llmwhisperer

What This Skill Does

The llmwhisperer skill is a specialized OCR (Optical Character Recognition) and layout extraction tool designed to interface with the LLMWhisperer API by Unstract. Unlike standard text extractors, this skill is purpose-built to handle complex document structures, tables, and handwritten notes. It excels at preserving the semantic layout of a document, making it an essential utility for digitizing physical paperwork, converting legacy forms into machine-readable data, and interpreting challenging visual media.

Installation

To begin using llmwhisperer, first ensure you have the OpenClaw environment configured. Install the skill by running: clawhub install openclaw/skills/skills/gumadeiras/llmwhisperer. Once installed, you must register for a free API key at the Unstract website. Save this key into your configuration file located at ~/.clawdbot/.env by appending LLMWHISPERER_API_KEY=your_key_here. The skill is designed to automatically detect this key from your environment, ensuring a seamless authentication process every time you invoke the tool.

Use Cases

This skill is highly effective for professionals dealing with high-density information. Common use cases include: digitizing handwritten medical or research notes, extracting financial data from non-searchable PDF invoices, reading text from complex marketing flyers with overlapping elements, and processing scanned government or legal forms where standard OCR tools fail to maintain column integrity.

Example Prompts

1. "OpenClaw, please run the llmwhisperer skill on the scanned_invoice.pdf file in my downloads folder and output the text directly to the console so I can review it."
2. "Process the image of my handwritten meeting notes using llmwhisperer and save the resulting text content into a new file called meeting_summary.txt."
3. "Use the llmwhisperer skill to extract all data from this image of a complex table; I need the layout preserved so I can parse it into a spreadsheet later."

Tips & Limitations

To maximize the performance of this skill, ensure that your input files (JPG or PDF) are clear and high-resolution. While LLMWhisperer is exceptionally robust, extremely low-light or blurry photos may impact accuracy. Note that the API has a usage limit of 100 pages per day on the free tier. When using the command, remember that the tool performs a direct binary upload; if your local environment has restricted internet access, the tool will fail as it requires a connection to the Unstract API endpoints. Always verify your .env file permissions if you encounter authentication errors.