diffusers-internal-dev/abot-world-interactive

Interactive ABot-World (a real-time, action-conditioned world model, Wan2.2-TI2V-5B) running natively on Modular Diffusers. The engine is one pipe.stream(action_source=...) loop over the ABotWorldStreamingBlocks preset (huggingface/diffusers#14159): the pipeline polls the held keys once per generated block and yields every block's decoded frames.

The serving shell — gradio.Server start/stop endpoints, a WebSocket that streams binary JPEG frames with steady pacing, per-session control queues that cross the ZeroGPU fork, and the front-end — is reused from acvlab/abot-world-interactive; only the ~40-line GPU rollout body is ours.