Lesson 3 5 min
Use a local model everywhere
A model you install once serves Aria, your agents and your workflows, with no re-download.
Try it in AWFlow
Opens this lesson's workflow in AWFlow. You review it before anything is saved.
In Aria
Pick an on-device model in the composer’s model picker, or click Set default on a chat model: the assistant prefers it, and new Web LLM nodes are pre-filled with it. In Ask mode, Aria can also classify an image, moderate text or transcribe audio with a local model and answer from the result. If no suitable model is installed, it tells you which kind to add.
In an agent
Choose an on-device model on the builder’s Model tab, or turn on the On-device AI skill.
In a workflow
- The Local AI nodes do one job each, such as Classify Image, Summarize, Translate, Transcribe Audio (Whisper) or Run Local Model.
- Web LLM and Transformers Chat plug into the Model slot of an AI node, such as Basic LLM Chain or Tools Agent.
- Local Classifier, Local Q&A Model and Local Embeddings plug in the same way, to classify text, quote answers from a knowledge base, or turn text into vectors on this device.
A model installed on the Local AI page is used with no re-download. The example below reads the page you’re on and summarises it with a Web LLM model: pick an installed model the first time.
Two more local options
- Chrome AI: a node that sends a prompt to the model built into Google Chrome (Gemini Nano), with no API key. It works only in Chrome, where Google is still rolling it out.
- Ollama: when Ollama runs on your computer, every model you have pulled appears in Aria’s model picker under Ollama · this computer. It isn’t in the browser, but nothing leaves your machine.
Key takeaways
- Set default makes a chat model your default in-browser model, which Aria prefers.
- Agents use it from the Model tab or the On-device AI skill.
- Workflows use it through the Local AI nodes, Web LLM and Transformers Chat.