123SUDODocs
Guides

Settings

General behaviour, appearance, language, the default models each surface uses, keyboard shortcuts, and exporting or restoring your data.

Open Settings from the sidebar, or press ⌘, (Ctrl+, on Windows and Linux). Six sections run down the left.

9xchat Settings with General, Appearance, Language, Default Models, Keyboard Shortcuts, and Data & Exports listed down the left
Settings covers how the app behaves rather than what any single answer says.

General

Thread housekeeping and a few things that apply everywhere.

  • Auto Generate Thread Name names a thread from its opening exchange, so the sidebar reads as a list of subjects rather than a list of timestamps.
  • Auto Archive Threads, with Auto Archive Days, moves threads out of the sidebar once they have been quiet for that many days. Archiving hides a thread; it does not delete it.
  • Thread Message Count sets how many messages a thread displays. This is a display limit, not a memory limit — it changes what you scroll through, not what the model is sent.
  • Auto Open Last Thread reopens whatever you were last reading at startup.
  • Spell Check underlines misspellings as you type in the composer.
  • Voice chat language is the language you speak, used for transcription. Naming it explicitly beats Auto detect on short replies, which are the ones detection most often gets wrong.
  • Automatic Backup takes a local backup on a schedule — Off, Daily, Weekly, or Monthly. The five most recent are kept and are listed under Data & Exports.

Appearance

Theme Appearance is System, Light, or Dark. System follows whatever the operating system is currently set to, including a scheduled switch at sunset.

Language

Sets the interface language. 9xchat ships 19, and the change applies immediately — no restart.

This is separate from Voice chat language in General, which is about the language you speak rather than the one you read, and from the language a model answers in, which follows how you write to it.

Default Models

Each surface starts new work on its own default, because the right model for a long chat is rarely the right one for generating a title.

SettingUsed by
Chat ModelNew chat threads
Quick ModelTitles, summaries, and prompt enhancement
Float Window ModelThe quick assist floating window
Notes ModelNotes and documents
Code ModelThe code editor
Image Generation ModelGenerated images

A default is a starting point, not a lock: you can change the model on any individual thread, and the thread keeps that choice.

Three more providers sit below the model list:

  • Audio Provider and Transcription Provider — 123sudo uses credits and needs no setup; OpenAI uses your own API key.
  • Knowledge Embeddings decides what powers semantic search over your Knowledge Base, and the choice is a privacy decision as much as a quality one. 123sudo uses credits. Local (Ollama) runs on your computer, so the files never leave the device. OpenAI or Gemini use your API key and send the file text to that provider.

Changing the embeddings provider offers to re-index your Knowledge Base, because embeddings from two providers are not comparable. Local indexing is free and private but can take a few minutes on a large library; the hosted options use credits.

Keyboard Shortcuts

Every binding below can be reassigned — click it, press the combination you want, and it saves. Esc cancels without changing anything. If the combination is already taken, the app says so rather than silently leaving two actions on one key.

ActionmacOSWindows and Linux
Open Global Search⌘FCtrl+F
Quick AI AssistAlt+SpaceAlt+Space
Quick Switcher⌘PCtrl+P
Toggle Sidebar⌘BCtrl+B
New Conversation⌘NCtrl+N
Private Mode⌘⇧NCtrl+Shift+N
New Tab⌘TCtrl+T
Delete Conversation⌘⌫Ctrl+Backspace
Open Preferences⌘,Ctrl+,
Send MessageEnterEnter
Add New Line⇧EnterShift+Enter

Send Message and Add New Line are fixed. They are what the composer itself listens for, and unlike the rest they name a physical key rather than an action.

Quick AI Assist needs an active trial or subscription. The other bindings work on a free account.

Data & Exports

Exporting

Export writes a single .json backup. Choose what goes in it — chats, notes, images, prompts, models, agents, skills, tools, MCP servers, attachments, knowledge collections, memories, and app settings — and leave out anything you do not want in the file.

Three of those are deliberately incomplete:

  • Tools and MCP Servers export with their credentials redacted.
  • Settings exports without your model API keys.
  • Knowledge exports collections and their scopes. The files themselves travel under Attachments, so a knowledge export without attachments restores the shape of a collection but not its contents.

Secrets are stripped so a backup can be moved and shared without leaking them, which also means restoring one does not bring your API keys back. Keep those where you normally keep passwords.

Restoring

Import takes a .json file exported this way and restores what is in it. The file is validated first, so a file that is not a 9xchat backup is refused rather than half-applied.

Automatic backups

Recent Backups lists the scheduled backups taken at the frequency set in General, newest first, with Restore and Delete on each. Only the five most recent are kept — each new one displaces the oldest.

If the list is empty, no frequency has been set yet; that setting lives in General, not here.

On this page