In documentation that contains multiple similar code examples (e.g. showing before/after or different configurations), SemanticDedup sometimes merges them as duplicates even though they differ in important ways.
Example: two Python snippets that differ only in one function argument get deduped, but that argument is the whole point of the example.
Maybe code blocks inside markdown should use a stricter similarity threshold?
In documentation that contains multiple similar code examples (e.g. showing before/after or different configurations), SemanticDedup sometimes merges them as duplicates even though they differ in important ways.
Example: two Python snippets that differ only in one function argument get deduped, but that argument is the whole point of the example.
Maybe code blocks inside markdown should use a stricter similarity threshold?