AIskNet 作者: Alberto García

Introducing "AIskNet" - a browser extension designed to enhance your browsing experience by leveraging open-source AI models locally. This extension fundamentally operates a local AI model that interacts with the content of your current web page, providing answers to your queries and thereby enabling quicker information gathering.

At its core, AIskNet employs state-of-the-art AI models optimized for compact size and efficiency. Operating directly on your local machine, this tool respects your privacy while autonomously employing advanced methods to retrieve information more accurately and swiftly.

Upon navigating to a new site and activating the extension icon, AIskNet automatically indexes the website content into "vector embeddings". When you pose a question, the extension sifts through these embeddings, identifying chunks of information most relevant to your query, and generates a response based on this information.

For transparency and accuracy, AIskNet highlights the text used as the source for generating its answers. This allows you to quickly double-check the answer's accuracy and form your own conclusions.

All data processed by this extension remains on your machine and is never transmitted to a remote server. The only instance where AIskNet interacts with remote servers is during the initial download of AI models upon installation.

AIskNet is a work in progress. With ongoing advancements in the field of artificial intelligence, it's expected to improve over time as more refined and compact open-source models become available. Currently, the model inference is carried out on the CPU, but future updates anticipate utilizing WebGPU to offload processing to your GPU. This will ensure an even more efficient and seamless user experience.

Source code: https://github.com/GarciaLnk/aisknet