Testing tasks in isolation

CrewAI Community Support
1

I am trying to do some task test in isolation.

That’s to say

determine_schema:
  description: >
    ...
  expected_output: >
    ...
  context:
    - input_task
  agent: query_manager

I want to fake input_task, so that when access its output it gives a value according to the test. e.g., providing an ad-hoc TaskOutput instance

Then instantiate the task under test and its agent according to my agents/tasks.yaml config, injecting the fake, build the crew and kick it off.

Observe the task output (or crew output which is solely depending on the single task).

It seems that there are various approaches to test the system, but I cannot find any to easily test the parts in isolation, without having to reinvent the wheel.

How normally is this done?

I have a high precision/low complexity use case, and this seems a pretty good approach.