test suite

[ltalirz]

experience shows that the docstrings describing the tools are rather brittle, even for a transition between gpt4 and gpt4o #48

we should implement a rudimentary test suite that can check, for a few exemplary use requests, whether (a) the right tools are used, and (b) whether the result is correct

see e.g. https://docs.smith.langchain.com/old/cookbook/testing-examples for inspiration