# Insights

### Recommendation Engine

The recommendations pipeline orchestrates multiple data sources and components to deliver personalised recommendations. The current architecture is designed to achieve data-driven, explainable recommendations and ease of model customisation for the end user

* Nuvolari system fetches a live and curated *opportunity catalog* of yield strategies, DeFi protocols, etc from the execution engine.
* It then generates for the queried wallets a *user profile* and *risk score* addressing 80+ parameters to understand the user behaviour, risk propensity and counterparties
* A feature vector is engineered for each user, combining their portfolio data with the catalog of executable opportunities.
* The system scores all possible user-opportunity pairs with a proprietary machine learning model, producing a ranked list.
* Opportunities are then post processed, to create in detail explanations using SHAP for feature importance and LLMs for the descriptions, the wallet address is never passed to LLMs and each wallet address data remains private.

### Insights

Your personalised recommendations are called insights, can be found in the insights home page, by toggling the details cards, you can discover how the recommendation was generated and how why tailored to you:

* Mood Alignment - explains how the insights matches your [mood](/ai-engine/mood.md)
* Behaviour - describes how your on-chain activity indicates a match with the insights
* Summary - TL;DR of why the insights are recommended to you
* Social Sentiment - Coming soon
* Shared Memory - Coming Soon

<figure><img src="/files/cZIJd84x1Onz6M6Sz82e" alt=""><figcaption></figcaption></figure>

***

<details>

<summary><em><strong>The importance of Trust</strong></em></summary>

Numerous industry reports highlight that most AI agents still operate like “black boxes”

> it’s impossible for users to see which inputs were used, how options were weighed, or why a particular transaction was triggered. This prompts a fundamental question: without visibility into how decisions are made, how can trust in the outcome can ever be established?

**Learn more about Deterministic AI in our latest article -**  [When will Crypto x AI be usable - Nuvolari Canvas Series #2](https://x.com/nuvolari_ai/status/1982874127665897764).

</details>

Unique Risk Assessment for each user is what makes Nuvolari insights personalised - learn more about your [**Mood Profile**](/ai-engine/mood.md).

Each insights is deterministic and actionable - powered by our *Execution Engine* - navigate to [**Shortcuts**](/execution-engine/shortcuts.md) to discover how personalised insights are executed.


---

# Agent Instructions: Querying This Documentation

If you need additional information that is not directly available in this page, you can query the documentation dynamically by asking a question.

Perform an HTTP GET request on the current page URL with the `ask` query parameter:

```
GET https://docs.nuvolari.ai/ai-engine/insights.md?ask=<question>
```

The question should be specific, self-contained, and written in natural language.
The response will contain a direct answer to the question and relevant excerpts and sources from the documentation.

Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.