Multi Agent Architectures
Last updated
Last updated
More complex use cases will require multiple Agents to work together, each handling a specific task or purpose within the larger process. By dividing responsibilities among specialized Agents, you can ensure that each step is executed with precision and that the overall system remains robust, scalable, and adaptable to changing requirements.
The Great Wave AI Platform leverages four powerful multi-agent architecture patterns to handle a wide range of workflows, from simple linear tasks to complex, multi-dimensional operations. Here’s a clear description of each pattern:
This pattern organizes agents in a sequential flow, managed by a Sequence Agent, where each agent completes a specific task before passing its output to the next. It is well-suited for workflows that require a highly structured, step-by-step approach, ensuring each stage builds on the outputs of the previous one.
In this pattern, a central Fusion Agent acts as a coordinator, managing multiple specialized agents that perform distinct tasks related to a larger goal. The Fusion Agent integrates the outputs of these agents into a cohesive result, providing a centralized approach to handling multi-faceted workflows.
An Adaptive Agent uses metadata prescribed against each agent to dynamically decide which agent to interact with given a user input. It then integrates the agent responses into a cohesive result.
Sequence - HR Workflow Automation: Employee Handbook Customization
Use Case: A new hire requires an employee handbook tailored to their role and department.
Business Area: Human Resources (HR)
Workflow:
Input Validation Agent: Ensures HR inputs (e.g., employee details, policies) are clean and complete.
Structuring Agent: Organizes the data into sections such as benefits, company policies, and role-specific guidelines.
Refinement Agent: Polishes the language for clarity and professionalism.
Content Creation Agent: Generates a fully customized employee handbook.
Result: Leverages collections of HR policies, role-specific guidelines, and writing style guides to ensure accuracy and consistency in the final handbook.
Use Case: Generate personalized evaluation summaries for job candidates.
Business Area: Recruitment
Workflow:
Fusion Agent (Recruitment Coordinator): Orchestrates and integrates outputs from multiple spoke agents.
Categorization Agent: Groups candidates based on skills and job requirements.
Knowledge Validation Agent: Verifies candidate credentials (e.g., certifications).
Content Creation Agent: Generates personalized evaluation summaries for each candidate.
Refinement Agent: Polishes the summaries for clarity and presentation.
Result: Utilizes candidate resumes, job descriptions, and industry benchmarks to deliver cohesive, data-backed evaluation summaries.
Use Case: A customer submits a support ticket with mixed inquiries about billing, technical issues, and product usage. The Adaptive Agent analyzes the request, dynamically routes specific parts to specialized agents, and then internally integrates and refines the responses into a single cohesive reply.
Workflow Overview:
Metadata Extraction Module:
Function: Analyzes the incoming ticket to extract key metadata such as inquiry type (billing, technical, product info), urgency, and context.
Outcome: Determines that the ticket contains multiple issues requiring distinct expertise.
Dynamic Routing Module:
Function: Based on the extracted metadata, it decides which specialized modules to engage:
Billing Module: Activated for payment and account inquiries.
Technical Module: Handles troubleshooting and error diagnostics.
Product Information Module: Provides guidance and tutorials for product usage.
Specialized Modules:
Billing Module: Reviews the customer’s account, recent transactions, and relevant billing policies to resolve payment concerns.
Technical Module: Analyzes error logs, device compatibility, and system status to address technical issues.
Product Information Module: Delivers tailored usage guidelines and support content for the customer’s product.
Internal Integration and Finalization:
Function: The Adaptive Agent internally gathers the outputs from the specialized modules and synthesizes them into a unified, clear, and comprehensive response.
Approach:
Integrates the information seamlessly, ensuring consistency in tone and clarity.
Performs a final internal review to verify accuracy and relevance before the response is delivered to the customer.
Result: The Adaptive Agent efficiently tailors its response by dynamically engaging the relevant specialized modules and integrating their outputs internally. This approach ensures that the customer's diverse inquiries are addressed comprehensively in one streamlined reply.
