Back to Skills
    🦞

    llmwhisperer

    Extract text and layout from images and PDFs using LLMWhisperer

    By @gumadeiras
    View on GitHub
    SKILL.md
    ---
    name: llmwhisperer
    description: Extract text and layout from images and PDFs using LLMWhisperer API. Good for handwriting and complex forms.
    metadata: {"clawdbot":{"emoji":"📄","scripts":["scripts/llmwhisperer"]}}
    ---
    
    # LLMWhisperer
    
    Extract text from images and PDFs using the [LLMWhisperer API](https://unstract.com/llmwhisperer/) — great for handwriting and complex forms.
    
    ## Configuration
    
    Requires `LLMWHISPERER_API_KEY` in `~/.clawdbot/.env`:
    ```bash
    echo "LLMWHISPERER_API_KEY=your_key_here" >> ~/.clawdbot/.env
    ```
    
    ### Get an API Key
    Get a free API key at [unstract.com/llmwhisperer](https://unstract.com/llmwhisperer/).
    - **Free Tier:** 100 pages/day
    
    ## Usage
    
    ```bash
    llmwhisperer <file>
    ```
    
    ## Script Source
    
    The executable script is located at `scripts/llmwhisperer`.
    
    ```bash
    #!/bin/bash
    # Extract text using LLMWhisperer API
    
    if [ -z "$LLMWHISPERER_API_KEY" ]; then
      if [ -f ~/.clawdbot/.env ]; then
        # shellcheck disable=SC2046
        export $(grep -v '^#' ~/.clawdbot/.env | grep 'LLMWHISPERER_API_KEY' | xargs)
      fi
    fi
    
    if [ -z "$LLMWHISPERER_API_KEY" ]; then
      echo "Error: LLMWHISPERER_API_KEY not found in env or ~/.clawdbot/.env"
      exit 1
    fi
    
    FILE="$1"
    if [ -z "$FILE" ]; then
      echo "Usage: $0 <file>"
      exit 1
    fi
    
    curl -s -X POST "https://llmwhisperer-api.us-central.unstract.com/api/v2/whisper?mode=high_quality&output_mode=layout_preserving" \
      -H "Content-Type: application/octet-stream" \
      -H "unstract-key: $LLMWHISPERER_API_KEY" \
      --data-binary "@$FILE"
    ```
    
    ## Examples
    
    **Print text to terminal:**
    ```bash
    llmwhisperer flyer.jpg
    ```
    
    **Save output to a text file:**
    ```bash
    llmwhisperer invoice.pdf > invoice.txt
    ```
    
    **Process a handwritten note:**
    ```bash
    llmwhisperer notes.jpg
    ```