MULTI-AGENT TASK MANAGEMENT GUIDED BY GENERATIVE ARTIFICIAL INTELLIGENCE

Smart overview of the Invention

The patent application discusses a system involving multi-agent task management guided by generative artificial intelligence (AI). This system is designed to manage tasks within software applications, utilizing a computing apparatus to assign tasks to an application assistant. The application assistant comprises multiple agents that interact with a generative AI model to complete tasks and update contextual information based on these interactions. The technology aims to enhance productivity in software applications by effectively integrating generative AI models.

Task Management Process

In this system, the task management process involves orchestrating multiple agents interacting with a generative AI model. The computing apparatus evaluates whether subtasks should be created and assigns an execution agent to carry out the task. Content generated by the generative AI model is assessed based on a call by the application assistant to the assigned execution agent. This structure allows for effective task completion by leveraging the capabilities of generative AI models.

Agent Roles

The agents involved in this system include task management agents and execution agents. Task management agents are responsible for generating content that assists in executing workflows for task completion. Execution agents focus on generating content specifically aimed at completing the task at hand. These agents work collaboratively under the guidance of a generative AI model to achieve efficient task management and content generation.

Application Assistant Functionality

The application assistant serves as a service within the application, generating prompts to elicit AI-generated content from the model. It configures prompts based on task attributes, project information, and other relevant data, submitting them to the generative AI model via an API. The orchestration layer of the application assistant coordinates agent activities until a task completion agent determines the task is complete, ensuring a streamlined workflow.

Generative AI Models

The generative AI models used in this technology include large-scale foundation models trained on diverse data using various learning techniques. These models can be based on architectures such as generative adversarial networks (GANs), variational auto-encoders (VAEs), and transformer models, including multimodal transformers. Multimodal models handle different data types like text, images, video, and audio, enabling them to generate coherent and contextually rich multimodal outputs, enhancing the system's versatility and effectiveness in managing diverse tasks.