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

# Knowledge and resources

> Give your agent product context, documentation, FAQs, pronunciations, and shareable resources.

Supply the agent with knowledge about your product through the **Knowledge** tab in the Navattic Agent editor. The more context the agent has, the better it can answer questions and demonstrate relevant features.

## Product context

Product context is general information about your product, key features, and value propositions. This helps the agent understand what to demonstrate and when. Click **Upload context** to have Copilot extract context automatically from a website URL or an uploaded document (PDF, plain text, or Markdown file). Copilot shows you a preview of the extracted content before applying it.

## Knowledge resources

Add documentation, support articles, or other materials that the agent can reference. You can add knowledge as a URL, plain text, or an uploaded file.

**Supported file formats for knowledge uploads**

| Category | Formats |
| :- | :- |
| Documents | PDF, Word (.docx, .doc), OpenDocument text (.odt), RTF, EPUB |
| Presentations | PowerPoint (.pptx, .ppsx, .ppt, .pps), OpenDocument presentation (.odp) |
| Plain text | TXT, Markdown (.md) |

One file per upload, up to 512 MB. Spreadsheets (Excel, CSV, Google Sheets) are not supported.

<Note>
  Word, PowerPoint, OpenDocument, RTF, and EPUB files must be under 25 MB. PDF and plain text files can be up to 512 MB.
</Note>

Navattic converts each uploaded document to text before the agent uses it. Headings, lists, and tables are kept, so a pricing table or spec sheet stays readable to the agent.

A large PDF can take a few minutes to import and shows as processing until it is ready. A password-protected document cannot be imported, so remove the password and upload the file again.

When you add a URL, Navattic crawls the site and formats the content for the agent to reference during demos. Full-website crawls appear as a single entry for the hostname. Click the entry to open a page browser where you can browse and search all crawled pages, organized by folder.

To manage a source, hover over it and click the **...** menu. Available actions depend on the source type:

| Action | Text | URL | File |
| - | - | - | - |
| Edit information | Yes | Yes | Yes |
| Refresh source | No | Yes | No |
| Download source | No | No | Yes |
| Flag as outdated | Yes | Yes | Yes |
| Remove | Yes | Yes | Yes |

Flag a source as outdated to mark it for review without removing it from the agent's context immediately.

## Viewing what the agent learned from a source

When you open a URL knowledge source in the drawer, the **Learnings** section shows the facts and information the agent extracted from that source. Each learning appears as a concise statement the agent can draw on when answering questions.

To remove a learning, hover over it and click the delete icon. Deleting a learning removes it from the agent's knowledge for that source. The page count in the drawer updates to reflect the change.

While a source is still processing or analyzing, the Learnings section updates automatically as extraction completes.

## FAQs

Common questions and their answers help the agent respond accurately to frequently asked questions.

## Pronunciations

Control how your agent pronounces specific words during voice synthesis. Use this for product names, technical terms, or brand names that the voice model might mispronounce by default.

Each entry has two fields:

* **Term**: The word or phrase as it appears in your product content (for example, "Navattic")
* **Pronunciation**: A phonetic respelling using hyphens to separate syllables (for example, "nuh-VAT-ick"). Capitalize syllables to indicate stress for your own reference; the agent normalizes the spelling before sending it to the voice model.

Pronunciations are applied automatically to everything the agent says. You do not need to add them to scripts or instructions separately.

## Sharing resources during demos

Resources are links and files you pre-configure in the **Knowledge** tab. The agent can drop them into the chat panel as rich-preview cards at the right moment in a conversation.

The agent decides when a resource is relevant, based on what the prospect is asking about and the **When to share** guidance you provide. It then sends the card into the chat automatically. The prospect sees the title, a short description, and for links, a thumbnail preview. They can open the link in a new tab or download the file without leaving the demo.

A badge on the chat toggle shows how many new resources have been shared while the panel is closed, so prospects don't miss them.

### Adding a link resource

<Steps>
  <Step title="Open the Resources section">
    In the Navattic Agent builder, open the **Knowledge** tab. Scroll to the **Resources** section and click **Add link**.
  </Step>

  <Step title="Enter the URL">
    Paste the link URL. Navattic automatically fetches the page title, description, and preview image. You can edit any of these fields.
  </Step>

  <Step title="Fill in the details">
    * **Title**: The headline shown on the chat card (required). - **Description**: A short summary shown beneath the title on the card (optional but recommended). - **When to share**: Instructions for the agent describing when to surface this resource, e.g. *"When the prospect asks about pricing tiers or wants enterprise pricing."*
  </Step>

  <Step title="Save">
    Click **Save**. The resource appears in the Resources list and is immediately available to the agent.
  </Step>
</Steps>

### Adding a file resource

<Steps>
  <Step title="Open the Resources section">
    In the **Knowledge** tab, click **Add file**.
  </Step>

  <Step title="Upload the file">
    Drag a file into the upload area or click to browse. Supported formats are the same as knowledge base uploads: PDF, Word, PowerPoint, OpenDocument, RTF, EPUB, TXT, and Markdown. One file at a time, up to 512 MB.
  </Step>

  <Step title="Fill in the details">
    * **Title**: The name shown on the chat card (required). - **Description**: A short summary visible to the prospect (optional). - **When to share**: Instructions for the agent, e.g. *"When the prospect asks for the security overview or SOC 2 report."*
  </Step>

  <Step title="Save">
    Click **Save**. The file is uploaded and the resource is available to the agent immediately.
  </Step>
</Steps>

<Tip>
  The **When to share** field is the most important part of each resource. Be specific about the conversation signals that should trigger sharing. Vague guidance leads to the agent sharing too broadly or not at all.
</Tip>

<Note>
  Each resource is only shared once per session. If the agent determines the same resource is relevant a second time, it will not resend it.
</Note>

## Brain suggestions

The **Brain suggestions** panel appears in the **Brain** tab when the agent has unanswered buyer questions from recent sessions. Navattic surfaces these as ranked suggestions so you can fill gaps in the agent's knowledge.

Each suggestion shows the buyer's question and an impact badge: **High**, **Medium**, or **Low**. Click a suggestion to expand it and see:

* The buyer's exact question, with a link to the source conversation
* The number of sessions where the same question went unanswered
* An **Answer** field to write the response the agent should have given

To act on a suggestion, write an answer in the **Answer** field and click **Add to brain**. Navattic creates a knowledge resource from your answer and the agent learns from it. To skip a suggestion without adding it, click **Dismiss**.

The panel hides when there are no pending suggestions. The badge on the panel header shows how many are waiting.


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.