Implement a function to sanitize the commit diffs before feeding them to the LLM.
Currently, we only ignore the diff for a poetry.lock file - those diffs tend to be large and fairly useless (the corresponding changes to pyproject.toml are far more concise and already tell us everything about what was changed or updated in terms of dependencies.)
But that's just one example. We should think of other commonly large and useless diffs to ignore (e.g. Pipfile.lock, binary files, other large assets... what else?)
May need to use a combination of a blacklist and heuristics to decide what to ignore.
Implement a function to sanitize the commit diffs before feeding them to the LLM.
Currently, we only ignore the diff for a
poetry.lockfile - those diffs tend to be large and fairly useless (the corresponding changes to pyproject.toml are far more concise and already tell us everything about what was changed or updated in terms of dependencies.)But that's just one example. We should think of other commonly large and useless diffs to ignore (e.g. Pipfile.lock, binary files, other large assets... what else?)
May need to use a combination of a blacklist and heuristics to decide what to ignore.