vector-db-cleanup

To remove orphaned chunks from ChromaDB, run the provided Python cleanup scripts with a specific profile parameter like --profile wiki, which loads the correct vector space and prunes entries pointing to non-existent source files.

Vector search returns stale results because the index retains orphaned chunks after file deletations or directory restructuring. Pruning the vector store removes these dangling entries and synchronizes the search index with the current filesystem.

The best way to clean up a vector store after a major refactor is to execute a pruning script that checks for source file existence and safely removes stale chunks, keeping the search index accurate.

You keep your vector search index synchronized with the filesystem by running periodic maintenance that prunes orphaned chunks, ensuring the vector store only contains entries for existing source files.

Yes, you need to specify a profile parameter such as --profile wiki to load the correct vector space before the cleanup script can identify and remove the orphaned chunks safely.