Pixeltable

Pixeltable is a unified multimodal backend that combines a database, orchestration and serving in a single Python file. It is designed for applications that process images, video, audio and documents alongside ordinary columns, and for the agents and AI pipelines built on top of that data. A developer declares tables as Python classes. Annotated fields become stored columns, while assignments become computed columns that run on insert and on update, so transformations are incremental rather than a pipeline that must be re run. Embedding indexes live on the table model and stay current with the data, which removes the need for a separate vector database. A FastAPI router in the same file defines endpoints, and command line tools create and migrate tables and start services. The documentation lists uses such as retrieval augmented generation, live APIs, agent tool calls as columns and media pipelines that treat frames and transcripts as rows, and includes migration guidance for teams coming from other stacks. The product suits Python developers who want to replace a glued together stack of object storage, warehouse, vector store and orchestrator with one catalog. It runs locally against a local catalog, and Pixeltable Cloud is described as being in limited beta, available by emailing the vendor. Pricing is not published in the documentation. The vendor does not list plan prices or usage rates.