---
title: "Operate PreciseMD through the built-in agent"
description: "Confirm the agent is available, start from a tutorial example, hand it files to OCR, and manage the thinking toggle and conversation history"
canonical_url: "https://docs.ardge.com/en/knowledge/precisemd/built-in-agent"
---
# Operate PreciseMD through the built-in agent

> Confirm the agent is available, start from a tutorial example, hand it files to OCR, and manage the thinking toggle and conversation history

## Overview

**Agent** is PreciseMD's built-in conversational interface, the first item in the sidebar. Describe what you want in plain language and the agent calls the matching features to do it, showing the parameters and result of every call in the conversation.

This article covers confirming the agent is available, starting your first conversation from a tutorial example, handing it files to parse, and managing conversation history.

## What you'll learn

- Confirm your deployment has the model settings the agent needs
- Start a first conversation from a tutorial example
- Upload a file, have the agent OCR it, and get a preview link
- Toggle **Thinking** as needed, and manage conversation history

**Estimated time: about 15 minutes**

## Applies to

- PreciseMD 1.4.0 and later

## Before you begin

- The model connection is set up — see [Set up the model connection in PreciseMD](https://docs.ardge.com/en/knowledge/precisemd/model-connection)

## Step 1: Confirm the agent is available

1. Click **Agent** in the sidebar

When the agent is available, the input box at the bottom of the page accepts text, with the placeholder "Ask anything…".

![The agent page and its examples](https://docs.ardge.com/screenshots/knowledge/en/precisemd/built-in-agent/step-1.webp)

<warning>

**Without a default VLM you can't send a message:**

While the top of the page shows "This deployment has no default VLM configured.", the input box stays disabled. The default VLM is the VLM connection marked **Default** on the **Model connections** page — check its **Model ID** per Step 3 of [Set up the model connection in PreciseMD](https://docs.ardge.com/en/knowledge/precisemd/model-connection), and confirm that model reads **Running** in AI Model Center.

</warning>

## Step 2: Start from a tutorial example

The *Tutorial examples* panel on the left offers three starter scenarios, each listing the steps it covers.

<table>
<thead>
  <tr>
    <th>
      Example
    </th>
    
    <th>
      What it covers
    </th>
  </tr>
</thead>

<tbody>
  <tr>
    <td>
      <strong>
        Upload a file for OCR
      </strong>
    </td>
    
    <td>
      Uploading a file, choosing the vision correction setting, getting a result preview link
    </td>
  </tr>
  
  <tr>
    <td>
      <strong>
        Set up a data source and a task
      </strong>
    </td>
    
    <td>
      Creating an SMB/NAS connection, setting the model connection, creating an OCR task that reads from it
    </td>
  </tr>
  
  <tr>
    <td>
      <strong>
        Model provider + Playground
      </strong>
    </td>
    
    <td>
      Adding a model connection, picking models in Playground, running OCR and reviewing the result
    </td>
  </tr>
</tbody>
</table>

1. Click any example card

That example's opening question fills the input box, ready to send as is or edit first.

![An example question filled in](https://docs.ardge.com/screenshots/knowledge/en/precisemd/built-in-agent/step-2.webp)

## Step 3: Send a message and watch what the agent does

1. Type your instruction in the input box
2. Press Enter to send

Below the input box it reads "Enter to send · Shift + Enter for a new line". While the agent works, the send button on the right turns into **Stop**; click it to interrupt the run.

Each time the agent calls a feature, a card appears in the conversation showing that call's **Parameters** and **Result**, with a status of **Running**, **Completed**, or **Failed**.

