mozilla released experimental expansion for a browser called MemoryCache . It includes a system of dialogs based on machine learning, which analyzes the content viewed by the user. This system differs from other AI chats in that it adapts to the user’s interests, using the data important for him to form answers to questions. Expansion is distributed under the license of mpl and so far may
All MemoryCache components are performed on the user system and does not require connection to external resources. The basis of knowledge for MemoryCache is the GPT4all-J model, trained on extensive English-language data, including Pile, Sharegpt and Dolly. Model is available in various versions, including 3, 7 and 13 billion parameters. To work with the model, the Privategpt tool is used to efficiently use large language models on personal computers.
MemoryCache retains the viewed web pages to the local catalog “~/downloads/MemoryCache”, from which they are then extracted by a periodically launched script and transfers to Privategpt (INGest.PY) to supplement the existing language model. Thus, the model adapts to the interests of the user and can be used to discuss the viewed content in a text chat with a bot. Users can conduct a dialogue with AI-bot in a natural language, asking questions about the preserved pages. The bot adapts to the interests of the user and provides answers using local data, without the need to connect to external sources.