Chat Assistant

Lida now includes a chat assistant available from any page in the app. It uses the Google Gemini API to generate quick responses based on the context you provide.

How to open it

Click the assistant's sparkle icon on the right side of the header. The assistant opens in a side panel that you can close at any time.

What can I ask?

• Summarize tasks and notes that you describe in the message.
• Suggest next steps for something you are planning.
• Get quick answers about productivity or organization.
• Turn plans and projects into task suggestions that are ready to review.
• Look up tasks, events, and notes previously saved in your account.

Responses are streamed in real time, follow the same direct, supportive tone as Lida, and support Markdown formatting (lists, bold text, tables, code blocks, etc.). When the assistant suggests tasks, each one appears in an action card. A task is only created after you click Add task.

Personalization with AI Memory

The assistant knows your name, plan, and summarized AI Memory from your profile (habits and preferences recorded over time). As a result, answers, task suggestions, and plans are adapted to your personal context. This memory is the same one shown on your Profile page and respects your plan's character limit.

Quick Capture nudge

While you type in the Quick Capture bar, Lida watches the text and suggests sending it to the assistant when it detects a question, a very long message without a clear date, or several comma-separated tasks. Tap the orange button that appears above the input or use the Cmd + Enter (Mac) / Ctrl + Enter (Windows) shortcut.

• Questions such as "how do I..." or messages containing a question mark become "Ask AI".
• Lists with two or more commas while the type is Task become "Split Tasks".

The text is cleared from the capture bar, the assistant opens automatically, and the message is sent.

Starter suggestions

When the conversation is empty, the assistant shows three quick suggestions to help you take the first step:

• "Summarize my bills this month"
• "Create a plan based on my profile"
• "How do the app limits work?"

Tap any suggestion to send it automatically.

Questions about the app

For questions about how to use Lida, commands, shortcuts, plan limits, or available features, the assistant checks the app's official documentation before responding. The documentation source is cited alongside the answer so you receive accurate, up-to-date information.

Searching your memories

When you ask about earlier data, the assistant acts as an investigator: it first fixes typos and then runs two searches at once — a vector semantic search with a natural sentence and an exact keyword search with expanded synonyms in your language. Both run in parallel across tasks, events, and notes you can access — including your own items, active Shared Trails, and Custom Lists where you are a member — and the results are merged and ordered by the most recent items. List names and tags also help the AI understand the context. Lida turns item content and context into vectors using Google Gemini.

The assistant treats tasks, events, and notes with equal importance unless you explicitly ask for a specific type. Only the ten most relevant results are used. If neither search finds anything, you simply see a quick "No records found" toast and the conversation stays free of error bubbles. A new or edited item may take a short time to appear while background processing finishes.

Exact and financial queries

For totals, counts, minimum or maximum values, spending evolution, exact periods, and contents of an exactly named list, the assistant uses a deterministic query without vector search. Detail rows are limited to keep the conversation concise, but calculations cover every filtered item and remain separated by currency. Notes with financial data are included; cancelled and migrated items are not.

List names must match exactly without case sensitivity. This prevents Nubank from being mixed with Nubank Archive 2025. Cards are currently identified by Custom List name; Lida does not maintain a separate institution or card field.

Conversation history

Lida keeps the history of your recent messages. When you open the assistant, your latest messages are loaded automatically so the conversation resumes where you left off, even after reloading the app.

Privacy and storage

All sent and received messages are processed by the Google Gemini API and stored in Lida linked to your account. As with insights, the data is used only to generate the real-time response and is not used to train AI models.

Usage limits

To keep the service reliable and sustainable, each plan has a daily chat message limit, reset at midnight in your timezone:

• FREE: up to 2 messages per day.
• PLUS: up to 50 messages per day.
• PRO: up to 200 messages per day.

When the limit is reached, you will see a message telling you to wait until the next day to continue. The limit and reset time use the same timezone as your task and financial record dates.

Launch limitations

Chat can create suggested tasks that you explicitly accept, but it cannot edit or delete items yet. Action cards and their Added status are saved in history and return after a page reload. For multi-day plans, the card shows the scheduled date and Lida saves the task directly on that day while keeping its title clean. If a response fails, your message returns to the input so you can retry without mixing it into the next conversation context.