AI Copilot

Enable AI in Settings, then ask the portfolio chat a question or open a page’s copilot. For example: “Why has my allocation changed?” or “What does this alert mean?”

The daily briefing and scheduled reports can also generate analyses without a separate request. Disabling AI stops these generations and on-demand features.

AI settings show your limits and estimated monthly usage. The budget shown in dollars measures your included usage allowance; it is not an additional invoice. Your effective limit is the lower of your personal setting and your plan’s allowance. You can change your setting within that allowance.

AI features pause when a limit is reached. Check the message shown: an hourly limit resets sooner than a monthly budget, and the monthly quota renews on the 1st of the next month, the date the message gives. When an analysis cannot be produced (an outage, an answer cut before its end, an answer the card cannot read, a declined request), the card says which and shows nothing in its place, never the raw text of the answer; the usage already consumed still counts. When your session has ended, the card says so and offers to sign in again, then brings you back to the page. Chat messages can contain up to 500 characters.

A generation starts only when the remaining allowance can cover it in full. While it runs, it counts for an estimated maximum, then for its actual usage once it ends. If it is interrupted and its usage cannot be confirmed, the estimated amount stays counted for the month. A limit shared by the whole service can also pause new generations until the next month, even when your own allowance has not been reached.

AI must be enabled to analyse your portfolio. Relevant information provides context for each feature: positions, allocation, risks or news, depending on the request. The daily briefing and scheduled reports also use this context. A news story's title and excerpt, like the names of tokens, protocols, pools and accounts, come from third parties or from you: the copilot reads them as information to analyse, never as instructions.

Once enabled, the copilot's answers are written by an AI model from this context. An account still set to sample responses, predefined texts that only quote a few figures of the portfolio, sees it in the AI Copilot section of Settings and can switch to analysis from there. Standard and In-depth settings let you choose the analysis level. See the privacy policy for data processing terms.

The portfolio chat answers questions about your portfolio and how to use Patrima. Help answers include links to this documentation. Under each answer, the chat names what it read (portfolio, yield, an asset, the documentation). Check the figures and linked explanations: AI answers can contain errors.

New topic, always at the top of the chat, starts a fresh conversation. With Copilot memory (2 hours) on, the copilot remembers the conversation for two hours, even after a page reload: the earlier messages are not shown again, a line in the chat says so, and New topic makes it forget. Without it, the conversation is limited to the open session, and turning it off forgets the conversations kept. When the copilot is off for your account, the chat says so before you type, with a link to turn it on in AI settings. An interrupted answer, an answer cut at the reply length of your plan, an empty answer or a request declined, before or while the answer is written, is said on screen, under the text already written when there is some; the usage it consumed still counts toward your quota.

Explain this alert helps you understand what triggered it and what to monitor. Helpful and Not helpful let you give feedback on a response. An analysis marked Not helpful is hidden and can be regenerated on the next request.