Chroma

From training data to inference context

Chroma’s CEO dates the company’s beginning to 2022, initially working on active-learning data curation. It shifted toward inference context and released ChromaDB in February 2023. The database and the later Foundation context work concern the material agents can retrieve, rather than authorship of the underlying foundation models. [1]

An agreement and two different continuity paths

Chroma announced a definitive agreement to join Replit on October 23, 2026. The primary establishes a signed agreement, rather than verified completion of company control. [1]

Its Cloud notice schedules writes to stop on November 22 and reads on December 22. Those are future service cutoffs. The notice separately says open-source ChromaDB remains available under Apache 2.0 and Replit will continue contributing, so the hosted-service winddown is not a shutdown of the code project. [2]