Assistant Ready
no model
Runs against your own endpoint. Nothing leaves your machine unless you configure a remote API.

Settings

Point this app at a model, then start chatting.
Ollama default: http://localhost:11434 · OpenAI: https://api.openai.com
Ollama examples: qwen2.5:7b, llama3.1:8b, mistral:7b
Kept in memory for this session only — never written to disk.
This is where you shape the assistant's style, tone and job. Make it yours.
Lower = focused and factual. Higher = more creative.
Use a safety-aware system prompt
Keep this on. The assistant will still be fully capable for coding, writing, study, business and automation — it just won't help with malware, scams or harming people.