Manticore Search 28.4.4 speeds up KNN rescoring, adds HTTP/custom-prompt conversational search, simplifies installs, improves facets and relevance defaults, and fixes auth, replication, SQL, and columnar/KNN issues.
A head-to-head benchmark on 975K 1536-dim vectors: turbopuffer's managed service vs Manticore on two $8.25/mo VPS. Matched recall (~0.966), ~4.5× lower cost, and comparable latency and throughput, with an honest look at where each approach wins.
Manticore can now build KNN indexes in parallel when saving chunks, merging chunks, and rebuilding KNN data. On a 16-core machine, building a KNN index for 1M vectors dropped from 8 minutes to 39 seconds.
How sharding works in Manticore Search, from splitting a table across CPU cores on a single node to automatic multi-node distribution that maintains a replication factor on its own.
Released in Manticore Search 27.1.5, the new ONNX Runtime backend makes auto-embeddings ~14× faster on average than the previous SentenceTransformers/Candle path on the same hardware, same model, same weights — and the margin holds whether you run 1 client thread or 32.
Run Manticore Search under systemd with accurate status reporting, cleaner reloads, journal-based logging, and safer shutdowns for RT workloads.