Shared Context Manager for Coordinated Multi-Agent Intents
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Consumer devices hosting multiple voice agents often lack awareness of each other's tasks, leading to confusion for users as commands are not coordinated, and information is not shared between agents, resulting in inconsistent responses.
Innovation Solution
A shared context manager (SCM) that receives and processes intents from multiple agents, forming modified intents and storing them in an intent stack to facilitate information sharing and resolve anaphoric references, ensuring coherent command execution across agents.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If multiple agents are hosted on a single device, then the device can respond to specific tasks with specialized agents, but the agents are not aware of tasks performed by other agents leading to user confusion
Solution Approach 1:
The patent merges the context management capabilities of multiple agents into a single shared context manager. This central component receives, processes, and stores context information from all agents in a unified manner, allowing any agent to access and utilize context generated by other agents, thereby eliminating information silos while preserving agent specialization
Solution Approach 2:
The shared context manager acts as an intermediary between multiple agents. It receives context information from agents, processes it through an intent stack, and makes it available to other agents. This mediator architecture enables indirect communication and information sharing between agents without requiring direct agent-to-agent interactions
2Adaptability or versatility
If each agent operates independently, then agents can specialize in specific tasks, but commands are not coordinated and information is not shared between agents
Solution Approach 1:
The system implements feedback mechanisms where agents submit their context and intents to the shared context manager, which then processes this information and feeds it back to the agents through the intent stack. This closed-loop feedback enables agents to adjust their behavior based on context from other agents, improving command coordination while maintaining specialization
Solution Approach 2:
The shared context manager provides universal functionality that serves all agents simultaneously. It handles context collection, processing, storage, and distribution in a unified manner, enabling coordinated operation across specialized agents without requiring separate coordination mechanisms for each agent pair
Data Source
AI summary
A user controls a computer application with a first multi-modal agent and a second multi modal agent. A shared context manager is configured to receive from the first agent a first intent having a first data field, and to receive from the second agent a second intent having a second data field. The shared context manager forms a modified second intent based upon the first data field, and provides the modified second intent to the second agent. An intent stack is configured to receive the first intent and the modified second intent from the shared context manager and to store the first intent and the modified second intent.


