Using an LLM-based agent to provide self-healing capabilities to a network

1 . A method comprising:

determining, by a device, that an issue identified in a network is eligible for self-healing; and

based at least in part on the issue being eligible for the self-healing, enabling, by the device, self-healing operations to address the issue in a closed-loop in the network, including:

determining, by the device, and using a large language model-based agent,

a root cause of the issue and a set of self-healing actions to address the root cause of the issue, and

triggering, by the device, the large language model-based agent to select one or more large language models to perform the set of actions in parallel, each of the one or more large language models performing the set of actions with an individual optimization criterion.

2 . The method as in claim 1 , further comprising:

determining, by the device, a criticality associated with the issue; and

determining, by the device, and based at least in the criticality, that the issue is eligible for self-healing.

3 . The method as in claim 1 ,

obtaining, by the device, a policy associated with the self-healing in the network;

determining, by the device, and based at least in part on the policy, that with the issue is associated with an application, the application being defined as eligible for self-healing in the policy; and

determining, by the device, and based at least in part on the issue being associated with the application, that the issue is eligible for self-healing.

4 . The method as in claim 1 , further comprising:

determining, by the device, a number of users reporting the issue via a chatbot; and

determining, by the device, and based at least in part on the number of users exceeding a threshold, that the issue is eligible for self-healing.

5 . The method as in claim 1 , further comprising:

determining, by the device, that the set of actions corrected the issue; and

training, by the device, the one or more large language models to perform troubleshooting in the network based at least in part on the set of actions.

6 . The method as in claim 1 , wherein the large language model-based agent selects a particular large language model to determine the root cause of the issue and the set of actions.

7 . The method as in claim 1 , wherein triggering, by the device, the large language model-based agent to select the one or more large language models to perform the set of actions in parallel, further comprising:

triggering, by the device, the large language model-based agent to:

select a first large language model to perform the set of actions to reduce a processing time, and

select a second large language model to perform the set of actions to improve an accuracy.

8 . The method as in claim 1 , further comprising:

determining, by the device, a location in the network associated with the issue; and

determining, by the device, and based at least in part on the location, that the issue identified in the network is eligible for self-healing.

9 . The method as in claim 1 , wherein triggering, by the device, the large language model-based agent to select the one or more large language models to perform the set of actions in parallel, further comprising:

causing, by the device, the large language model-based agent to send an instruction by calling an application programming interface (API) to the one or more large language models to perform the set of actions.

10 . The method as in claim 1 , wherein the set of actions are governed by a policy specified by an administrator for the network.

11 . An apparatus, comprising:

one or more network interfaces;

a processor coupled to the one or more network interfaces and configured to execute one or more processes; and

a memory configured to store the one or more processes that, when executable by the processor, cause the processor to perform operations including:

determining that an issue identified in a network is eligible for self-healing; and

based at least in part on the issue being eligible for the self-healing, enabling self-healing operations to address the issue in a closed-loop in the network, including:

determining, using a large language model-based agent,

a root cause of the issue and

a set of actions to address the root cause of the issue, and

triggering the large language model-based agent to select one or more large language models to perform the set of actions in parallel, each of the one or more large language models performing the set of actions with an individual optimization criterion.

12 . The apparatus as in claim 11 , wherein the one or more processes, when executable by the processor, cause the processor to perform further operations including:

determining a criticality associated with the issue; and

determining, based at least in the criticality, that the issue is eligible for self-healing.

13 . The apparatus as in claim 11 , wherein the one or more processes, when executable by the processor, cause the processor to perform further operations including:

obtaining a policy associated with the self-healing in the network;

determining, based at least in part on the policy, that the issue is associated with an application, the application being defined as eligible for self-healing in the policy; and

determining, based at least in part on the issue being associated with the application, that the issue is eligible for self-healing.

14 . The apparatus as in claim 11 , wherein the one or more processes, when executable by the processor, cause the processor to perform further operations comprising:

determining a number of users reporting the issue via a chatbot exceeding a threshold; and

determining, based at least in part on the number of users exceeding a threshold, that the issue is eligible for self-healing.

15 . The apparatus as in claim 11 , wherein the one or more processes, when executable by the processor, cause the processor to perform further operations comprising:

determining that the set of actions corrected the issue; and

training the one or more large language models to perform troubleshooting in the network based at least in part on the set of actions.

16 . The apparatus as in claim 11 , wherein the large language model-based agent selects a particular large language model to determine the root cause of the issue and the set of actions.

17 . The apparatus as in claim 11 , wherein triggering the large language model-based agent to select the one or more large language models to perform the set of actions in parallel, further comprising:

triggering the large language model-based agent to:

select a first large language model to perform the set of actions to reduce a processing time, and

select a second large language model to perform the set of actions to improve an accuracy.

18 . The apparatus as in claim 11 , wherein the one or more processes, when executable by the processor, cause the processor to perform further operations comprising:

determining a location in the network associated with the issue; and

determining, based at least in part on the location, that the issue identified in the network is eligible for self-healing.

19 . The apparatus as in claim 11 , wherein triggering the large language model-based agent to select the one or more large language models to perform the set of actions in parallel, further comprising:

causing the large language model-based agent to send an instruction by calling an application programming interface (API) to the one or more large language models to perform the set of actions.

20 . A tangible, non-transitory, computer-readable medium storing program instructions that cause a device to execute a process comprising:

determining that an issue identified in a network is eligible for self-healing; and

based at least in part on the issue being eligible for the self-healing, enabling self-healing operations to address the issue in a closed-loop in the network, including:

determining, using a large language model-based agent,

a root cause of the issue and

a set of actions to address the root cause of the issue, and

triggering the large language model-based agent to select one or more large language models to perform the set of actions in parallel, each of the one or more large language models performing the set of actions with an individual optimization criterion.