LLM Workflow Orchestration for External API Task Execution
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current AI systems struggle to handle queries outside their training data and require cumbersome integration with external applications, necessitating manual configuration and expertise, which is resource-intensive and not scalable.
Innovation Solution
A system that uses a large language model (LLM) to generate workflows, incorporating building blocks like semantic graphs and external applications, enabling automated integration and execution of tasks without manual coding, by determining appropriate applications and making API calls.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If current LLMs are used to respond to user queries, then they can handle queries within their training data, but they cannot handle queries outside their training data and require manual integration with external applications
Solution Approach 1:
The system enables LLMs to autonomously generate and execute workflows with external applications without requiring manual configuration by software engineers. The LLM automatically creates integration logic, making the system self-configuring and eliminating the need for expert intervention in each integration scenario.
Solution Approach 2:
The system transforms static LLM responses into dynamic workflow executions. Instead of relying on fixed training data, the LLM can dynamically generate workflows that adapt to various external applications and query types, allowing the system to handle unseen queries through automated workflow generation and execution.
2Adaptability or versatility
If manual integration of external applications is performed, then the AI system can use external applications, but the process requires several man hours and expertise
Solution Approach 1:
The LLM automatically generates workflow configurations and integration logic for external applications without requiring manual intervention from software engineers. This self-service capability eliminates the time-consuming process of manual configuration, testing, and debugging that previously required expert knowledge and multiple man hours.
Solution Approach 2:
The system performs preliminary workflow generation and configuration automatically before execution. By pre-generating the integration logic and workflow structure through LLM automation, the system eliminates the need for subsequent manual configuration steps, significantly reducing the time required for application integration.
3Adaptability or versatility
If conventional integration methods are used, then external applications can be connected, but the process is cumbersome and expensive
Solution Approach 1:
The LLM autonomously generates workflow configurations and integration logic for connecting external applications, eliminating the need for manual configuration by experts. This self-service approach makes the integration process simple and accessible without requiring specialized knowledge or cumbersome setup procedures.
Solution Approach 2:
The system introduces an intermediary workflow generation layer between the LLM and external applications. This intermediary automatically translates user intent into configured workflows, simplifying the integration process and making it easy to connect external applications without direct manual configuration.
4Device complexity
If LLMs operate as standalone systems, then they are simple to deploy, but they cannot integrate other applications effectively
Solution Approach 1:
The system transforms the static standalone LLM into a dynamic system that can automatically generate and execute workflows with external applications. This dynamic capability allows the LLM to adapt to various external applications while maintaining ease of use through automated workflow generation, effectively resolving the contradiction between simplicity and integration capability.
Data Source
AI summary
Systems and methods described receive user input or query at a user interface associated with a large language model (LLM) to perform a task. The user intent and persona are determined based on the user input and other user related data. An LLM automatically generates a workflow consisting of a plurality of steps for performing the task received. Each step of the workflow is mapped to a building block that is to be used for processing the step of the workflow. Selection of the building blocks is based on a plurality of factors, including relevancy, complexity, cost, and accuracy. One of the building blocks used is to perform an application programming interface (API) call to one or more external applications for executing a step of the workflow. Parameters for the API call may be obtained from a generated catalog.Results from the workflow may be customized based on the persona.