary]\n"
)
Possible Fixes:
Put the imports of each agent in the agents .yaml files.
Could also be in the global_controller.yaml, leaning towards this as to reduce dependecy on the agents.yaml to eventually phase them out.
The requirements to be installed in ventis build are currently being declared in the stub_generator.py file, which only works for the current workflows, and will fail any workflows that try to call a library not currently included.
Decoupling this would also open up the possibility of speeding up docker build as redundant installs won't have to happen, though a cache currently fixes that.
Example snippet of the code:
requirements = (
"grpcio\ngrpcio-tools\nredis\npyyaml\nflask\nboto3\nyfinance\npsutil\nipdb\nipython\n"
"sqlalchemy\npsycopg[binary]\n"
)
Possible Fixes:
Put the imports of each agent in the agents .yaml files.
Could also be in the global_controller.yaml, leaning towards this as to reduce dependecy on the agents.yaml to eventually phase them out.