Intermediate Storage for Local LLM Access and Data Privacy

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing technologies fail to efficiently address the challenge of securely storing and invoking large language models (LLMs) while ensuring data privacy and optimizing response times, particularly in local computer systems.

Innovation Solution

An intermediate storage device with a model storage unit, data storage unit, AI selecting unit, and access control unit are implemented, allowing for local invocation of LLMs and secure data access, enabling differentiation between model operation and data access commands, and supporting various application scenarios.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Adaptability or versatility

If LLMs are transmitted through the Internet for processing, then users can access language model capabilities, but personal information or commercial confidences may be leaked

Engineering Contradiction:

Improveaccess to language model capabilitiesVSAvoidinformation leakage risk

Core Design Contradiction:

Adaptability or versatilityVSObject-affected harmful factors

Solution Approach 1:

The patent introduces an intermediate storage device as a mediator between the user's computer and the language model. This device locally stores multiple language models and intercepts commands intended for online processing, routing them to local models instead. This intermediary approach enables language model functionality while keeping sensitive data within the local system, preventing information leakage to external servers.

Inventive Principle:

Principle #24Intermediary (Mediator)

2Loss of time

If multiple language models are stored locally, then response time is accelerated and data security is improved, but storage space and device complexity increase

Engineering Contradiction:

ImproveLLM invocation response timeVSAvoidintermediate storage device structure

Core Design Contradiction:

Loss of timeVSDevice complexity

Solution Approach 1:

The patent segments the language model storage and processing function into a separate intermediate storage device, distinct from the main computer system. This segmentation allows the computer to offload the complexity of managing multiple language models to the specialized storage device, which handles model storage, selection, and local processing independently, thereby reducing the overall system complexity while maintaining fast local response times.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The intermediate storage device is designed with multi-functionality, serving as both a storage unit for language models and an active processing unit that can execute commands locally. It integrates storage, model selection, and inference capabilities in a single device, reducing the need for separate components and simplifying the overall system architecture while enabling fast local processing.

Inventive Principle:

Principle #6Universality (Multi-functionality)

3Reliability

If an intermediate storage device is introduced to store and process LLMs locally, then data security and response time are improved, but the system complexity increases

Engineering Contradiction:

Improvedata securityVSAvoidsystem architecture

Core Design Contradiction:

ReliabilityVSDevice complexity

Solution Approach 1:

The patent merges storage and processing functions into a single intermediate storage device. Rather than having separate storage units and processing units, the device combines both capabilities, allowing it to store language models and execute commands locally without requiring additional communication interfaces or coordination mechanisms. This merging simplifies the system architecture while maintaining data security and fast response times.

Inventive Principle:

Principle #5Merging (Combining)

Data Source

PatentUS20260003776A1Intermediate storage device, computer system, and computer command preprocessing method

Publication Date: 2026.01.01 GIGA BYTE TECH CO LTD

AI summary

An intermediate storage device a computer system, and a computer command preprocessing method are provided. The intermediate storage device includes a model storage unit, a data storage unit, an AI selecting unit, and an access control unit. The model storage unit stores a plurality of language models. The data storage unit stores a plurality of file data. The AI selecting unit selects a language model based a model operation command, and the language model executes the model operation command and generates a generation result. The access control unit connect to the data storage unit and the model storage unit. The access control unit determines whether an input command is a model operation command or a data access command. The input command is the model operation command, the access control unit forwards the model operation command to the AI selecting unit and send the output data to the access control unit.