arc53 / doc2md

Convert pdf and image files into markdown

This project helps users to convert Documents (.pdf, .png, .jpg, .jpeg) into Markdown for you ease of ingestion into LLM workflows.

It uses a public LLM endpint (doc2md) here This endpoint simply gives images or pdfs (converted to images) to visual model and asks it to conver it into markdown.

Here is a quick snippet using python to perform such task:

# Client is your OpenAI compatible client
model = 'meta-llama/Llama-3.2-11B-Vision-Instruct'
prompt = "Convert the following image to just the markdown text, respond only with text and description of it if relevant."
messages = [
    {
        "role": "user",
        "content": [
                    {
            "type": "text",
            "text": prompt,
            },
            {
            "type": "image_url",
            "image_url": {
                "url":  f"{base64_image}"
            },
            },
        ]
    }
]
response = client.chat.completions.create(model=model,
    messages=messages,
    stream=False,
    max_tokens=int(max_new_tokens),
    **kwargs)