scarletkc/colab-playground
Developer ToolsDesktop front end for the Google Colab CLI that manages accounts, GPU runtimes, cost guards, port forwarding and runtime plugins.
What it is
Colab Playground wraps the official Google Colab CLI (colab, package google-colab-cli) in a desktop GUI built with Rust, GPUI and Ely GPUI Components. It is not a notebook editor and does not replace the CLI; every operation goes through colab. Its main job is to make runtime management less error-prone: keep several Google accounts, start CPU/GPU/TPU runtimes with hardware requirements, stop them, forward ports, and run plugin scripts on the remote runtime.
Cost control is central. The app shows balance, burn rate and remaining time, and can stop runtimes automatically on the runtime side when the app is closed. The bundled "Deploy a GGUF model" plugin serves a Hugging Face GGUF model through llama.cpp with an OpenAI-compatible API and web chat, choosing a quantisation that fits the assigned GPU. Plugin authoring is described in docs/PLUGINS.md.
Get started
Release builds are offered for Windows x64, macOS Apple silicon and Linux x64 via the latest release; the release page and install guide explain which file to pick and what first launch asks. On Windows, the Colab CLI must run inside WSL.
To run from source, install a Rust toolchain and, in the repository root, run:
cargo run -p colab-playground # demo data: no Google account, nothing billed
cargo run -p colab-playground --release # the real Colab CLI
Demo mode is the safe first step because it uses synthetic data and creates no billable runtimes. Real mode requires the guided setup to find or install google-colab-cli and sign in to a Google account. Backend flags, demo scenarios and settings storage are covered in docs/USAGE.md.
Operating constraints
Colab charges compute units while a runtime exists, whether busy or idle; closing a port forward or the app does not stop billing. Stopping a runtime wipes its files and downloaded models. The spending cap and balance floor only work while the app is running; auto-stop and idle stop depend on a watchdog started on the runtime, and the app reports when it could not start one. The CLI keeps one signed-in account per machine, so switching accounts in the app affects every other use of colab, and runtimes left on the previous account keep billing until their own auto-stop. Runtimes opened outside the app get no auto-stop and are shown so they can be stopped manually. Plugins execute scripts on the paid runtime, so only trusted plugins should be installed. When usage is unknown, the app marks it as unknown rather than zero, and guards report that they cannot verify a limit.
HotRepo's take
The repository is new and records only an early growth window: 20 stars added between 2026-11-17 23:46 UTC and 2026-11-18 01:16 UTC, taking the project to 127 stars and 6 forks in about 90 minutes. That is too short a series to describe a trend or compare it with the wider field, and the project has no recorded GitHub daily rank. What stands out technically is the narrow scope: it delegates to the Colab CLI instead of reimplementing Colab, and it puts cost guards and runtime-side auto-stop at the center of the design. Those choices fit the main failure mode of Colab usage — forgetting a runtime is still billing — better than a generic notebook client would. Whether the downloaded binaries correspond to the published Rust source cannot be established from the repository materials alone, so verify releases against the source before using real accounts.
Written by AI from the project's README, updated . The project's own docs are the source of truth.
You might also like
- googlecolab/google-colab-cliRun and manage Google Colab CPU, GPU, and TPU runtimes from the terminal.
- TizenryA/colab-gpu-studioSelf-hosted mobile web workbench that runs ComfyUI image and video workflows on on-demand Google Colab GPUs.
- vivaui/vivaBuild native Go desktop apps that render through each OS's own window and GPU APIs, without a WebView or bundled runtime.
- longbridge/gpui-kitA Rust UI framework for building cross-platform desktop apps on GPUI, with a full component library and virtualization included.
- tw93/PakePackage any website into a lightweight Tauri desktop app with a single CLI command on macOS, Windows, and Linux.