We currently lack sufficient integration/end-to-end testing for our AI Assistants. Running tests against a real LLM is expensive in terms of money and time, so it is unrealistic that we will include these test suites as part of every PR commit. The cadence and triggers for these tests are outside the scope of this issue.
Our first step is to provide a FTR suite which exercises AI Assistant functionality. Once we have this running locally, we can explore options for how to securely run these in an automated fashion.
We currently lack sufficient integration/end-to-end testing for our AI Assistants. Running tests against a real LLM is expensive in terms of money and time, so it is unrealistic that we will include these test suites as part of every PR commit. The cadence and triggers for these tests are outside the scope of this issue.
Our first step is to provide a FTR suite which exercises AI Assistant functionality. Once we have this running locally, we can explore options for how to securely run these in an automated fashion.