Everyday use

Your Account & Preferences

Open Account from the sidebar or go to /account. These settings apply to your account unless noted otherwise. Most save automatically; the active section shows its save state. Installed plugins may add their own account sections.

Account profile

The Account section contains your profile, linked platform identities, and display preferences.

Name and Email update your profile.
Upload avatar accepts PNG, JPEG, WebP, and GIF. Avatars are visible to every signed-in account of the instance.
Interface scale ranges from 80% to 150% in 5% steps; the default is 100%.
Visual effects offers Auto, Full range, Reduced, and Off. Auto follows your operating system's reduced-motion preference.

Interface scale and visual effects are stored in your browser, so they can differ between devices. The language switcher is in the top bar and supports English, Čeština, and Slovenčina. On a phone the chat page leaves it out to make room for the conversation, so switch the language from any other page. When signed in, Elowen also remembers your language for account-scoped alerts and push notifications.

Linked platform identities

Link your platform identity so Elowen can resolve messages to your account:

Discord ID is your numeric user ID. In Discord, right-click your profile and choose Copy User ID.
Microsoft Teams identity is your Entra object ID or a 29:... sender ID. Teams normally resolves this from your verified platform email.
Telegram ID is your numeric user ID; send /start to @userinfobot to find it.
WhatsApp number uses international format with digits only, such as 420111222333.

Each platform identity can belong to only one Elowen account. These fields save automatically.

Personal plugin settings

An enabled plugin may add a personal configuration form or custom panel to Account, under Personal plugin settings. A plugin can instead show its personal settings on its own page. These values belong to your account and do not change the instance-wide settings under Settings → Plugins. Non-secret values save automatically. A secret saves when you leave its field or press Enter; an empty field deletes the stored secret, and Esc cancels a replacement. A disabled plugin or one no longer granted to you is not available in Account.

Models

The Models section sets personal model and execution preferences for Web chat and elowen chat. Instance-wide providers and runtime options are under Settings → <assistant name> AI.

Default model

Primary model selects the fallback for chats without their own model choice. Choose Inherit to follow the instance default. A conversation keeps its selected model across reconnects and routine respawns. A project-specific preference takes priority when applied to a new conversation. Use the conversation model picker or /model to switch explicitly.

Saving a model applies through the normal session respawn. If the model is unavailable or no longer allowed, Elowen uses an applicable project, account, or server default.

Reasoning effort

Reasoning effort sets how much effort the selected model spends before answering. Available values depend on the model. Default leaves the choice to the model; an unsupported saved level is cleared when you change models. Higher effort can take longer and use more tokens.

Vision model

Vision fallback model is an optional fallback for image-bearing turns when the selected chat model cannot process images. Leave it set to inherit to use the primary model; if that model cannot read images, Elowen says so next to the option.

Fast mode

Fast mode is chosen per model: select the models that should use priority processing. Only models whose provider supports Fast are listed: ChatGPT sign-in, OpenAI and Azure OpenAI API keys and Anthropic API keys, for the models the model catalog lists with Fast mode. Models on a Claude subscription never qualify. The catalog list is refreshed together with the model catalog, including from Refresh catalog, so a new model appears without an update. When none of your configured models qualifies, the row says that Fast mode is not available for your models and names the routes that would make it available. A model you did not select never receives Fast options, and the status line shows FAST only while the model of the current conversation has it on. Live sessions use a changed selection on their next provider request. /fast turns Fast on or off for the model of the current conversation only and reports when that model does not support it.

Automatic compaction

When a conversation approaches its context limit, Elowen can summarize older context to make room.

Automatically shorten context is on by default.
Shorten when context is full sets the threshold that starts it. The default is 80%; accepted values are 30–95%.
Compaction per model overrides the global threshold for selected models.
Compaction model chooses the model that writes summaries, including for /compact. If left empty, every provider uses the conversation model, including ChatGPT OAuth. An unavailable or revoked pick also uses the conversation model.

