> For the complete documentation index, see [llms.txt](https://docs.screendesk.io/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.screendesk.io/debug-with-ai/data-sources.md).

# Data sources

Connect trusted documentation, files, text snippets, and resolved support conversations to Screendesk AI.

Data Sources is your workspace's knowledge library. It gives AI access to selected product documentation, procedures, and support history so recommendations can reference your product's actual behavior.

Open **Settings → Data Sources**. The library has four tabs: **Website**, **Files**, **Text snippets**, and **Tickets**.

{% hint style="info" %}
In this interface, **training** means processing and indexing source content so Screendesk can retrieve it for a relevant question. Adding a source alone does not mean all its content is ready to use.
{% endhint %}

## Choose the right source

| Source            | Use it for                                                | How it is maintained                                                           |
| ----------------- | --------------------------------------------------------- | ------------------------------------------------------------------------------ |
| **Website**       | Help centers, documentation sites, or individual articles | Discover pages, choose what to include, then train; optional weekly retraining |
| **Files**         | PDF, DOCX, TXT, and Markdown documents                    | Upload and train                                                               |
| **Text snippets** | Focused procedures or answers maintained in Screendesk    | Edit the text and save with retraining                                         |
| **Tickets**       | Resolved conversations from connected support providers   | Import and sync as internal case memory                                        |

Start with current, authoritative documentation. Add ticket history when you want examples of how similar cases were handled.

## Connect a website

1. Click **Add source → Website**.
2. Choose whether to crawl a website, read a sitemap, or add a single page.
3. Enter the website or sitemap URL.
4. Use **Advanced options** if you need a source name, audience, crawl limits, or path filters.
5. Submit the source and wait for link discovery.
6. Open **Review & train**. Review the discovered links and select the pages AI should use.
7. Save the selection and start training. Check that the selected pages reach **Trained**.

Discovery finds links; training makes their content available for retrieval. A discovered page is not yet trained knowledge.

### Control what is included

Use **Manage links** to review trained, untrained, missing, failed, and excluded pages. Exclude obsolete articles, unrelated content, or pages the agent should not rely on.

For a large site, use include and exclude path patterns, maximum crawl depth, and the maximum page count. Following external domains is a separate option; enable it only when those sites belong in the source.

### Keep it current

* **Find new links** discovers additional pages for review.
* **Retrain source** refreshes selected content.
* **Auto retrain weekly** schedules recurring refreshes.

Review new and failed pages after a refresh. If a page is **Not found**, check that its URL still exists and is accessible.

## Upload files

1. Click **Add source → Files**.
2. Select PDF, DOCX, TXT, or Markdown files, up to **20 MB per file**.
3. Choose the audience and submit the upload.
4. Wait for training to complete.

Open a file source to inspect the extracted text and training status. Check that the extracted text contains the instructions you expect before relying on it for recommendations.

Uploaded files do not automatically follow changes to the original file on your computer.

## Add a text snippet

1. Click **Add source → Text snippet**.
2. Enter a title and the procedure or answer.
3. Set the source name and audience, then submit it.
4. To update it later, open the source, edit the content, and click **Save and retrain**.

A useful snippet describes the symptom, the conditions where the procedure applies, the steps to take, and how to verify success.

## Import resolved support conversations

Open the **Tickets** tab to add an available support connection.

| Provider      | Imported case history     |
| ------------- | ------------------------- |
| **Zendesk**   | Solved and closed tickets |
| **Intercom**  | Closed conversations      |
| **Freshdesk** | Resolved support tickets  |
| **Freshchat** | Resolved conversations    |

1. Connect the provider in Screendesk's integration settings if needed.
2. Return to **Data Sources → Tickets**.
3. Click **Start import** for the connection.
4. Review the imported count, source status, and last successful sync.

Historical sources sync weekly. Use **Sync now** for a manual refresh. If **Reconnect required** appears, restore the integration connection.

Check **Conversation scope** on the source: imported history may contain public replies and private notes. Treat this material as internal support context.

### How ticket history is used

Resolved conversations are **internal case memory**, separate from trusted documentation. A past workaround may have applied only to a particular customer or an older product version.

AI Resolution can consult similar cases, but a recommended fix must also have support from trusted documentation. Ticket history alone is not sufficient.

Importing Intercom, Freshdesk, or Freshchat history does not enable current-ticket AI Resolution in those helpdesks. See [AI in your helpdesk](/debug-with-ai/ai-in-your-helpdesk.md).

### Review learned article drafts

When learned drafts are available, click **Review learned articles**.

Review the supporting conversations, edit the title and content, and save your changes before publishing. **Publish** promotes the draft to **internal trusted knowledge** and starts indexing it. **Reject** keeps it from being published.

Publishing a learned article here does not publish it to your public help center.

## Choose an audience

| Audience              | Meaning for the current AI Resolution workflow                                               |
| --------------------- | -------------------------------------------------------------------------------------------- |
| **Public**            | Documentation that may support customer-facing information                                   |
| **Internal**          | Trusted knowledge for the internal investigation; not material to copy into a customer reply |
| **Restricted**        | Content excluded from the current AI Resolution documentation search                         |
| **Historical ticket** | Internal case memory, assigned to imported ticket sources                                    |

Setting an audience does not publish the source to a public website. Review content before choosing **Public**.

Tightening access limits retrieval. Broadening access to already indexed content requires retraining.

## Pause or remove knowledge

A website source can be disabled and enabled again. Removing a source stops its use in retrieval while cleanup runs in the background.

Removing a ticket source stops similar-case retrieval. It does not erase notes or recommendations that were already generated.

## If knowledge is not being used

Check that AI processing is enabled, the source is connected and trained, the relevant pages are included, and the audience permits use by the workflow.

Also check relevance: a trained article about a different feature may not be retrieved for the current issue. If AI Resolution has no supported fix, it can require human review instead.

Continue with [AI Resolution setup](/debug-with-ai/ai-resolution.md).


---

# Agent Instructions
This documentation is published with GitBook. GitBook is the documentation platform designed so that both humans and AI agents can read, navigate, and reason over technical content effectively. Learn more at gitbook.com.

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

Perform an HTTP GET request on the following URL with the `ask` and `goal` query parameters:

```
GET https://docs.screendesk.io/debug-with-ai/data-sources.md?ask=<question>&goal=<user_goal>
```

`ask` is the immediate question: it should be specific, self-contained, and written in natural language.
`goal` is what the user is ultimately trying to achieve, the reason they need the answer. Sharing it helps GitBook give you a better, more relevant answer. A goal is most helpful when it describes the outcome the user wants rather than restating the question. For example, with `ask=how do I create an API token`, a goal like `automate deployments from our CI pipeline` lets GitBook tailor the answer to that use case.

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.