Update Gradio to version 5.16.0

Configure clean Gradio HF Space deployment

fix: add unique remote port support and binary check to bore tunnel

fix: bind default internal registry URL to PORT env var (10000 on Render)

fix: add int64 and int32 tensor dtypes to transport protocol DTYPE_MAP

perf: move embed_tokens to Node 0 GPU and convert Orchestrator to zero-weight CPU container (< 40 MB RAM)

fix: auto-connect orchestrator on first chat completion request

fix: add 10s background heartbeat ping loop in colab_runner.py

fix: mount registry routes directly on main Gateway app for single Render service URL

feat: add native bore.pub tunnel support to colab_runner.py

feat: add scripts/colab_runner.py for automated 3-node Colab execution

feat: implement dynamic topology layer assignment and partial-token fault tolerance

Update README with live Render URL and Zero-RAM Meta Loader benchmarks

Implement zero-RAM meta device shell loader in layer_loader.py

Set reuse_address=True on NodeServer socket

Add --registry-url, --public-host, and --public-port flags to node CLI

Add single overall run_demo.py script

Fix setuptools build backend in pyproject.toml

Fix SSE streaming token decoding for OpenAI SDK

Add root GET / endpoint to Gateway

Deploy ShardFlow infrastructure on Render

updates

first commit