Use direct tokenized Gradio queue calls and lock app height

Patch ZeroGPU handshake for embedded client decode

Use Gradio client for token-aware live decode

Reduce ZeroGPU decode cost and restore same-origin API calls

Make frontend Gradio API calls resilient

Extend live decode wait for ZeroGPU queue

Add Local Lens cultural translator interactions

Improve live decode resilience on ZeroGPU

Fallback when Gradio decode returns empty event

Point frontend to Gradio API prefix

Use Gradio API for decode and narration

Implement award-ready interaction flow

Use official Gradio client for ZeroGPU calls

Prevent decode lockups and normalize app height

Call GPU endpoints through Gradio bridge

Route live decode directly through ZeroGPU

Improve decode endpoint diagnostics

Wire pixel journey flow to Gradio AI APIs

Deploy: gr.Server + React SPA + /api/decode (PNGs via hf CLI)