Enterprise ERP hybrid authorization and AI calling agent method and system supporting offline deployment

Abstract

The application discloses a kind of enterprise ERP hybrid authorization and AI calling agent method and system supporting offline deployment.The method loads RSA signature license file, and machineId is empty when activated;Local signature is verified when starting, and machine fingerprint is activated and bound once online when activated;Offline verification function code is run;AI request is forwarded after being checked customer subscription, scene white list, machine code binding and data sensitive level by proxy gateway.The application makes intranet ERP zero continuous networking available, and prevents AI from calling beyond authority and License illegal migration.

Description

Technical Field

[0001] This invention relates to the field of enterprise software licensing and artificial intelligence service agency technology. Background Technology

[0002] The enterprise ERP is usually deployed on the internal network and cannot be continuously verified online.

[0003] The existing solution has no intranet availability or AI gateway linkage. Summary of the Invention

[0004] Load RSA signature license; if machineId is empty, activation is pending. Start local signature verification; if activation is pending, activate and bind machine fingerprint in one go. Runtime is purely offline function code verification. AI forwards after verifying subscription, sceneCode, machineId, and DataLevel via proxy. Corresponding systems include client, LicenseServer, and AiProxy. Attached Figure Description

[0005] Figure 1 Architecture; Figure 2 activation; Figure 3 AI verification; Figure 4 migrate. Detailed Implementation

[0006] LicenseService offline signature verification.

[0007] Activate LicenseStartupValidator.

[0008] Binding migration.

[0009] RequireFeature function code.

[0010] AiProxyService uses multi-level verification for forwarding and billing.

Claims

1. A method for hybrid authorization and AI invocation proxy in enterprise ERP that supports offline deployment, characterized in that, include: The system loads an authorization license file using an asymmetric digital signature. When the machine identifier in the license file is empty, it is marked as pending activation. Upon application startup, the signature and validity period of the license file are locally verified. When in the pending activation state, a one-time activation request is sent to the authorization server. This activation request carries the local machine fingerprint. The system receives the license file with the machine identifier written and re-signed by the authorization server and saves it locally. During application runtime, the authorization code is verified offline based on the locally saved license file, preventing network requests to the authorization server from being initiated at runtime. When the ERP application requests AI services, the proxy module of the authorization server verifies the consistency of the customer identifier subscription status, scene code whitelist, and machine identifier with the binding record. After successful verification, the request is forwarded to the AI ​​service provider.

2. The method according to claim 1, characterized in that, Maintain a one-to-one binding record between customer identifiers and machine identifiers, and provide unbinding and reauthorization migration processes.

3. The method according to claim 1, characterized in that, The local verification includes using an RSA public key to verify the license file signature and check the license validity period and the maximum number of users.

4. The method according to claim 1, characterized in that, The function authorization code verification is implemented through the API entry feature tagging and function code normalization module. Unauthorized functions return a rejection response.

5. The method according to claim 1, characterized in that, The agent module verification also includes: verifying the paid validity period or trial validity period of the AI ​​subscription; and verifying that the scenario code is in the list of allowed scenarios.

6. The method according to claim 1, characterized in that, The proxy module rejects calls to the public cloud big language model based on the data sensitivity level of the request, which exceeds a preset level.

7. The method according to claim 1, characterized in that, The proxy module verifies the integrity of the assistant message and tool call message chain in the multi-turn dialogue before forwarding.

8. The method according to claim 1, characterized in that, Record the token usage for AI calls and associate it with customer identifiers for billing purposes.

9. The method according to claim 1, characterized in that, The front-end documentation and change log dynamically filter unauthorized module sections based on the list of licensed features.

10. A hybrid authorization and AI invocation agent system for enterprise ERP that supports offline deployment, characterized in that: include: The licensed client module is configured to perform the local verification and function code control as described in claims 1 to 4; The license server module is configured to perform activation, binding, and reauthorization; the artificial intelligence agent module is configured to perform the verification and forwarding as described in claims 5 to 8.

Images