lO-1 works from live web evidence + its trained core — every claim traceable, every build complete. Chat history, projects, and banks persist in this browser.
⚙ Settings
The mode matrix (Ask = fluid prose + quiet sources; Research = dense tables, zero chat; Build = complete code + Worklog + Dependency Log) is part of the default law — keep it unless you mean to change lO's behavior.
Reply layout (lO-1's gathered answers)
Answer-first = fluid prose, the direct answer leading, sources quiet at the bottom. Research vault = dense Markdown matrices (master matrix + timeline). Compact = evidence only.
lO-core · local model vparams
loading…
🔌 Connection & keys
Connectivity …
lO-1 needs only an internet connection and a writable disk. Red light = no connection (chat falls back to the offline banks). No other service is required for max-quality output.
AI (OpenAI-compatible) disconnected
Saved providers (quick-switch — creds remembered per provider in this browser):
Free Cloudflare Worker proxy: deploy cloudflare-worker.js (ships with lO) — one worker serves ALL providers via the X-Target-Upstream header; paste its URL as the Base URL.
Connect GitHub not connected
Hugging Face not connected
Optional — Space management and uploads (publish projects to your HF Space right from the workspace).
Snippet library
All fields are saved to localStorage on this device. No server, no sync.