deploy demo space from 9889c231ec4c

deploy demo space from de407f7ae660

deploy demo space from ad13168d36ff

deploy demo space from 17b6c53765c3

deploy demo space from c93bb72fbc91

deploy demo space

deploy demo space from 5a902dad15e8

deploy demo space from 597e5b1fba43

Update Armnet organization references

deploy demo space from fdef0e8c0b7a

deploy demo space from 5d8c586f64ef

deploy demo space from b9604a838b2e

deploy demo space from f355e79dd1a0

deploy demo space from c2cbfbd3aa74

Migrate demo to the single-arm BusyBox cell

drop an accidentally uploaded lint cache

two-task selector (buttons + left switch), per-task boards, rename to armnet eval

Pin the environment-aware client and eval image

Bump eval image to armnet-065cc39c6691

Add armnet.dev link

Add armnet.dev link

Leaderboard: add Training framework column; tag demo runs as lerobot

Add Model type column (act/diffusion/smolvla/...) to leaderboard

Add live cell availability light (green/yellow/red) + gate Run on availability

Move demo API key to a Space secret (no keys in repo) + add Instructions tab

Leaderboard: small inline avatar via injected CSS (Gradio 6 strips img sizing)

Fix job secrets: reference a named orchestrator secret (uppercase HF_TOKEN) instead of inlining the token value

Add persistent pooled leaderboard (HF dataset) + 20 rollouts; pin policy revision

Fail fast when cell busy (allow_queue=False, synchronous submit) instead of queueing

Unlock faster on lost contact (give-up 120s -> 10s)

Robust follow: poll backoff + give-up unlock on lost contact; backfill rollouts from terminal result; ws reconnect backoff

Reduce request rate: slow timers + run video timer only during a run

Pin Space to Python 3.12 (armnet-client requires <3.13)

Ship armnet demo Space (policy eval on real robots)

initial commit