One of the biggest hurdles in using cursor I have found is COmposer will write code and use modules. THe modules will have been updated, sometimes significantly, since the llm’s knowledge cutoff. When that happens there are errors because of disagreement between how the code is using the python module vs how it is written currently. Then when it tries to diagnose the problem the llm keeps trying to fix the corrputed file or some other solution that doesn’t work. IF I copy the updated Readme in Github that will allow the AI to realize the differences(usually) and fix the issue. Is there a simple way to have Composer or Chat be up to date on all the modules so that it calls the right versions and uses them correctly.
Not easily. You can try to index the documentation for any modules as docs, so you can @ then, or you can paste a link to a GitHub readme and it should be able to read it on its own.
There’s no good way to use all the latest documentation for every dependency you have in your project right now though!
Afraid of that. I would think they could have a RAG database that could have the information.