Dynamic execution of artificial intelligence agents through device management
Systems and methods are described for dynamic execution of artificial intelligence (“AI”) agents. A server can receive, from a client device, an input associated with an AI agent. Based on a manifest file or user profile, the server can identify a management policy that applies to the AI agent. The server then dynamically configures access to the agent objects based on applying the management policy. The management policy is applied to a device status of the client device, a user profile of a user of the client device, and/or a network configuration of the client device. The server then executes a modified workflow based on the dynamically configured access, wherein the modified workflow bypasses or changes operation of at least one of the agent objects. Based on the modified workflow, the server transmits an output to the client device.
1 . A method for dynamically managing execution of an artificial intelligence (“AI”) agent, comprising:
receiving, at a client application of a client device, a request to execute a first AI agent, wherein a manifest file specifies relationships between agent objects of the first AI agent;
identifying, by a second AI agent, a management profile that applies to the first AI agent, the second AI agent being configured to evaluate compliance of the management profile based on at least one of:
a device state of the client device; and
a user profile of a user of the client device;
dynamically configuring execution of the first AI agent, comprising selecting an agent object for inclusion in a workflow based on the compliance evaluation of the second AI agent;
executing the workflow for the first AI agent based on the dynamic configuration; and
causing display of an output of the first AI agent at the client device,
wherein the agent objects comprise at least one of: an AI model object, an AI dataset object, an AI tool object, an AI prompt object, and an AI code block object; and
wherein the manifest file specifies execution linking between the agent objects.
2 . The method of claim 1 , comprising logging the execution of the first AI agent, wherein the logging includes the compliance evaluation and the executed workflow.
3 . The method of claim 1 , wherein the workflow changes operation of at least one of the agent objects.
4 . The method of claim 1 , wherein the management profile includes an agent object policy that is identified based on an identifier in the manifest file, and operation of a first agent object is based on the device state or user profile.
5 . The method of claim 1 , wherein the management profile includes a user management policy that is compared against a group identified in the user profile.
6 . The method of claim 1 , wherein the device state is noncompliant with respect to at least one of device security, operating system, software, and performance.
7 . The method of claim 6 , wherein, based on the noncompliance, the workflow bypasses sending an input to an AI model.
8 . The method of claim 1 , wherein the workflow of the first AI agent is dynamically configured to include local execution of a first AI model at the client device instead of accessing a second AI model over a network.
9 . The method of claim 1 , wherein the management profile specifies a security requirement, and wherein the dynamic configuration comprises comparing the security requirement against a network configuration.
10 . The method of claim 1 , wherein the management profile includes a list of allowed or disallowed applications, and wherein the list is compared against applications installed or executing on the client device.
11 . The method of claim 1 , wherein the management profile specifies use of different AI models for different groups, and wherein the dynamic execution configuration is based on a group identified in the user profile.
12 . The method of claim 1 , wherein the device state includes a location of the client device, and wherein compliance is determined based on comparing the location to a geofence.
13 . The method of claim 1 , wherein the device state is received from a management controller that executes at the client device.
14 . The method of claim 13 , further comprising sending a management controller profile to the client device, wherein information included in the device state is based on the management controller profile.
15 . The method of claim 1 , further comprising assigning tools to an agent object based on the evaluated compliance.
16 . The method of claim 1 , wherein the dynamic configuration includes selecting the agent object based on comparing a policy identifier of the agent object to a management policy.
17 . The method of claim 16 , wherein the policy identifier is part of an agent object policy that applies to the selected agent object, wherein the agent object policy specifies build parameters for the first agent object.
18 . The method of claim 1 , wherein access to a dataset is bypassed based on the evaluated compliance.
19 . A non-transitory, computer-readable medium including instructions that are executed by a processor and cause the processor to perform stages for dynamically managing execution of an artificial intelligence (“AI”) agent, the stages comprising:
receiving, at a client application of a client device, a request to execute a first AI agent, wherein a manifest file specifies relationships between agent objects of the first AI agent;
identifying, by a second AI agent, a management profile that applies to the first AI agent, the second AI agent being configured to evaluate compliance of the management profile based on at least one of:
a device state of the client device; and
a user profile of a user of the client device;
dynamically configuring execution of the first AI agent, comprising selecting an agent object for inclusion in a workflow based on the compliance evaluation of the second AI agent;
executing the workflow for the first AI agent based on the dynamic configuration; and
causing display of an output of the first AI agent at the client device,
wherein the agent objects comprise at least one of: an AI model object, an AI dataset object, an AI tool object, an AI prompt object, and an AI code block object; and
wherein the manifest file specifies execution linking between the agent objects.
20 . A system for dynamically managing execution of an artificial intelligence (“AI”) agent, comprising:
a memory storage including a non-transitory, computer-readable medium comprising instructions; and
at least one hardware-based processor that executes the instructions to carry out stages comprising:
receiving, at a client application of a client device, a request to execute a first AI agent, wherein a manifest file specifies relationships between agent objects of the first AI agent;
identifying, by a second AI agent, a management profile that applies to the first AI agent, the second AI agent being configured to evaluate compliance of the management profile based on at least one of:
a device state of the client device; and
a user profile of a user of the client device;
dynamically configuring execution of the first AI agent, comprising selecting an agent object for inclusion in a workflow based on the compliance evaluation of the second AI agent;
executing the workflow for the first AI agent based on the dynamic configuration; and
causing display of an output of the first AI agent at the client device,
wherein the agent objects comprise at least one of: an AI model object, an AI dataset object, an AI tool object, an AI prompt object, and an AI code block object; and
wherein the manifest file specifies execution linking between the agent objects.