Pixeltable is an open-source Python framework that unifies the database, orchestration and serving layers AI applications usually stitch together from separate systems. Images, video, audio and documents live in tables, transformations are declared as computed columns, and indexes and HTTP routes are declarations in the same application file, so inserting a row automatically runs everything downstream. It is an open-source project on GitHub with around 1,600 stars, actively developed with fresh commits in 2026.
Instead of juggling object storage, a vector database, an orchestrator and glue endpoint code, Pixeltable collapses the stack into one declarative file that a human or a coding agent can write and review as a single diff. Tables are defined with typed columns for text, images, video, audio and documents, and computed columns apply user-defined functions or model calls on insert and update. Built-in indexing supports vector search for RAG and multimodal retrieval, and the same schema can be exposed as HTTP endpoints with the bundled serving layer, locally or on Pixeltable Cloud.
Pixeltable is free and open source under the Apache 2.0 license for self-hosting. Pixeltable Cloud, the managed hosted version, is in limited beta, which makes the project Freemium overall.
AI engineers and agent builders who want one declarative backend for multimodal data, pipelines and serving instead of wiring together four separate systems.