Replies: 1 comment
|
Hey there @vggg, check out https://github.com/mpfaffenberger/openpup - this is kinda what you're thinking right? |
0 replies
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Uh oh!
There was an error while loading. Please reload this page.
Hi @mpfaffenberger — first, thanks for code-puppy; the plugin/callback architecture is a joy to build against.
Idea: a
remote_controlplugin that lets you monitor and steer a locally running code-puppy session from another device (phone/browser) — see proposed file changes, approve/reject them remotely, and send messages back — similar to Claude Code's recent "Remote Control" feature, but local-first and transport-agnostic.I've been mapping it against the codebase and it looks like it can be a pure plugin, no core changes, because the seams already exist:
file_permissionandrun_shell_commandhooks (relocate the decision to the remote surface, return it).frontend_emitterplugin (its docstring already anticipates "an embedder … a WebSocket backend").agent_run_start/agent_run_end.custom_commandfor/remote-control.dbos_durable_exec(opt-in).The design keeps the transport pluggable and outbound-only (no inbound ports): a small
RemoteControlTransportinterface with swappable backends — a chat-app bridge (Telegram/Discord = relay + UI + push, lowest friction) first, and a self-hosted FastAPI+WebSocket relay (uvicornis already in the tree) for richer diffs.I've got an inert scaffold + a runnable Telegram approve/reject spike working locally already. Before I invest further, two questions:
file_permission(and theget_user_approvalflow) a stable extension point you're comfortable having a plugin build on, or is it likely to change? Anything you'd want me to avoid coupling to?Happy to share the design doc and open a draft PR for the scaffold if useful. Either way, thanks for considering it.
All reactions