Michael Limberger
Need me? Email mike@limberger.ca
AI
Quick Reference
The card you keep nearby
Everything from tonight. Print it or bookmark it.
Show me.
Ollama: brew install --cask ollama.
OpenWebUI compose:
services:
open-webui:
image: ghcr.io/open-webui/open-webui:main
ports:
- "3000:8080"
volumes:
- open-webui:/app/backend/data
restart: always
volumes:
open-webui:
Then docker compose up -d.
Open http://localhost:3000.
The first account is admin.
ollama pull qwen3:30b-a3b 19GB General, code, reasoning
ollama pull qwen3-next 50GB Heavyweight, long context
ollama pull mistral-small3.2 15GB Text + vision, multilingual
ollama pull qwen3-vl 6GB Vision specialist
ollama pull qwen3-vl:4b 3.3GB Light vision
8GB: qwen3-vl:4b only.
16GB: qwen3-vl 8B, one small model at a time.
32GB: 30b-a3b, mistral-small3.2, qwen3-vl.
64GB+: all of the above, including qwen3-next.
Workspace, Models, Create: name, base model, system prompt, starters, parameters, capabilities, save.
Temperature is the creativity dial (0.2 precise, 1.0 creative, default about 0.7).
Top P is the filter (default 0.9).
Top K default is 40.
Repeat penalty: 1.0 off, 1.1 mild.
Context is memory and it costs RAM.
Max tokens -1 means unlimited for a single reply.
Seed 0 is random.
Starting points: coding 0.2/0.9, conversation 0.7/0.9, creative 0.9/0.95, facts 0.3/0.85.
Variables: {{CURRENT_DATE}}, {{CURRENT_TIME}}, {{USER_NAME}}.
Enter sends.
Shift+Enter is a newline.
Ctrl+Shift+O is a new chat.
Ctrl+Shift+S toggles the sidebar.
Tools: Workspace, Tools, paste Python, assign via bindings.
Built-in web search can use a Perplexity API key; community tools often default to sonar-pro.
Compose commands from the yml folder: up -d, down, logs -f, ps, pull && up -d.
down -v wipes everything.
Nothing loads: docker compose ps.
No models: ollama list.
Cannot connect: Admin, Connections.
Slow: memory pressure.
Gibberish: lower temperature.
Forgetting: new chat or raise num_ctx.
Docs: docs.openwebui.com, ollama.com, ollama.com/library.