See Brain & Chat for more about compaction.

Execution safety

YOLO mode (labeled Automatically approve requests to use tools (YOLO)) skips interactive approval prompts. It does not override rules that explicitly deny a command or tool, and it does not override Block in unattended runs.

Unattended runs controls rules that would normally ask for approval when nobody is present, such as scheduled jobs, platform conversations, or sub-agent work:

Allow is the default and lets an ask rule proceed automatically.
Block stops an ask rule in unattended work.

Explicit Allow and Deny rules still apply.

Command permissions

Command permissions manages personal rules for shell commands and Tool rules for tool names. Choose Allow, Ask, or Deny for each pattern. The last matching rule wins; a new rule is added at the end, and adding an existing pattern moves it to the end. Elowen ships no built-in restriction on shell commands: a command that no rule matches runs without asking. Only your own rules, and the limits of unattended runs and Plan mode, restrict or prompt for commands.

git status*   Allow
rm *          Deny

Commands approved with Always allow also appear here.

Voice

Account → Voice holds your personal settings for voice calls and dictation. Changes save automatically. Calls and dictation need voice mode enabled by an administrator with an eligible provider, plus voice access on your account. When voice is unavailable, the section says so, and you can still change your language and call-back preferences.

Voice chooses the voice Elowen uses on calls from the voices your instance offers. Press the speaker icon beside a voice to hear a sample.
Language sets the language for calls. Choose Account language to follow your account's language, pick Czech, Slovak, or English, or choose Other language and enter a language code.
Call me back lets Elowen call you after a voice brief when it needs an answer or has the result.
Stop dictation after a pause ends dictation by itself after 3, 5, 10, or 20 seconds of silence, or never. The text stays in the message field, and you can send it at any time. The default is 5 seconds.

Personality

Under Account → Personality, choose a communication style or add Agent instructions, optional Markdown guidance for how Elowen should work with you. The default style is Concise; other choices are Professional, Friendly, and Detailed. Both settings save automatically and apply to your account's conversations.

For tone, detail and format, a specific request in your current message takes precedence over your saved writing preferences, which take precedence over the selected style and default guidance. Friendly can still be brief, and Concise gives a full explanation when you ask for one. Elowen matches a deliverable to its audience, so a formal email can accompany a friendly chat reply. These preferences guide communication; they do not change permissions or how thoroughly Elowen works and verifies a result. Code-mode keeps its own work prompt and uses the shared communication preferences through the existing persona composition.

Memory

Under Account → Memory, manage these account preferences:

Search memory after each message adds relevant memories to the reply context; on by default.
Search for memories again while preparing a reply searches again as Elowen works; on by default.
Save new memories after each reply lets the background curator save durable facts; off by default.
Learn skills from completed work lets Elowen save a reusable way of working as a personal skill. It appears only when your instance offers skill learning; off by default.

Automatic memory maintenance

Automatic memory maintenance merges similar notes, corrects outdated facts and shortens memories so they stay useful. It is off by default; while it is off, nothing runs by itself.

When it is on, a pass runs in the background once enough time has passed since the last one and enough of your other conversations hold a finished exchange. You can also start a pass yourself with Run now, which ignores those conditions. Maintenance model chooses which model reconciles the memories; Same model as the conversation follows the model of the conversation the pass runs from. A chosen model that is no longer offered is shown as Unavailable and is never swapped for another.

Under the switches, Past maintenance passes lists when each pass ran and what it changed. Open one to see the before and after of every memory it touched, with a link to the Memory register. Stop ends a running pass; memories it already changed stay changed, and a memory it moved to the trash can be restored there. A pass works only on your own personal memories, never a shared Project pool, and it never removes a memory permanently: it can merge, rewrite or move memories to the trash. Your conversations see only a short note when a pass improved something.

Changes apply to subsequent work without restarting Elowen. Manage individual memories and categories on the Memory page.

Terminal

Account → Terminal controls the appearance and behavior of elowen chat. The Web UI previews these settings; changes save automatically.

