Prompt Hospital Turns AI Model Orchestration into a Visual Office

Summary

Prompt Hospital is an open-source project that presents an AI team as an isometric office inspired by Theme Hospital. Each model occupies a room, while tasks appear as envelopes in a coordinator's tray; dragging an envelope onto a model records the assignment and can trigger the model to run. The frontend uses Vite, React, TypeScript, and SVG and can operate offline, while a journal.json file serves as the source of truth for models, tasks, statuses, and history. A Python API exposes journal operations, and dispatch.py polls the journal and invokes configured executors such as codex exec or claude -p. The project supports a progression from read-only monitoring to journal writes and full execution, with the dispatcher recording completion or blocking states. Developers can add a model's visual appearance to the frontend catalog and connect its identifier to a custom CLI or API executor in the dispatcher. The repository includes local setup instructions, Docker and nginx deployment guidance with Basic Auth, and tests, linting, and build commands. Its reference configuration uses DeepSeek as coordinator, Claude Code and Codex as local executors, and several models through OpenRouter for review; planned work includes per-model statistics and a mechanism for returning idle models to the coordinator.