GroupBy

Chroma adds first-class sparse vector search support, enabling hybrid retrieval that combines dense semantic embeddings with sparse keyword-based signals for more precise results. Supports BM25 and SPLADE embeddings via a new Search() API with hybrid search via Reciprocal Rank Fusion (RRF), backward-compatible with existing collections.

Oct 29, 2025

Chroma now supports regular expression filtering via $regex and $not_regex operators in the where_document field, allowing users to search and filter collection documents by pattern — such as finding function definitions, email addresses, or URLs. Available through the Python API and the Chroma Cloud UI.

Jun 24, 2025

Factory AI uses Chroma Cloud to power semantic and full-text code search for its AI software development agents (Droids). Chroma enables semantic similarity search, regex/full-text search, and collection forking to incrementally index git diffs without full re-ingestion. Serverless scalability handles unpredictable parallel workloads across many concurrent Droid instances. Factory's Droids hold the #1 score on Terminal-Bench, resolving 58.8% of tasks.

Oct 7, 2025

Chroma Cloud can now automatically crawl and index websites. Users supply a starting URL and Chroma handles scraping, converting pages to markdown, chunking via Tree-sitter, and generating embeddings — no custom ingestion pipeline required. The resulting collections support hybrid search blending dense embeddings for semantic matching with BM25 sparse vectors for keyword precision. Designed for product docs, blogs, and support pages.

Nov 4, 2025