Colors: Auto uses the dark terminal palette; Custom lets you edit the palette. Presets include Elowen Dark, Elowen Light, Solarized Dark, Dracula, and Gruvbox Dark.
Font: System, Menlo / Consolas, IBM Plex Mono, or Courier; size 10–20 px.
Cursor: Block, Bar, or Underline, with optional blink.
Show reasoning (Thought) rows in the CLI shows or hides streamed thinking.
Prompt history sets how many sent prompts the CLI keeps for recall with the up arrow; range 20 to 1,000; default 100.
Double Esc window sets how soon a second Esc must follow the first to interrupt a running turn; range 0.5 to 5 seconds; default 1.8 seconds.
Scrollback sets how many lines of history the terminal keeps; range 500 to 50,000 lines; default 1,000.

Notifications

Under Account → Notifications:

Notifications on this device enables Web Push for the current browser or device. Enable it separately on each device and allow notifications in the browser.
Minimum push notification severity selects Off, Info, Warning, or Critical. The Alerts bell still shows alerts when push is off.
Muted sources accepts comma-separated source names, such as sandbox. Those alerts remain in the bell but are not sent as push notifications.

The top-bar Alerts bell shows durable alerts about startup, plugin or service problems, usage thresholds, and disk or recovery issues. Opening it marks alerts as read. Use Mark all read or a row's Mark read action to acknowledge them; use Dismiss to hide a row. Dismissal does not clear the underlying issue.

API tokens

Under Account → API tokens, create personal tokens that let your own scripts and CI drive the agent over HTTP, see what each token has done, set its limits, and delete it. See Agent API.

Security and saving

Under Account → Security, change your password by entering the current password, then the new password twice. A new password must contain at least 8 characters.

Two-step verification

In the same Account → Security section, Two-step verification protects password sign-in with a six-digit code from an authenticator such as iPhone Passwords or Google Authenticator. The controls are hidden for SSO-only accounts. If an older account uses both Microsoft and a password, sign in with its password once before the controls appear. Microsoft sign-in remains subject to the organization's policy; Elowen does not claim Microsoft enforced MFA.

Choose Enable two-step verification and enter your current password.
Scan the QR code or copy the Setup key into the authenticator. If the QR cannot be generated, use the manual key.
Enter a current code and choose Confirm and enable. Other browser sessions are signed out.

The setup key is shown only during this setup. Leaving the section, cancelling or expiry discards it; an unfinished setup can be cancelled and restarted, not reopened to reveal its key. At the next password sign-in, enter the authenticator code in the same login card. Use Back to restart sign-in. Keep leading zeroes and wait for a new code after using one.

A verified session remains trusted until the existing session lifetime expires, or until logout, cookie removal or revocation. There is no separate remembered-device setting. Anyone who steals that session cookie can access the account during that time; a long lifetime increases this risk. Signed-in sessions names each session by the browser and operating system it was signed in from, with its start and expiry times, its sign-in method (Password, Password and authenticator or Microsoft) and the current session marked; sessions started before this was recorded show Unknown device. Revoke session signs out the selected session and briefly reconnects the account's other browser connections. API tokens and paired computers are unchanged.

Disabling requires a new, unused authenticator code and signs out every browser session, including the current one. There are no recovery codes, password bypass, hidden reset or CLI recovery. If the authenticator is lost, a different administrator must independently verify the person and explicitly reset their factor in Users → Account. Before an administrator enables this feature, arrange another administrator or usable Microsoft sign-in. A lone administrator without either has no supported recovery.

Profile, identity, model, voice, personality, memory, terminal, and non-secret plugin settings save automatically. A password change needs its Change password button. A plugin secret saves when you leave its field or press Enter. Enabling device notifications also requires browser permission.

Account settings do not change other users' preferences. Administrators manage instance-wide options, users, models, and plugins under Settings and Users & Access.

For terminal use, see the CLI guide. For commands, see Slash Commands.