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

# Generate answers

> Where suggested answers come from: your library, your entities, and your evidence—and how we pick between reusing and drafting something new.

When you ask something in Chat, we either **reuse** an answer your team already approved or **write a new one** from the material in your workspace. Here is what that material is; below that is **how we use it**, then **how you steer it**.

## Answer Library

Answers you (or a teammate) saved from earlier questionnaires—**approved** and ready to use again. When the question lines up, we can suggest **the same wording** so you stay consistent with what you already committed to. In Chat this may show up as a **Similar answer** you can copy or open in **Review**.

<div className="bg-[#FDFDFD] rounded-lg flex items-center justify-center overflow-hidden outline-[0.5px] outline outline-black/10">
  <img src="https://mintcdn.com/passionfruit/JDR9w0VtF6zQ6gRN/images/changelog/REL27-answer-library-match.png?fit=max&auto=format&n=JDR9w0VtF6zQ6gRN&q=85&s=3bd607266d3d702ffcedc79caa47b4a6" alt="Similar answer from the Answer Library in chat, with Copy and Review" className="w-full h-auto my-0" width="2566" height="1376" data-path="images/changelog/REL27-answer-library-match.png" />
</div>

## Entities

Structured records for a **company, site, or subsidiary**—and the answer data tied to them. That is how we know *who* the answer is for and we keep search from mixing up different parts of the business.

## Evidence

**Unstructured** files: PDFs, Word, scans, and anything else you upload. You **tag** them so you can filter and find the right document when things pile up. We read these when we need to **draft** a fresh answer and we **cite** what we used.

## How we use it

1. **Search** — We look across the sources above using your question, your questionnaire, and tags—within the limits you set (see below).
2. **Library first** — If we find a strong **Answer Library** match, we suggest that reusable answer (sometimes with a colleague’s note so you know why it fits).
3. **Otherwise we draft** — If there is no good library hit, we **generate** a new answer from **Evidence** (e.g. PDFs), **entity-linked** information, and other approved answers in your workspace. You get **citations** so you can open **Review** and check before you use it.

## Your setup

Here you **select context** for the chat: **select sources** (which types we look at—entity, Answer library, evidence), **customize the setup** (which company or site, and optionally specific files), and keep everything scoped to the right place. You can change it anytime during a session.

All of that happens in **[search configuration](/docs/fill-questionnaires/search-configuration)**—the panel where you pick entities first, then source toggles, then manual file selection if you need it.

Suggested answers always include **citations**; open them when you want to double-check. If something important is missing from your workspace, we may ask for clarification or be less certain about brand-new or unapproved content.
