NVIDIA lets RTX owners build custom AI plugins

Tabbing out of a game just to change Spotify tracks, adjust fan speeds, or look up a walkthrough is annoying.

Nvidia is stepping in to handle all of that with local AI. On April 23, Nvidia rolled out Project G-Assist as an experimental feature inside the NVIDIA app, alongside a new ChatGPT-powered G-Assist Plug-In Builder.

Project G-Assist is an on-device assistant built specifically for GeForce RTX systems. It lets you monitor performance, tweak hardware settings, and execute commands through voice or text.

Small local models with cloud superpowers

Under the hood, G-Assist runs on a responsive small language model (SLM) executing locally on GeForce RTX GPUs. That means zero latency and completely private on-device inference.

Best of all, users can summon the assistant directly inside the NVIDIA overlay without minimizing their current game or creative app.

For custom tricks, the new Plug-In Builder uses ChatGPT to turn natural language prompts into proper JSON manifests and Python logic. You simply drop the generated config files into a designated folder, and G-Assist loads them automatically.

While local models keep things snappy, they have limits. To solve that, Nvidia provided a sample Google Gemini plug-in using a free Google AI Studio API key. If you are solo queuing in Apex Legends and cannot decide which Legend to pick, or wondering whether level 25 is safe for Nightmare mode in Diablo IV, G-Assist can consult Gemini in the cloud and report back.

Controlling peripherals and external services

Beyond games, the plug-in ecosystem taps into hundreds of public APIs and local hardware controls.

Nvidia shared several starter plug-ins on its GitHub repository. A Spotify integration allows hands-free playback and volume control, a Twitch tool checks if a specific streamer is live, and lightweight modules can fetch weather or stock data in real time.

For hardware enthusiasts, G-Assist can trigger benchmarks, ramp fan speeds, or sync RGB lighting across gear from Logitech G, Corsair, MSI, and Nanoleaf. It also provides C++ and Python bindings and works with multi-agent orchestration tools like Langflow.

Why it matters

For gamers and tinkerers with an RTX rig, this transforms your GPU into an always-on command center that does not force you to alt-tab. If you like automating your setup, it is worth playing with right away on GitHub. That said, dealing with Python environments and API keys means it is still very much a hobbyist tool rather than a consumer-ready one-click feature.

From an industry standpoint, Nvidia is quietly securing the desktop interface layer. While Microsoft pushes Windows Copilot for general office productivity, Nvidia is staking out the gaming and creative overlay where power users spend their actual time.

We think open-sourcing the plug-in specs and leveraging ChatGPT to generate glue code is a smart move to scale the ecosystem quickly. However, relying on user-generated Python scripts running alongside intense games introduces risks around performance overhead and system stability, which Nvidia will have to tightly monitor as the feature matures.