Rabbid's Lab

🦝 About me

🧪 I run an AI crew. Rick, Morty, Birdperson, Beth and the gang babysit my backups, home lab and gadgets, then report in, in character, every single day.
🎩 My butler swears. Joshua sleeps until he hears his name, then answers out loud with the manners of Jeeves and the mouth of George Carlin.
🧠 Real GPU, zero subscriptions. Started on a laptop with no GPU at all. Now an RTX 5060 runs my local AIs about 10x faster, with free cloud AIs as backup. The internet goes down, the crew keeps working.
🐧 Linux, all in. A few weeks ago I didn't know what sudo did. I cut my teeth on Omarchy (Arch + Hyprland), now run Ubuntu on a real rig, and still break things on purpose in my lab.

📣 Latest

Oct 23, 2026 🦝 New iPad AI chat app "hitchhikers‑guide" lets your PC power the conversation.
Oct 23, 2026 🖖 Bridge released – a Star Trek LCARS UI for your PC with AI crew doors.

🚀 Things I've built

🧪

Ask Rick `try it`

Chat with a PG Rick Sanchez whose AI brain runs right inside your browser. No server, no account.

🦙

Rick & Joshua `ollama run`

Run my Rick or my Carlin-mouthed butler Joshua on your own computer: ollama run 2db2du/rick or 2db2du/joshua. No GPU needed.

🛸

Mission Control

A Rick and Morty themed dashboard and API for my home AI crew: reports, launch pad, offline library search.

🧪

Rick's crew

Scheduled AI jobs on n8n plus a "brain switch": free cloud AIs when online, local AIs when the internet's out.

🎩

Hey Joshua

An offline, always-listening voice butler with a British accent. Silent until you say his name.

🧠

`joshua`

Terminal chat that sends every message to the best local AI model for the job.

🏠

homelab

One command to run self-hosted apps: Nextcloud, Paperless, Mealie, Immich and friends.

🔎

Private search

My own SearXNG search engine, started on demand, that my AIs search through too.

🖖

bridge

Bridge: a Star Trek LCARS control center for a home PC with GPU viewscreen and AI crew doors.

🦝

hitchhikers-guide

An iPad home‑screen app that routes AI chats to your PC, supporting both cloud and local Ollama models.

All free, all MIT licensed. Take them, break them, make them better.

🛠️ What I work with

Linux, Arch, Mint, Bash, Git, GitHub, Docker, Python, Neovim, Obsidian, Cloudflare, Discord, Apple, Bots, Markdown

🧠 The AI stack (one RTX 5060, 8 GB)

Local brains (Ollama on the GPU)gpt-oss 20B for the heavy lifting · deepseek-r1 14B for thinking · qwen3 14B · qwen3.5 9B · qwen3-vl 8B and gemma4 12B (vision) · llama3.2 3B at ~83 tokens a second · qwen2.5-coder 7B · qwen3 1.7B as the router · nomic-embed-text for search

Free cloud brainsOne LiteLLM "brain switch" in front of the free tiers of Groq, Gemini, Mistral Codestral, OpenRouter, Cloudflare Workers AI, NVIDIA NIM (Kimi K3 for the hard thinking) and Ollama Cloud. Falls back to the local brains when the internet's out.

VoiceVosk hears (offline), Kokoro speaks in a British butler voice, Piper as backup. WhisperX for transcripts.

Pictures, music, video, 3DComfyUI on the GPU: Z-Image Turbo pictures, LTX-Video clips and Hunyuan3D models from a photo. ACE-Step for full songs with vocals, plus rembg cutouts, Demucs karaoke and MoneyPrinterTurbo shorts.

KnowledgeAn offline library on Kiwix (Wikipedia, WikiMed, Stack Exchange, survival guides): the AIs read the matching passages, not just the first paragraph.

The crewn8n runs Rick, Morty, Summer, Birdperson and friends on a schedule. Claude is the supervisor.

📜 House rules

If it ain't free, skip itFree and open source first, every time.

Local beats cloudPrivate, self-hosted, works when the internet doesn't.

Prototype firstTest on cheap hardware before committing to a real build.

Nothing at bootUnless I say so. My rig, my rules.