AI Assistant Chat: Interface and FeaturesElite

The first time you open the AI assistant, you’ll see a welcome screen with example request buttons. These are ready-made prompts you can use to get started quickly. Tap the one that fits what you need, and the assistant will handle it.

The chat interface has three main parts: the top bar, the message input panel, and the chat history. To make the assistant chat wider or narrower, grab its left edge and drag it left or right.

Top bar

• Chat title — shows the name of the current conversation; click it to open your chat history. The title is generated automatically by the AI assistant based on your first message in the chat. Hover over the title to see the chat’s last activity: date and time.
• “+” button (new session) — starts a new chat.
• Chat history button — opens the list of saved chats.
• Context indicator — shows the percentage and number of tokens used by the AI model within a single conversation; the limit depends on the model you’re using.
• Close button “✕” — closes the chat interface.

Message input panel

• Conversation context — shows the current app context (project, task, and so on) that the assistant takes into account when handling your requests. It appears above the message input field and helps the assistant give more relevant answers.
• Mode — a switch between Agent Mode and Ask Mode.
• Model — picks which of your connected AI models handles the request.
• Voice input button — turns on speech recognition in the web version and desktop app. The button is always visible. If speech recognition is not configured, clicking it opens the AI Assistant settings.
• Text input field — where you type your requests and commands.
• Send button — sends your message to the assistant.

Chat history

The area in the center shows your messages and the assistant’s replies. As you scroll through the conversation, your messages pin to the top and expand on hover so you can read them in full. As you keep scrolling, the pinned field shows the message at the level you’re currently at. Each message has a copy button, and you can rate the assistant’s latest reply with a like or dislike. When you dislike a reply, you can pick a reason and undo the data changes made by that request with a single button.

You can talk to the AI assistant in plain language — there are no special commands or syntax to memorize. Just write what you need done, the way you’d say it to another person.

For example, instead of hunting for the create-task menu, you can write “Create a task to buy milk tomorrow” or “Add a client meeting on Friday at 10:00 with high priority.”

You can also ask for help: “How do tags work in the app?” and the assistant will answer.

The AI assistant understands your request and carries it out. You can refine the details in the conversation if the assistant misunderstood something or needs extra parameters. If the result isn’t what you wanted, try rephrasing your request, add more detail, or switch to a different model.

The AI assistant works in two modes, and you can switch between them depending on what you need.

• Agent Mode — the full-featured mode with everything for creating, editing, and deleting items, plus advanced search and analytics. Use this mode when you need to perform an action or get a detailed analysis.
• Ask Mode — a lighter mode for simple conversation, answering questions, and reference information about the app. Use it when you just need information or advice rather than a specific action.

In Agent Mode, the assistant may ask permission to work with your data. If you want it to act on the data in your account, grant permissions for the chat or on a permanent basis by clicking the matching button.

Contexts tell the assistant which data in your SingularityApp account to take into account when handling a request. The context changes automatically as you move around the app. Open “Today” and the context becomes “Today”; open a project and that project becomes the context; click a specific task and the context switches to that task. The context you’re in directly shapes the result of the assistant’s work.

For example, if you’re in the “Work” project and say “create a task,” the assistant adds it to that project automatically.

Besides the automatically selected contexts, you can pick and add contexts to a message by hand. The add-context button sits above the input field. You can choose specific items as contexts — a particular project, several projects with subprojects, or several contexts of different types at once.

You can also use special characters to open the context picker right while you’re typing a message.

Type @ in your message to open the list of all available contexts:

• Projects
• Contexts
• Sections
• Tasks
• Tags
• Filters
• Habits

Or type # in your message to open the list of system contexts:

• Inbox
• Today
• Plans
• Someday
• Archive
• Trash
• Search
• Project
• Tags
• Settings
• Task and note filter
• Calendar
• Habits

After you type @ or #, keep typing — the context list filters based on what you enter. Press Enter to select the first match in the list, or use the up and down arrows to move through it and then press Enter. You can add any number of contexts marked with @ and # in a single message.

If you want to send a request that shouldn’t take the context into account, click the “eye” icon next to the context to turn it off.

You can drag tasks, projects, tags, sections, and habits into the chat with the mouse — into the input field or into the message feed. While you hold an item over the field, a hint appears: “Drop to add to the request.”

After you drop the items, they are added to the request as one group. Above the field you can see how many of each were added — for example, “3 tasks” and “1 project.” Large selections work the same way: every selected item goes into the request together.

Voice input works in the web version and in the desktop app. Click the microphone button next to the input field and start speaking. When the system notices a pause, a short countdown starts, and the message is sent automatically. You can also send it by hand with the send button. The recognition method is chosen in the AI Assistant settings, in the “Voice input” section.

The assistant remembers facts about you and uses them in later chats: preferences, your usual order of work, and limitations. Memory is stored on this device only.

When the assistant remembers something, the reply shows a “Remembered” card with a “Forget” button. Clicking “Forget” removes the fact right away.

All remembered facts are listed in the AI Assistant settings, in the “Memory” section. There you can turn memory off with the “Use memory” toggle. Each fact has two icons: a pencil to edit the text and a basket to remove the fact. Hovering over an icon shows “Edit” or “Forget.” The “Clear all” row removes every fact at once, and the app asks for confirmation first.

Below the list you can see how many characters of memory are used. When the limit is reached, the app shows “Memory is full — shorten or delete other facts.”

The assistant takes the whole conversation history into account when it generates answers. Because of that, we recommend starting a new chat for each new question or discussion. If you cover different topics in one chat, the assistant may return inaccurate results as the topics blend together. If you notice it answering in the context of a previous topic, start a new chat.

All your conversations with the AI assistant are saved automatically as sessions. Each session gets an automatically generated title based on the conversation’s content. You can browse your history by your most recently created chats.

The AI assistant also builds a summary of what was discussed and pulls out the key decisions made in the conversation. That helps you jump back to the important moments without rereading the whole thing.