Reindex

Reindex means re-running the ingestion pipeline (e.g., parse, chunk, embed, upsert to vector index) on documents that are already in the system. Reindexing is useful when you change chunking strategy, embedding model, or other settings so that existing chunks in the vector database are replaced with new ones that match the new config.

When to Reindex

• Chunking change — You updated chunk size, overlap, or strategy; old chunks no longer match.
• Embedding change — You switched embedding model or vector config; vectors must be regenerated.
• Schema or metadata — You changed how metadata or filters are set and want the index to reflect it.

Reindexing can be scoped to a knowledge base or document set. It may take time for large corpora; jobs or webhooks can track progress.

Related Concepts

Reindex repeats ingestion for existing documents. It updates the vector index and knowledge base search. Chunking and embedding settings drive what gets reindexed.