(RFE from Shirly Radco, See: https://github.com/openshift/lightspeed-service/issues/889)

As an SRE/manager I want to be able to use the chat bot to ask in natural language what I'm interested about and that get the PromQL query from it that I will be able to just copy and paste to Grafana/OCP metrics page.

Describe the solution you'd like
A clear and concise description of what you want to happen.

Prometheus is installed by default in most of the the OCP installation.
It would be useful to integrate and teach the the LLM how to query Prometheus for our users.
Prometheus enables the users to query information for current state and also for a period of time.
For example a question that a user can ask the AI:
write a PromQL query, Show me the trend of requests latency in the network during the last day.

I think that a plain Prometheus LLM would not be sufficient since we need to teach the module the OCP language vocabulary. For example, what is a VM and a Virtual Machine and how they are connected to the metrics.

There are two fundamental approaches to this:

1) fine tuning a model by adding promql syntax understandings as well as teaching it the OCP metric names/descriptions
2) providing some RAG references that define example promql queries for answering specific questions, along w/ descriptions of those queries.

(2) is simpler to implement but likely will only result in good answers for questions very similar to the set of example queries provided in the RAG index.