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

# PDF Reader

The PDF Reader node extracts text content from PDF files with flexible reading modes to handle everything from simple text-based documents to complex scanned files.

<div align="center">
  <img src="https://mintcdn.com/agenthub/d1VfJKPM3_oWRs3O/images/pdf_reader.png?fit=max&auto=format&n=d1VfJKPM3_oWRs3O&q=85&s=fb4cc3631b292fec7380aa4a27127535" alt="PDF Reader" width="800" data-path="images/pdf_reader.png" />
</div>

## Overview

<CardGroup cols={3}>
  <Card title="Standard Reading" icon="file-lines">
    Extract text directly from PDFs at no additional cost
  </Card>

  <Card title="Advanced Reading" icon="brain">
    AI-powered structured extraction optimized for LLM processing
  </Card>

  <Card title="OCR Mode" icon="eye">
    Read scanned documents and image-based PDFs with AI vision
  </Card>
</CardGroup>

## Reading Modes

Choose the right reading mode based on your PDF type and use case:

<Tabs>
  <Tab title="Standard">
    **Best for:** Text-based PDFs with selectable text

    * Uses direct text extraction
    * Fastest processing speed
    * **Cost:** 0 additional credits
    * **Limitations:** Cannot read scanned images or handwritten content
  </Tab>

  <Tab title="Advanced">
    **Best for:** Complex PDFs being processed by AI nodes

    * Structured content extraction
    * Optimized for Large Language Models
    * Page-level chunking
    * **Cost:** +5 credits per execution
    * Handles large documents (up to 5 minute timeout)
  </Tab>

  <Tab title="OCR">
    **Best for:** Scanned documents, image-based PDFs, handwritten content

    * Uses AI vision models for text recognition
    * Processes handwriting and multiple languages
    * **Cost:** 2-20 credits depending on AI model:
      * GPT-5.4 Mini Vision: 2 credits
      * Claude 4.5 Haiku: 2 credits
      * GPT-5.4 Vision: 20 credits
      * Claude 4.6 Sonnet: 20 credits
      * Gemini 3.5 Flash/3.1 Pro: Variable
  </Tab>
</Tabs>

## Configuration

### Required Inputs

<ParamField path="PDF File Name" type="file" required>
  The PDF file to extract text from. This is a **file picker** that allows you to:

  * Upload a new file directly
  * Select an existing file from storage
  * **Dynamically pass in a file** from other nodes (like Google Drive)

  <Note>Only shown when "Use Link?" is disabled. Accepts `.pdf` files only.</Note>
</ParamField>

### Dynamic File Input

To pass PDF files dynamically from other nodes (such as files retrieved from Google Drive):

<Steps>
  <Step title="Enable Dynamic Input">
    Hover over the PDF Reader node and click **"Configure inputs"**
  </Step>

  <Step title="Activate PDF File Name">
    In the configuration panel, enable **"PDF File Name"** as a dynamic input

    <div align="center">
      <img src="https://mintcdn.com/agenthub/d1VfJKPM3_oWRs3O/images/pdf_reader_dynamic_input_config.png?fit=max&auto=format&n=d1VfJKPM3_oWRs3O&q=85&s=a0328d0e066a3b92785a7e9908726cc9" alt="Configure PDF File Name as dynamic input" width="600" data-path="images/pdf_reader_dynamic_input_config.png" />
    </div>
  </Step>

  <Step title="Connect File Source">
    Connect the file output from another node (like Google Drive File Reader) to the PDF File Name input

    <div align="center">
      <img src="https://mintcdn.com/agenthub/d1VfJKPM3_oWRs3O/images/pdf_reader_dynamic_file_connection.png?fit=max&auto=format&n=d1VfJKPM3_oWRs3O&q=85&s=e41ed8037b486d45acbf3a6d274f7538" alt="Connect file from Google Drive" width="700" data-path="images/pdf_reader_dynamic_file_connection.png" />
    </div>
  </Step>
</Steps>

### Optional Settings

<AccordionGroup>
  <Accordion title="Source Options">
    <ParamField path="Use Link?" type="boolean" default="false">
      Enable to read PDF from a URL instead of an uploaded file.

      When enabled, you'll provide a **File URL** instead of uploading.
    </ParamField>

    <ParamField path="File URL" type="string">
      Direct link to a publicly accessible PDF file.

      Example: `https://example.com/document.pdf`

      <Warning>URL must be publicly accessible without authentication.</Warning>
    </ParamField>
  </Accordion>

  <Accordion title="Reading Configuration">
    <ParamField path="Reading Mode" type="enum" default="Standard">
      Choose how the PDF should be processed:

      * **Standard**: Direct text extraction (0 credits)
      * **Advanced**: AI-powered structured reading (+5 credits)
      * **OCR**: Optical character recognition (cost varies by model)
    </ParamField>

    <ParamField path="Specify Pages?" type="boolean" default="false">
      Enable to read only specific pages instead of the entire document.
    </ParamField>

    <ParamField path="Page Numbers" type="string">
      Comma-separated page numbers and ranges.

      **Format examples:**

      * `1-5` (reads pages 1 through 5)
      * `1, 3, 5` (reads pages 1, 3, and 5)
      * `1-5, 8, 11-13` (reads pages 1-5, 8, and 11-13)

      <Note>Page numbers are 1-indexed (first page is page 1).</Note>
    </ParamField>
  </Accordion>

  <Accordion title="Output Format">
    <ParamField path="Split PDF Content by Page" type="boolean" default="false">
      Controls how extracted text is returned:

      * **Enabled**: Returns a list where each item is one page
      * **Disabled**: Returns all content as a single combined text string

      <Tip>Enable this when you need to process pages individually in Loop Mode.</Tip>
    </ParamField>
  </Accordion>

  <Accordion title="Password Protection">
    <ParamField path="Is Protected by Password?" type="boolean" default="false">
      Enable if your PDF requires a password to open.
    </ParamField>

    <ParamField path="Password" type="string">
      The password needed to decrypt and read the PDF file.

