It is a demonstration application that uses ScreenshotOne API to render a full-page screenshot of website pages, apply OCR to it, and search for given patterns.

It uses the following technologies:

• ScreenshotOne API—the best screenshot API;
• OpenAI Vision API (for OCR).

Check out more examples in the ScreenshotOne examples repository.

You provide an URL via the CLI argument and the application will:

1. Take a full-page screenshot of the given URL;
2. Split the screenshot into multiple parts;
3. Apply OCR to each part;
4. Search for the given patterns in the OCRed parts by asking AI to answer that if the pattern is present in the part;
5. Then it will get the HTML content of the page;
6. Parse links and navigate to the internal links.
7. And repeat the process from step 1 for the new page till it finds the match of the given patterns on the OCRed part of the page.
8. Then it will print the results.

The code was written with the help of Cursor as specified in the instructions.

1. Clone the repository:

git clone https://github.com/screenshotone/examples.git

2. Go to the examples/python/vision-researcher directory:

cd examples/python/vision-researcher

3. Install the dependencies:

pip install -r requirements.txt

4. Create a .env file and set the following environment variables:

SCREENSHOTONE_API_KEY=your_screenshotone_api_key
OPENAI_API_KEY=your_openai_api_key

5. Run the application:

python vision_researcher.py <url> <prompt> <max pages>

For example, to search for the content containing "testimonials" on the ScreenshotOne website:

python vision_researcher.py https://screenshotone.com "Does the website page or the website page parts contain testimonials?" 5

The results will be printed in the console: