> ## Documentation Index
> Fetch the complete documentation index at: https://docs.pluvo.io/llms.txt
> Use this file to discover all available pages before exploring further.

# Assistant

> Ask about your financial data in plain language and get answers where every number traces back to its source.

## What it does

The Assistant is the front door for questions about your financial and operational data. You ask in plain language, and Pluvo answers from your connected data, with each important number shown as a link you can click to see exactly where it came from.

The point is trust. Rather than a figure you have to take on faith, every load-bearing number in an answer is backed by a real query against your data, and Pluvo tells you when something in its answer could not be verified.

## How to use it

<Steps>
  <Step title="Ask a question">
    Open **Assistant** and type your question, for example "Do a budget vs actuals for last month" or "Which expense lines are trending over budget?" Press Enter to send, or Shift and Enter to add a line break.
  </Step>

  <Step title="Or start from a suggestion">
    Three suggested questions sit below the box, tailored to the data you have connected. Clicking one fills it into the box so you can adjust the wording before you send it.
  </Step>

  <Step title="Answer any clarifying question">
    If a detail would change the answer materially, the Assistant asks you one question and offers a few options to choose from. Pick one, add a detail if you like, then continue.
  </Step>

  <Step title="Read the answer and check the numbers">
    Click any highlighted figure to see how it was derived, the source it came from, and the inputs behind it. From there you can also open the underlying rows.
  </Step>

  <Step title="Keep going">
    Ask a follow-up in the same chat to go deeper. Your conversation keeps its context, and each chat has its own link you can come back to.
  </Step>
</Steps>

## Seeing the work

Under each answer you can open **Thoughts** to see the research the Assistant did, step by step, in plain language, such as querying your warehouse or reading your data schema. Steps that returned rows can be expanded to preview them.

Each answer also carries a short summary of how well grounded it is, including how many figures trace back to a real query and how many came from your Library. If any figures could not be verified, the Assistant flags them so you can treat them as narrative rather than fact.

## Working with your data

**Browse the rows.** When you ask to see the detail behind an answer, a table panel opens beside the chat. You can sort it, page through the full result, and use **Export** to download the loaded rows as a spreadsheet.

**Save what it learns.** If you ask, the Assistant can save context about your organization so it applies to future conversations, and it can save a query you trust to your **Library** so future answers reuse the exact approved logic. Nothing is saved silently, it only happens when you ask.

## Organizing your chats

Your chats are listed alongside the Assistant, grouped by **Today**, **Yesterday**, **Last 7 days**, and **Older**, with a search box to find one again. You can start a **New chat**, create folders, drag chats into them, and rename or delete a chat from its menu.

## Tips

<Tip>
  Be specific about the period and the part of the business you mean. "Compare Q4 actuals to budget for marketing" gets you further than "how are we doing."
</Tip>

<Tip>
  Click through on a number before you circulate it. The provenance card shows the formula, the inputs, and when the data was last synced, which is the fastest way to sanity check a surprising result.
</Tip>

<Tip>
  Approving definitions in your **Library** and writing **Skills** both make Assistant answers better, because it reuses your approved logic and follows your house rules.
</Tip>

## What's next

Ask the Assistant to build a **Report** when you want an answer written up as a document you can publish and keep.