AI · Edge
Edge AI on VegaDūta
Edge AI on VegaDūta means running models on the user's own device. In-browser LLMs run via WebLLM — a model marketplace lets you browse, download, and manage installed models, and inference then happens locally in the browser rather than on a server.
Alongside that, device pairing connects desktop and mobile companion agents to the platform, giving agents device tools such as notifications and screen reading. To be precise about the term: 'edge' here means on-device and in-browser models plus paired devices, not edge-datacenter compute.
LLMs in the browser
WebLLM runs open models directly in the browser using the device's own hardware. Once a model is downloaded, inference happens locally — useful when connectivity is unreliable or when the text being processed should not leave the machine.
A marketplace for edge models
The model marketplace is the management layer for on-device AI: browse the catalog, download the models that fit your hardware, and manage what is installed. It turns 'run a local model' from a developer exercise into a click-through experience.
Paired devices as companions
Desktop and mobile devices pair with the platform as companion agents, extending what your agents can reach. Device tools such as notifications and screen reading let an agent surface information on the device you are actually looking at, rather than only in a chat window.
What 'edge' means here — honestly
This capability is about on-device and in-browser models and paired companion devices. It is not edge-datacenter compute or a CDN-adjacent inference network. If you need heavier models than a browser can hold, the LLM connectors — including self-hosted Ollama — are the right tool.
Frequently asked questions
Can I run an LLM entirely in my browser?
Yes. VegaDūta uses WebLLM to run models in-browser on your own hardware: browse the model marketplace, download a model, and inference runs locally on the device. Text processed this way is not sent to a server for generation.
How do I choose and install an edge model?
Through the model marketplace: browse the catalog, download a model suited to your device, and manage your installed models from the same place. Larger models need more memory, so the practical choice depends on your hardware.
What can a paired device do?
Device pairing connects a desktop or mobile companion agent to the platform, and agents gain device tools such as sending notifications and screen reading. It extends agents from the chat window onto the devices you actually use.
Is this the same as edge computing in a datacenter?
No. On VegaDūta, 'edge' means on-device: in-browser models via WebLLM and paired desktop or mobile devices. It is not edge-datacenter or CDN-based inference — for heavier self-hosted models, use the Ollama LLM connector instead.
See it working in two minutes
The sandbox provisions a real tenant — describe an agent in one sentence and test it, no account, no card. Or browse ~90 industry workflow recipes to see what teams build.
