[FEAT] labbotbrain — LLM Q&A cog with role-gated access and read-only tool use

[hwcopeland]

Overview

A new cog to give LabBot a voice. He has had to listen to #general for years. It is now his time to speak.

Includes an persona declaration generated from #general and #homelabquotes text dumps. Direction and persona was then fine tuned based on user suggestions. Current profile reflects "homelab-veteran" which includes a snarky, IT overlord who dunks first, then helps after.

Commands

[p]ask [question]

DoD

• Questions are answered by a LLM
• Context and Persona to assimilate to Homelab
• Role-Gated Permissions

Stretch

• API tooling for additional services (i.e. eBay, CPUBenchmark, LabBot Observability, GitHub, etc. )
• Additional Personas profiles
• Community Funded API
• Rage-Bait (0.5% message hits) - Confidently Wrong, Message output in italics for safety

[hwcopeland]

Will include xAI API key. At current usage, and with role-gating (boosted-labbers, pin of the month, lab pic of the month, mods), I would expect the API credits to last ~6 months. This assumes that the bot doesn't get abused, and would be at <100 req/day.

[goochs]

I feel there are enough snarky IT overlords dunking first and helping after in homelab that we don't need a LLM pretending to be one piling on.

Personally, I fear the low-effort chat that already exists at times is discouraging for newbies and veterans alike, and I don't believe that adding this sort of functionality would improve that.