Sequential RAG Agent Planning for Deterministic Task Execution

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Current large-language model (LLM) based agents lack determinism and predictability in complex, multi-step tasks, making them resource-intensive and difficult to debug, with high computational and time-wasting costs.

Innovation Solution

A sequential retrieval-augmented generation system comprising four specialized agents: High-Level Planning, Detailed Planning, Action, and Writing Agents, which break down complex tasks into structured steps, progressively adding detail and ensuring logical execution.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Adaptability or versatility

If current LLM-based agents are used to perform complex multi-step tasks, then task completion capability is achieved, but determinism and predictability are lost

Engineering Contradiction:

Improvetask completion capabilityVSAvoiddeterminism and predictability

Core Design Contradiction:

Adaptability or versatilityVSReliability

Solution Approach 1:

The patent divides the complex task into multiple sub-tasks handled by specialized agents (planning agent, retrieval agent, generation agent, evaluation agent). Each agent handles a specific portion of the task, which provides determinism and predictability in each step while maintaining overall task completion capability through the coordinated sequence of these segmented operations.

Inventive Principle:

Principle #1Segmentation

2Adaptability or versatility

If LLM-based agents perform complex tasks with continuous re-planning, then task adaptability is improved, but computational cost and time consumption increase

Engineering Contradiction:

Improvetask adaptabilityVSAvoidcomputational cost

Core Design Contradiction:

Adaptability or versatilityVSUse of energy by moving object

Solution Approach 1:

The planning agent performs preliminary action by creating a detailed plan before task execution begins. This pre-planned sequence of steps reduces the need for continuous re-planning during execution, thereby lowering computational cost and time consumption while maintaining task adaptability through the structured approach.

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The evaluation agent provides feedback on the execution results, allowing the system to verify correctness and make adjustments only when necessary. This targeted feedback mechanism maintains task adaptability while avoiding the computational overhead of continuous re-planning, as corrections are made only at specific evaluation points rather than continuously.

Inventive Principle:

Principle #23Feedback

3Adaptability or versatility

If multiple agents interact in complex systems, then task completion capability is enhanced, but debugging complexity increases

Engineering Contradiction:

Improvetask completion capabilityVSAvoiddebugging complexity

Core Design Contradiction:

Adaptability or versatilityVSDevice complexity

Solution Approach 1:

By segmenting the system into distinct agents with well-defined roles and interfaces, the patent reduces debugging complexity. Each agent can be independently tested and debugged, and the standardized interaction protocols between agents make it easier to trace issues compared to monolithic complex systems.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The evaluation agent acts as an intermediary that coordinates between the planning, retrieval, and generation agents. This intermediary role provides a standardized interface for interaction and error handling, simplifying the debugging process by centralizing the coordination logic and providing clear feedback mechanisms.

Inventive Principle:

Principle #24Intermediary (Mediator)

4Loss of time

If LLM-based agents operate without structured planning, then response time is reduced, but task execution reliability decreases

Engineering Contradiction:

Improveresponse timeVSAvoidtask execution reliability

Core Design Contradiction:

Loss of timeVSReliability

Solution Approach 1:

The planning agent performs preliminary action by generating a detailed execution plan before task execution begins. This pre-planned structure ensures reliable task execution while minimizing actual execution time, as the steps are predetermined and can be executed efficiently without on-the-fly decision-making during the task execution phase.

Inventive Principle:

Principle #10Preliminary action

Data Source

PatentUS20260023934A1Systems and methods for completing complex tasks using sequential retrieval-augmented generation

Publication Date: 2026.01.22 ELSEVIER INC

AI summary

A method of automatically completing a task includes receiving, by a high-level planning agent, a query. The high-level planning agent outputs a high-level plan including a plurality of steps and information needed to complete the plurality of steps. The method also includes receiving, by a detailed planning agent, the query and the high-level plan. The detailed planning agent outputs a detailed plan that includes, for each step, one or more tools and one or more parameters for each tool. The method also includes receiving, by an action agent, the query and the detailed plan. The action agent automatically generates an action agent prompt that includes a function call for each tool and one or more parameters for each tool. The method also includes receiving, by a writing agent, the query and the execution output. The writing agent outputs a description of results of a completion of the tasks.