Letronna®GitHub ↗

שימוש ב־LetronnaהאפליקציהThe parts of the window

Settings

Every page of the window's Settings, and what each option does.

העמוד הזה עדיין לא תורגם, ולכן הוא מוצג באנגלית. אפשר לתרגם אותו ב־GitHub. GitHub ↗

Open Settings with Ctrl+, (Cmd+, on macOS), from the account menu in the top bar, or from the command palette. Up and Down move between the pages, and Tab moves through a page. Escape, or the close button, closes it.

PageHolds
Accountwho is signed in, your plan and credit, and who pays
Usagetokens and cost, added up from your sessions
Accountsyour gateways and their keys
Searchwhere the agent searches the web
Modelthe default model, fallbacks and session names
Appearancetheme, film grain and motion
Thinkingeffort, reply style, and what a reply may do
Memorywhat Letronna keeps between chats
Privacythe default mode, notifications, keeping sessions, and Discord
Walletthe agent's wallet
Shortcutsthe Quick Entry key
Pluginsoptional integrations
Aboutthe version, and updates

Account

Signed out, Sign in in the browser opens letronna.ai. If the browser does not open, Open the page again or Copy link. Hosted models and cloud runs need a Letronna account; your own keys do not.

Signed in, the page shows your Name, Email, Plan, Credits and Team.

  • Bill hosted models to: Personal, or a team you are in. One pays for each call, never both. It is the same choice as /team in the terminal. See Teams.
  • Open dashboard opens your dashboard. Refresh reads your account again.
  • Sign out signs out here and in the terminal alike. Your chats and settings stay where they are.

Usage

Tokens and cost of the model calls, added up from the saved sessions. Ctrl+Shift+U opens it too.

  • This session, The last 7 days and By model, the last 7 days add them up, with how much of the input was cached.
  • Insights shows your recent weeks: Tokens by day, Replies by hour of the day, Top tools and Top models.
  • Cache misses in this session lists replies the provider served without its cache, and why.

The same is /usage and /insights in the terminal.

Accounts

Each gateway, with where it stands, its key, and a way to sign in or set it up. See Gateways.

Where the agent searches the web. It asks you before each search.

  • Signed in, it searches through your Letronna account, paid by the search from your credit, with DuckDuckGo when that cannot answer.
  • A service of your own saved here is used instead. Add key or Add address, then Save and use. Reset goes back to the service's official address, and Remove forgets it.
  • Use alone on DuckDuckGo keeps every search off what is paid.
  • Search with DuckDuckGo when no search key is set turns that fallback on or off.

Keys stay on this computer. The Web switch in the composer is different: it is the gateway's own search inside a reply, where it has one. See What the agent can do.

Model

  • Default model: the model for new messages. Pick one in the composer, or type its id here.
  • Fallbacks: models asked in order when yours is overloaded, out of credit or gone. Write gateway:model for one on another gateway. One that fails is passed over for 5 minutes.
  • Save keeps both.

Session names

Name sessions with a model: once a session's first reply is in, a model gives it a short name. Off, it keeps the first line of what you asked.

ChoiceNames it
AutomaticLetronna's quick model for a chat on Letronna; otherwise your side model, or the chat's own
This chat's modelthe model the chat is on

Or type a model of your own and press Save. See Sessions.

Appearance

The theme, film grain and motion. See Appearance.

Thinking

For the session on screen and new ones.

  • Effort: how hard the model thinks. Whatever the model likes, or one of the levels the model offers. Models that cannot be told carry on as they are.
  • Style: how replies are written. Default is as Letronna usually writes; your own styles are listed too. See /style in What the agent can do.
SwitchDoes
Show the thinkingprints the model's reasoning as it arrives, not only the answer
Search the weblets the gateway look things up for an answer, and cite the pages it read
Read replies aloudspeaks each reply as it streams; see Voice
Check changes before finishingafter changing files, asks the model once to build or test before it says the work is done

Memory

What Letronna keeps from one chat to the next. Edit or delete each entry, and write SOUL.md. Ctrl+S saves. See Memory.

Privacy

Agent

Notifications

  • Notify when Letronna needs you: while the window is hidden or behind others, the computer tells you when a reply ends, a card or question is waiting, or a helper or scheduled run is done. Clicking one opens its session.

Sessions

  • Delete sessions older than N days: sessions untouched for that many days are deleted. Empty keeps them all. Pinned sessions and the one in use are kept. Letronna looks at most once a day, when it starts with no other copy open. Press Save.

Discord Rich Presence

  • What to share: Show what I am doing, or Hide activity details. Turn the integration on under Plugins first.

Wallet

The wallet's networks and limits. See Wallet.

Shortcuts

The key that opens Quick Entry from anywhere. See Quick Entry.

Plugins

Optional integrations, each marked Built in or External. Switch one off to unload it.

PluginDoes
ChatGPTthe Codex models your ChatGPT plan includes, signed in with ChatGPT instead of an API key
Claudethe models your Claude subscription includes, signed in with Claude instead of an API key
Discord Rich Presenceshows Letronna on your Discord profile; what it shows is set in Privacy

The choices are kept in plugin-switches.json in your data folder.

About

The version and build. Check for updates looks for a newer version, and Download gets it when there is one. The tray's Check for updates opens this page.

עריכת העמוד ב־GitHub ↗