A method for dynamic isolation and sharing of a multi-brand AI agent platform based on configuration driving

Abstract

The application discloses a kind of based on configuration driven multi-brand AI intelligent agent platform dynamic isolation and sharing method, it is related to multi-tenant AI platform architecture technical field.The method includes: receiving brand registration request and loading layered configuration package including brand performance layer, intelligent agent system layer, business logic layer and data authority layer;According to the identification of request source, identify target brand and route to corresponding configuration space;Load brand performance layer configuration and render brand UI interface;According to intelligent agent system configuration, load brand specific intelligent agent role;According to business logic configuration, route API request;According to data authority configuration, execute data isolation.The application realizes the decoupling of brand layer, intelligent agent layer, business layer and data layer by layered configuration architecture, and new brand can be deployed by modifying configuration only.

Description

Technical Field

[0001] This invention relates to the field of multi-tenant AI platform architecture technology, specifically to a method and system for achieving complete isolation of multi-brand AI intelligent agent platforms in terms of UI presentation, persona system, business logic and data permissions on the same technical foundation through configuration-driven implementation. Background Technology

[0002] Existing AI agent platforms typically employ single-brand monolithic architectures, multi-tenant shared architectures, or microservice architectures. These approaches suffer from the following drawbacks: high brand switching costs, requiring modifications to front-end code, back-end configurations, and databases; hard-coded agent systems, lacking support for hot switching via configuration; coarse-grained data isolation, requiring either complete sharing or complete isolation; and inconsistent user experiences. Therefore, a configuration-driven method is needed to achieve flexible isolation and sharing across UI, agents, business logic, and data layers for multiple brands.

[0003] A search revealed that existing technologies include: traditional multi-tenant SaaS architectures focus on data isolation and tenant management, but do not involve brand-differentiated configuration and hot-loading switching of AI agent roles; white-label SaaS platforms typically only allow changing the logo and brand color, not the complete agent system. The innovation of this application lies in "achieving complete isolation and sharing of AI agent platforms across multiple brands at the UI, agent system, business logic, and data levels through layered configuration packages," demonstrating novelty and inventiveness. Summary of the Invention

[0004] This invention aims to provide a configuration-driven method for dynamic isolation and sharing of multi-brand AI agent platforms. By using a layered configuration architecture, it decouples the brand layer, agent layer, business layer, and data layer, allowing new brands to be deployed simply by modifying their configurations.

[0005] A configuration-driven method for dynamic isolation and sharing of multi-brand AI agent platforms includes the following steps: Step S1 – Brand Registration and Configuration Loading: The system receives a registration request for a new brand and loads the complete configuration package corresponding to that brand. The configuration package includes brand presentation layer configuration (brand name, logo, main color scheme, font, domain name binding), agent system configuration (agent role list, persona prompt, tool permission whitelist, model specifications), business logic layer configuration (API base address, billing strategy, free quota rules, function switches), and data permission layer configuration (user data isolation rules, knowledge base access permissions, dialogue record storage strategy). Step S2 – Request Routing and Brand Recognition: The system identifies the target brand based on the source domain name or brand identifier field of the request and routes it to the corresponding configuration space. Step S3 – Brand Presentation Layer Rendering: The system loads the corresponding configuration based on the brand identifier and renders it as a UI interface specific to that brand. Step S4 – Dynamic Loading of Agent System: The system loads the agent role list, persona parameters, and tool permissions as needed based on the brand configuration. Step S5 – API Multi-Base Address Routing: The system routes the request to the specified API base address based on the business logic layer configuration. Step S6 – Data Isolation Execution: Execute full isolation, shared, or field-based isolation modes according to the data permission layer configuration.

[0006] Further technical solutions: Brand presentation layer configuration and intelligent agent system configuration support hot reloading, and modifications take effect immediately without restarting the service. API multi-base address routing supports three schemes: complete isolation, base address sharing plus identifier isolation, and hybrid mode.

[0007] Beneficial effects: Zero-code brand addition, adding a brand only requires writing a configuration package; dynamic switching, brand configuration supports hot reloading; flexible isolation, data isolation granularity can be configured as needed; unified maintenance, underlying business logic is shared to reduce redundant development.

Claims

1. Claim 1: A configuration-driven method for dynamic isolation and sharing of multi-brand AI agent platforms, characterized in that, The process includes the following steps: receiving a registration request for a new brand, loading the hierarchical configuration package corresponding to the brand, wherein the configuration package includes brand presentation layer configuration, intelligent agent system configuration, business logic layer configuration and data permission layer configuration; Upon receiving a user access request, the system identifies the target brand based on the request's origin domain or the brand identifier field in the request header, and routes the request to that brand's configuration space. It loads the corresponding brand presentation layer configuration based on the identified brand identifier and renders it as a UI interface specific to that brand. Based on the intelligent agent system configuration in the brand configuration, it loads the brand's list of intelligent agent roles, their persona parameters, and tool permissions as needed. Based on the business logic layer configuration in the brand configuration, it routes the request to the specified API base address. Finally, it performs data isolation based on the data permission layer configuration in the brand configuration, supporting three modes: complete isolation, shared isolation, and isolation by field.

2. Claim 2: The method according to claim 1, characterized in that, The layered configuration package includes the following four layers: Brand presentation layer configuration (brand name, logo, main color, font, domain name binding), Intelligent agent system configuration (intelligent agent role list, persona prompt, tool permission whitelist, model specifications), Business logic layer configuration (API base address, billing strategy, free quota rules, function switch), and Data permission layer configuration (user data isolation rules, knowledge base access permissions, dialogue record storage strategy).

3. Claim 3: The method according to claim 1, characterized in that, The brand presentation layer configuration and intelligent agent system configuration support hot reloading updates. The update scope includes the intelligent agent welcome message, color scheme, logo, and function visibility. The update takes effect immediately without restarting the service.

4. Claim 4: The method according to claim 1, characterized in that, The API multi-base address routing supports three modes: fully isolated mode using independent domain names and backend services, base address sharing plus identifier isolation mode using the same domain name for isolation through brand identifier, and hybrid mode with partial service sharing and partial service isolation.

5. Claim 5: The method according to claim 1, characterized in that, The same backend API engine can generate different personas based on different intelligent agent system configuration parameters, including response style, tone and rhythm, and professional ability boundaries.

6. Claim 6: A configuration-driven dynamic isolation and sharing system for multi-brand AI intelligent agent platforms, characterized in that: include: The brand registration module is used to receive brand registration requests and load the hierarchical configuration package; The route identification module is used to identify the target brand based on the request source identifier; The UI rendering module is used to load the corresponding brand interface based on the brand presentation layer configuration. The agent loading module is used to load brand-specific agent roles on demand according to the agent system configuration. The API routing module is used to route requests to a specified API base address and attach a brand identifier according to the business logic configuration. The data isolation module is used to implement multi-brand data isolation based on data permission configuration.