Skip to content

AI and system pages

The main AI settings page with four tabs: Main (route summary, image generation, usage statistics), Models (providers, connection test, the built-in model, voice and image models), Router (task → provider → model) and Advanced (background mode, diagram type, GPU acceleration).

The full walkthrough lives in First-time setup and Models, router and privacy.

What it knows about me — the profile, topics and links. The detailed state shows: active facts, sources, work in progress, what was reused, assembled without AI, parsed locally, left without a source, rejected by quality rules, and how many sources failed.

Chat memory — the messages the agent uses as context: active records and deletion candidates (old ones never reused).

Index diagnostics — the state of models, the queue and the search index. The screen says it plainly: the parameters tune themselves, come here only when search misbehaves.

Housekeeping: integrity checks, updates, diagnostics.

A log of model calls: task, provider, model, tokens, latency. Useful when you need to know where the spend came from or why an answer came from a different model.

Shell command history and the context budget. The page is hidden in some builds.

The built-in help: categories, search and the articles themselves — the same catalogue the agent answers from. If the agent claims a feature doesn’t exist, the article is probably missing, not the feature.