Public: anyone can read this; the crew can change it.

Migrated from github.com/Trevato/SO-ARM100

workbenchthis machine

TOP: the hosted app connects to the real arm and camera - it still serves the simulation with ARM_SERVER set and both addresses granted #40

shipped merged opened this has it

Landed in pull #43 · run

opened this

Probe 2026-10-20 20:25Z, everything outside the app is in place:

  • ARM_SERVER is set on the production app (app page env list).
  • Both destinations are GRANTED for this repository (forge 0.7.210, #3437): 100.89.95.68:8101 and 100.89.95.68:8080 (granted_by trevato).
  • forge.json declares both (pull #10, merged).
  • The production app was redeployed after the grants (run 01m5dfej5naqwpedys9e5ekwz4, ok).
  • The real arm answers on the Pi: curl 127.0.0.1:8101/status → shoulder_pan position 2017, 23 °C, torque true.

Yet: the public /status returns the simulation (every joint position 2048, torque false, temperature 0), and on the Pi ss shows NO connection from the forge (100.84.55.62) to :8101 or :8080 - only the owner's Mac. Before the grants the deploy log said egress refused: 100.89.95.68:8101 ...; after, nothing.

Find why and fix, smallest change first. Likely places: the production entry (forge.json app.run = Software/so101/bench/fake-serve.ts) may never read ARM_SERVER, or it reads it only at a different path; the WebSocket/MJPEG client may not go through the cell's egress (check how Bun fetch/WebSocket reach the outside inside a cell - an egress proxy env such as HTTP(S)_PROXY, or a direct connect); or it falls back to the simulation silently. Make the fallback loud: /status says "source":"live <host>" or "simulated: <why>", and the UI shows it. Done: the public /status shows the real joints (not 2048s), the camera tab shows the Pi's stream, and ss on the Pi shows the forge connected. The crew may use the Pi over SSH once trevato/forge#3470 lands ([email protected]).

Sign in to join this conversation.

↑↓ navigate ↵ open ⌘↵ its app esc close

Keyboard shortcuts

Docs