[Question]: Why follow a programmatic approach to finding issues when integrating AI ?

[ConsularParadi]

Checklist

• I've searched for similar issues and couldn't find anything matching
• I've included steps to reproduce the behavior

Affected Components

• K8sGPT (CLI)
• K8sGPT Operator

K8sGPT Version

No response

Kubernetes Version

No response

Host OS and its Version

No response

Steps to reproduce

NA

Expected behaviour

NA

Actual behaviour

No response

Additional Information

Could you please guide me on why you have chosen a programmatic approach to error finding instead of relying solely on the LLM to find issue and debug? This may cause hallucination but it is also a more dynamic approach and it also adds more context for the LLM to be able to resolve the issue. It may also help companies detect internal custom errors or even find application level errors.
To reduce the context size, we can extract metadata information based on query and apply filters accordingly.

[AlexsJones]

This project was based on codifying what real SRE have learnt.
We don't believe in solely trusting an LLM to find the right issues. It would need all resources as context to be effective which is costly, slow and ineffective.

We use the LLM's to help simplify error messages.

We now are using LLM's more to suggest fixes and have an early pilot of auto remediation available for our operator https://github.com/k8sgpt-ai/k8sgpt-operator/blob/main/AUTO_REMEDIATION.md