Lumos

Lumos is a Chrome extension for asking questions about web pages using models running on your own machine. It's for readers working through long discussions, product reviews, news articles or technical documentation who want summaries and answers tied to the material they're reading. Ollama handles inference locally, without a remote AI server.

The extension uses retrieval-augmented generation (RAG) to bring page content into the conversation. You can ask about a whole page or narrow the context to text you've highlighted. Site-specific content extraction lets it focus on relevant sections, such as an article body or forum comments.

Documents work too. Lumos accepts PDFs, CSV and JSON files, along with plain text formats such as Markdown and source code. You can also use clipboard text as an attachment. An attached file becomes the conversation's source material in place of the current page.

With a multimodal model, Lumos can include images from the page in your prompts or process attached JPEG and PNG files. It also lets you save and reopen chats, regenerate the last response and cancel a request.

Lumos is open source under the MIT license. The browser extension depends on a local Ollama server for its model and embedding workflow; the model doesn't run inside Chrome itself. You can choose an Ollama language model such as llama2 and an embedding model such as nomic-embed-text. The local backend can also run in Docker.