We Asked for Socks, Got Pants: Six Ways to Search the Same Product Catalog
We tested full-text, vector, hybrid, image, geo, and AI search on 82,524 products. Here's where they go wrong, what they find, and the SQL behind the results.
Manticore Blog – posts about tech, search database, and open-source
New Percentile and Median Absolute Deviation Aggregations in Manticore
Manticore’s new percentile and MAD aggregations, showing how to analyze tail behavior and outliers in SQL and JSON with tunable accuracy via compression.
Manticore Search: mmap speeds up columnar KNN rescoring
See how memory-mapped database files make columnar KNN rescoring in Manticore Search 2.57–3.13x faster without slowing out-of-core search.
How search works in Kiva's SaaS, where every customer has their own database
How Kiva uses Manticore Search in a SaaS platform where every customer has their own database and schema: combining a periodically rebuilt table with an RT table, processing changes in real time, and running search without dedicated servers.
Beyond MariaDB full-text search: how wine.co.za made about 145,000 records easier to search with Manticore
How wine.co.za uses Manticore Search alongside MariaDB to improve full-text search across its wine industry archive, provide better search-as-you-type suggestions, and keep maintenance to a minimum.
Better Vector Search for Long Documents: Chunking Strategies Inside Manticore Search
An embedding model reads only the first few hundred tokens of a document and silently drops the rest. Manticore Search now splits long documents for you at INSERT time: add chunk_strategy to the vector column and pick one of five chunking strategies. No ingest pipeline, no text splitter library. On our own manual, recall@5 for deep content went from 55% to ...