Closed RebelOfDeath closed 4 months ago
I would argue we can do integration/end-to-end tests for the endpoints, which should also test the LLMs. I've already added a test_generation
method that does this for one sample; but ideally we stress-test with a batch of generations as well.
Test the logic for the sessions in 233f189b07466bc8441007244dc338268255e2ca along with improvements bound to the feedback received from the testing phase.
achieved 93% line coverage as of 5a9c7163ec80048cc381f88f155fd2055acc4026 in main.py
As part of #10 I would like to propose to do some rather robust testing before we move on to the next milestone i.e. developing the actual plugins. For this, we still need to test quite a bit of logic in order to ensure the quality of the server and the soundness of the approach taken in terms the decisions made. Below I have included a checklist of the test suites we still have to write.
The Actual endpoints
The Conversion Interface from Application to DB
The Sessions and Session Managers and the Clean-Up Logic
The LLM Side of things
TBD