Ideating #1
Description
Most useful product to me:
well theres actually several
- nice easy dashboard to view what actions have been done on the overall cluster last x minutes
- what the control plane is going to do next
- this seems interesting - is there a way for me to not reinvent the wheel?
I should be able to say "why are there so many pods in my cluster? I only wanted there to be 1 pod per deployment" and it should
a. tell me
b. either tell me how to fix it, or fix it by applying a shell script (dry run first ofc, maybe using the --dry-run=client flag)
Surely there exists a way I could tack on an extra embedding layer that takes a minified version of "overall state" + config files (this should be everything necessary), and spits out next steps.
- what would super easy step 0 be? Just pass in the config file of the relevant deployment, nothing else.
this is a really interesting way to construct the prompt.
So overall product would be cli to pass in (prompt + relevant config file), and get steps for how to fix it.
v0 product:
run locally, flask server.
takes in a config file and a question
returns an answer
Note: no finetuning on model, no vector database. Proof of concept for overall thing. putting guardrails in place.