Shumai, an open‑source platform for creative teams, bundles asset storage, frame‑by‑frame feedback, secure sharing and AI‑assisted metadata into a single workspace. The project mirrors Frame.io’s polished UI while remaining self‑hostable. Users can spin up a full instance with Docker Compose in minutes, or install via npm with a PostgreSQL + pgvector backend.

Shumai supports S3‑compatible or local storage, role‑based access controls and custom metadata fields that fit any production pipeline. Heavy tasks such as video transcoding run on a Temporal‑orchestrated worker pool, letting large deployments scale independently. Integrated AI agents leverage Google Gemini to auto‑tag assets and answer queries via a sandboxed execution environment. Semantic search with vector embeddings finds assets by concept, not filename.

Installation guides cover three paths: a one‑click Docker Compose launch, an npm package requiring manual PostgreSQL setup, or a source checkout for developers. The CLI lets users upload files, create versions and manage projects from the terminal. Documentation also details sandbox security settings for safe AI script execution. The platform’s modular design lets organizations replace storage layer, swap AI provider, or scale Temporal workers without touching code.