Manticore Search adds auto-chunking so long docs don't silently lose content in vector search

Better Vector Search for Long Documents: Chunking Inside Manticore Search

Manticore Search now supports auto-chunking inside the table definition—set chunk_strategy on a vector column and it splits long docs, embeds each chunk, and searches them all. On a 189-page manual, recall@5 for content beyond the model window jumped from 55% to 83%, at roughly 2.5× RAM and 4× ingest time. Queries are never chunked; only stored documents are split.