API-Grounded LLM Workflow Generation with Privacy and User Feedback
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing robotic process automation (RPA) systems struggle with spontaneous or unpredictable tasks due to their reliance on pre-defined workflows and lack of objective reasoning, while existing large language models (LLMs) face challenges in generating executable code for workflows that integrate user feedback and preserve data privacy.
Innovation Solution
A method and system using a large language model to generate executable code for workflows that integrate user feedback and adjust as needed, while preserving data privacy, by leveraging Application Programming Interfaces (APIs) and incorporating user feedback through a two-way interaction loop.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If pre-defined workflows are used in RPA systems, then task automation efficiency is improved, but the system cannot handle spontaneous or unpredictable tasks
Solution Approach 1:
The system transitions from static pre-defined workflows to dynamic workflow generation. LLMs generate workflows on-demand based on user queries and task requirements, allowing the system to adapt to spontaneous tasks while maintaining automation efficiency through structured generation processes.
Solution Approach 2:
The system enables automated workflow generation without requiring expert knowledge or manual configuration. Users can directly query the system for task automation, and the LLM generates executable workflows autonomously, making the system self-sufficient for handling diverse tasks.
2Loss of information
If LLMs directly interact with private data for workflow generation, then task understanding is improved, but data privacy is compromised
Solution Approach 1:
The system introduces an intermediary layer where LLMs interact with synthesized task representations rather than direct private data. The LLM processes task descriptions and generates workflows without direct access to sensitive information, maintaining both task understanding and data privacy.
Solution Approach 2:
Instead of using actual private data, the system uses synthesized or anonymized task descriptions as inputs to the LLM. This copying approach preserves the essential task characteristics needed for workflow generation while eliminating direct exposure to sensitive information.
3Manufacturing precision
If user feedback is integrated into workflow generation, then workflow accuracy is improved, but system complexity increases
Solution Approach 1:
The system implements a feedback loop where user responses to generated workflows are captured and used to refine subsequent workflow generation. This iterative process improves accuracy by learning from user corrections and preferences while maintaining a relatively simple architecture through direct integration with the LLM interface.
Data Source
AI summary
A method for using a large language model to generate executable code for workflow execution in a manner that integrates user feedback and adjusts the workflow as needed while preserving data privacy is provided. The method includes: receiving first information that relates to a workflow context, second information that relates to at least one application programming interface (API), and third information that relates to a code generation request; using the received information to generate a lecture, and transmitting the lecture to a language model; receiving a user query that relates to performing a task, and transmitting the query to the language model; receiving a workflow that is automatically generated by the language model based on the lecture and the query; and executing the workflow in order to generate an output that is responsive to the query.