      Works with both Standard and Advanced reading modes.
    </ParamField>
  </Accordion>
</AccordionGroup>

## Output

<ResponseField name="PDF Contents" type="string | string[]">
  The extracted text content from the PDF.

  **Output type depends on configuration:**

  * If "Split PDF Content by Page" is **enabled**: Returns `string[]` (list of pages)
  * If "Split PDF Content by Page" is **disabled**: Returns `string` (combined text)

  Each page's content is preserved in order. When combined, pages are separated by newline characters.
</ResponseField>

## Common Use Cases

<Steps>
  <Step title="Simple Document Extraction">
    Extract all text from a standard PDF document at no additional cost.

    **Configuration:**

    * Reading Mode: Standard
    * Split PDF Content by Page: Disabled

    **Result:** Complete document text as a single string
  </Step>

  <Step title="LLM-Optimized Processing">
    Process complex PDFs with tables and formatting for AI analysis.

    **Configuration:**

    * Reading Mode: Advanced (+5 credits)
    * Connect output to Ask AI or Extract Data nodes

    **Result:** Structured content optimized for AI processing
  </Step>

  <Step title="Scanned Document Digitization">
    Convert scanned PDFs or image-based documents to text.

    **Configuration:**

    * Reading Mode: OCR
    * Choose appropriate AI model (Mini models for cost savings)

    **Result:** Extracted text from non-selectable content
  </Step>

  <Step title="Page-Specific Processing">
    Extract and analyze specific pages from large documents.

    **Configuration:**

    * Specify Pages: Enabled
    * Page Numbers: "1-3, 10"
    * Split PDF Content by Page: Enabled

    **Result:** List containing only selected pages
  </Step>
</Steps>

## Credit Costs

| Reading Mode | Additional Cost | Best For                               |
| ------------ | --------------- | -------------------------------------- |
| **Standard** | 0 credits       | Text-based PDFs with selectable text   |
| **Advanced** | +5 credits      | Complex documents for AI processing    |
| **OCR**      | 2-20 credits    | Scanned documents, depends on AI model |

<Info>
  **Cost optimization tips:**

  * Use Standard mode whenever possible to save credits
  * Choose Mini models (GPT-5.4 Mini, Claude 4.5 Haiku) for OCR when quality permits
  * Test with single documents before batch processing
  * Use page selection to process only needed sections
</Info>

## Troubleshooting

<AccordionGroup>
  <Accordion title="Empty or partial text extraction">
    **Problem:** PDF Reader returns blank text or missing content

    **Solutions:**

    * Check if PDF contains selectable text (try highlighting text in a PDF viewer)
    * For scanned documents, switch to **OCR mode**
    * For image-based PDFs, use **OCR mode** instead of Standard
    * Verify the PDF isn't corrupted by opening it in another application
  </Accordion>

  <Accordion title="Password-protected PDF fails to open">
    **Problem:** Error message when trying to read a password-protected PDF

    **Solutions:**

    * Enable **"Is Protected by Password?"** option
    * Enter the correct password in the **Password** field
    * Verify password works by testing in a PDF viewer first
    * Some PDFs have restrictions on copying/extraction - OCR mode may help
  </Accordion>

  <Accordion title="URL-based PDF won't load">
    **Problem:** Cannot read PDF from provided URL

    **Solutions:**

    * Ensure URL points directly to a PDF file (ends in `.pdf`)
    * Verify URL is publicly accessible (no login required)
    * Check URL doesn't expire or require authentication
    * Try downloading the PDF manually to test URL validity
  </Accordion>

  <Accordion title="Processing takes too long or times out">
    **Problem:** PDF processing exceeds timeout limits

    **Solutions:**

    * Use **page selection** to process only needed pages
    * Split large documents into smaller files
    * Consider using Standard mode instead of Advanced for faster processing
    * For very large documents, process in batches using Loop Mode
  </Accordion>
</AccordionGroup>

## Related Nodes

<CardGroup cols={2}>
  <Card title="Extract Data" icon="database" href="https://docs.gumloop.com/nodes/using_ai/extract_data">
    Pull structured information from extracted PDF text
  </Card>

  <Card title="Ask AI" icon="comment-dots" href="https://docs.gumloop.com/nodes/using_ai/ask_ai">
    Query document content with natural language questions
  </Card>

  <Card title="File Reader" icon="file" href="https://docs.gumloop.com/nodes/file_operations/file_reader">
    Read non-PDF document formats
  </Card>

  <Card title="AI Fill PDF" icon="pen-to-square" href="https://docs.gumloop.com/nodes/pdf/ai_fill_pdf">
    Fill PDF forms with AI-generated content
  </Card>
</CardGroup>

## Batch Processing

The PDF Reader node supports **Loop Mode** for processing multiple PDFs in a single workflow.

<Tip>
  When using Loop Mode:

  * Each PDF in the input list is processed independently
  * Consider credit costs when processing large batches
  * Use "Split PDF Content by Page" to handle per-page analysis across multiple documents
</Tip>

**Example batch workflow:**

1. Provide a list of PDF files as input
2. Enable Loop Mode on PDF Reader
3. Each PDF is read and processed sequentially
4. Combine node aggregates all results