The agent can operate across:

<table>
<thead>
  <tr>
    <th>
      Area
    </th>
    
    <th>
      What it covers
    </th>
  </tr>
</thead>

<tbody>
  <tr>
    <td>
      <strong>
        Data sources
      </strong>
    </td>
    
    <td>
      Managing SMB/NAS connections
    </td>
  </tr>
  
  <tr>
    <td>
      <strong>
        Tasks
      </strong>
    </td>
    
    <td>
      Creating and managing OCR tasks
    </td>
  </tr>
  
  <tr>
    <td>
      <strong>
        Runs
      </strong>
    </td>
    
    <td>
      Running and monitoring OCR jobs
    </td>
  </tr>
  
  <tr>
    <td>
      <strong>
        Results
      </strong>
    </td>
    
    <td>
      Reviewing OCR results
    </td>
  </tr>
  
  <tr>
    <td>
      <strong>
        Playground
      </strong>
    </td>
    
    <td>
      Prompt experiments
    </td>
  </tr>
  
  <tr>
    <td>
      <strong>
        Prompt templates
      </strong>
    </td>
    
    <td>
      Managing prompt templates
    </td>
  </tr>
  
  <tr>
    <td>
      <strong>
        Model connections
      </strong>
    </td>
    
    <td>
      Managing model connections
    </td>
  </tr>
  
  <tr>
    <td>
      <strong>
        Users
      </strong>
    </td>
    
    <td>
      User management
    </td>
  </tr>
  
  <tr>
    <td>
      <strong>
        Access tokens
      </strong>
    </td>
    
    <td>
      Managing personal access tokens
    </td>
  </tr>
  
  <tr>
    <td>
      <strong>
        Overview
      </strong>
    </td>
    
    <td>
      A read-only system overview
    </td>
  </tr>
</tbody>
</table>

![The tool call card and the answer](https://docs.ardge.com/screenshots/knowledge/en/precisemd/built-in-agent/step-3-1.webp)

![The complete answer](https://docs.ardge.com/screenshots/knowledge/en/precisemd/built-in-agent/step-3-2.webp)

<tip>

**When several fields are needed, the agent hands you a form link:**

Creating a data source, a task, or a model connection, the agent may reply with a link. Click it to open that feature's own creation dialog, complete with the folder picker and shared folder list. Finish there and the agent picks up the result and carries on.

</tip>

## Step 4: Upload a file for the agent to parse

1. Click the paperclip icon at the bottom left of the input box, **Attach a file for OCR**
2. Select the file to parse, or drag it onto the input area

Once uploaded, the file's name appears above the input box; click the × beside it to remove it.

1. Type what you want the agent to do and send

The agent puts the file through the parsing pipeline. When it's done, the agent gives you a preview link for that result in the conversation; click it to see the output inline.

<note>

**Where uploaded files go:**

Files the agent uploads land in a **Playground** workspace named `Agent Uploads`, where you can review and delete them.

</note>

## Step 5: Toggle thinking as needed

The **Thinking** toggle at the bottom right of the input box controls whether the agent streams the model's reasoning.

- Off — faster responses, more reliable tool calls
- On — a **Thinking process** section appears in the conversation, and it runs slower

The interface describes it as "Off is faster with more reliable tool calls. On streams the model's reasoning (slower)."

## Step 6: Manage conversation history

The conversation history panel on the right is sorted by last update, and each conversation is titled with the first 40 characters of its first message.

- Start a new conversation: click **New conversation**, or send `/new` or `/clear` in the input box
- Switch conversations: click any conversation in the history panel
- Delete a conversation: hover over it and click the trash icon, **Delete conversation**

Both side panels collapse from the icons in the title bar: **Toggle tutorial examples** on the left, **Toggle chat history** on the right.

![Chat history and the delete icon](https://docs.ardge.com/screenshots/knowledge/en/precisemd/built-in-agent/step-6.webp)

## Related articles

- [Set up the model connection in PreciseMD](https://docs.ardge.com/en/knowledge/precisemd/model-connection)
- [Create a personal access token for AI agents](https://docs.ardge.com/en/knowledge/precisemd/access-tokens)
- [Tune prompts and compare OCR results in Playground](https://docs.ardge.com/en/knowledge/precisemd/playground-prompt-tuning)


## Sitemap

See the full [sitemap](https://docs.ardge.com/sitemap.md) for all pages.