NVIDIA has announced an updated version of its ChatRTX application which is designed to enhance interaction with local data on RTX-powered Windows PCs and workstations. We previously reported on NVIDIA’s ChatRTX and how it allows you to have a fully local LLM on your computer.
This update, which was initially showcased at the GTC event in March, brings additional features and compatibility with new language models to the RTX-accelerated chatbot application. ChatRTX employs a retrieval-augmented generation approach, using NVIDIA TensorRT-LLM software and RTX acceleration technology to enable chatbot functionalities on local devices.
The application, supported by large language models (LLMs), allows users to search and query their personal notes and documents. ChatRTX responds by rapidly generating pertinent replies while operating on the user’s own device. The updated version introduces support for several new LLMs, including Gemma, an open local LLM trained by Google. Gemma is based on the research and technology underlying Google’s Gemini models and prioritizes responsible AI development.
Additionally, the update extends support to ChatGLM3, an open bilingual LLM capable of understanding both English and Chinese, based on a general language model framework. The application now also integrates with OpenAI’s Contrastive Language-Image Pre-training (CLIP), a neural network model that learns visual concepts from natural language. With the inclusion of CLIP, ChatRTX users can interact with images on their devices using descriptive words and phrases, eliminating the need for intricate metadata tagging.
A new feature with the release is the implementation of voice interaction capabilities. ChatRTX now supports Whisper, an automatic speech recognition system that processes spoken language through AI. This enables users to submit voice queries to the application and receive text-based responses.
Be sure to check out NVIDIA’s video about the update Below. For more information on ChatRTX and its features, check out NVIDIA’s website here.
