We launched the alpha version of the PocketLLM app about two months ago and the response has been overwhelmingly positive. Now, we’re taking it to the next level. The new PocketLLM is a comprehensive GenAI ecosystem that provides precise control over retrieval augmented generation (RAG). It also features a constantly expanding Model Bazaar, where users can access a variety of specialized pre-trained models. Additionally, PocketLLM offers an integration with OpenAI’s ChatGPT for generating the final answer. The integration with ChatGPT requires a network call to OpenAI’s API and is completely optional.
User Guide
You can find a complete user guide at this link. The compact 250MB application includes advanced GenAI tools that can transform large amounts of raw text (PDF or DOCX) into a domain expert AI Agent. Our Dynamic Sparse Technology enables quick processing of thousands of pages in just a few minutes using your ordinary laptop’s CPU. The two figures below introduce the user interface and the options available in the Beta version of PocketLLM. With PocketLLM, you can leverage an ever-growing bazaar of pre-trained models, teach the model new associations, provide real-time reinforcement learning for improved performance, and save and share the model for further enhancement or deployment.
With an array of cutting-edge features, PocketLLM is arguably the most comprehensive GenAI ecosystem app on the planet.
Unleashing Personalized AI-Agents: PocketLLM’s Vision of Decentralized and Private AI for All
In the current AI landscape, a handful companies with access to expensive specialized hardware have the capability to manipulate large AI models. This creates a situation where data needs to be transferred from user-facing hardware to AI-ready hardware, compromising privacy and control. At ThirdAI, we’ve developed an efficient brain-like dynamic sparse algorithm that allows us to train neural networks of any size without the need for specialized hardware. With our PocketLLM tool, you can build sophisticated AI models on simple CPU cores found in laptops and desktops. You have the flexibility to choose the text corpus for pre-training, fine-tuning with supervised data, and even guide the model through real-time reinforcement learning. You can refine and share the model as a base for others to customize, creating a continuous cycle of improvement. PocketLLM ensures “air-gapped” data privacy by keeping it within your hardware, whether on-premises or on a virtual private cloud.
More Information
The new version of the app is available and fully free for personal use. Download it at https://www.thirdai.com/pocketllm/.
Developers: The app utilizes our Neural Database Engine in the backend. The complete backend python notebook and APIs are located here. We’re excited to see the innovative applications you can create! To begin, apply for a free license by visiting our website.
