Private deployment console · JSON storage · No database
Deploy, tunnel, and schedule MiniMax-H3 (fl2va & ref2va video + audio studio) on Vast.ai or Google Colab Pro+.
Read FastAPI logs from your Vast.ai instance.
Terminate permanently destroys the selected instance and its disk (billing stops, data is lost) and asks for confirmation. Start boots the existing Vast.ai instance, deploys the configured repository if its remote folder is missing, starts tunneling, and waits for MiniMax-H3 readiness. Stop halts the instance without destroying it; storage charges may continue.
Save specifications first. Select an offer to review its compute price before creating. Disk and network may be billed separately.
Available offers will appear here.
One click: searches offers using your saved GPU, RAM, disk and maximum price, rents the cheapest match, then clones, configures and starts MiniMax-H3. Only available when no instance is selected.
Read FastAPI logs from your active Google Colab session.
google-colab-cli is installed & authenticated inside your DevOps server Python venv (checking venv…), this controller runs colab new, colab exec, and colab stop directly to allocate GPUs, clone GitHub/GitLab repos, start FastAPI + tunnel, and release runtimes. If the CLI venv is not configured or interactive re-auth is required, use Open in Colab → select GPU → Run all.
Waiting for Colab CLI session or notebook registration (`POST /api/colab/register`)…
View all Google Colab CLI sessions on this server, connect to any session to inspect its GPU & MiniMax-H3 status, stop the MiniMax-H3 service inside a session, or terminate the Colab VM to stop Compute Unit usage.
Loading Google Colab sessions…
No cache observation yet.
Read FastAPI logs from your saved runtime.
Your deployment activity will appear here.
Cancellation closes the SSH channel. Remote child processes may continue; inspect logs before retrying.