Enhance README :: Add Cost-Effective PDF to Markdown Conversion Using AWS Bedrock Models

[igorlima]

This PR introduces another model option for converting PDFs to Markdown using vision models via the zerox library.

• here's a code snippet for you to give it a go.
  • snippet

    # PDF EXTRACT via zerox
    from pyzerox import zerox
    import pathlib, os, asyncio
    
    import litellm
    # # https://github.com/BerriAI/litellm/blob/11932d0576a073d83f38a418cbdf6b2d8d4ff46f/litellm/litellm_core_utils/get_llm_provider_logic.py#L322
    litellm.suppress_debug_info = True
    # https://docs.litellm.ai/docs/debugging/local_debugging#set-verbose
    litellm.set_verbose=False
    # https://docs.python.org/3/library/logging.html#logging-levels
    # https://github.com/BerriAI/litellm/blob/ea8f0913c2aac4a7b4ec53585cfe9ea04a3282de/litellm/_logging.py#L11
    os.environ['LITELLM_LOG'] = 'CRITICAL'
    
    kwargs = {}
    custom_system_prompt = None
    
    # AWS Bedrock requires the library `boto3`
    # !pip3 install boto3
    model = "bedrock/amazon.nova-lite-v1:0"
    
    # Define main async entrypoint
    async def main():
      file_path = "data/input.pdf" ## local filepath and file URL supported
      ## process only some pages or all
      select_pages = None ## None for all, but could be int or list(int) page numbers (1 indexed)
      output_dir = "./data" ## directory to save the consolidated markdown file
      output_dir = None
      result = await zerox(file_path=file_path, model=model, output_dir=output_dir,
                          custom_system_prompt=custom_system_prompt, select_pages=select_pages, **kwargs)
      return result
    
    # run the main function:
    result = asyncio.run(main())
    md_text = "\n".join([page.content for page in result.pages])
    
    # print markdown md_text
    print(md_text)
    # save markdown to file
    pathlib.Path("data/output-zerox-pdf.md").write_text(md_text)
    print("Markdown saved to output-zerox-pdf.md")
    
    """
    HOW TO RUN THIS SCRIPT:
    AWS Bedrock requires the library `boto3`: `pip3 install boto3`
    
    ({
    export AWS_ACCESS_KEY_ID=XXXXXXXXXXXXXXXXXXXX
    export AWS_SECRET_ACCESS_KEY=xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx
    export AWS_REGION_NAME=us-east-1
    export LITELLM_LOG=CRITICAL
    python3 snippet.py
    })
    """

    # HOW TO RUN THIS SCRIPT:
    # AWS Bedrock requires the library `boto3`: `pip3 install boto3`
    export AWS_ACCESS_KEY_ID=XXXXXXXXXXXXXXXXXXXX
    export AWS_SECRET_ACCESS_KEY=xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx
    export AWS_REGION_NAME=us-east-1
    export LITELLM_LOG=CRITICAL
    python3 snippet.py

What makes this so cool? The zerox library leverages the litellm library behind the scenes, giving us a wider range of options. Plus, litellm lets us tap into vision models from the AWS Bedrock platform and a few other choices.

I opened this PR for two main reasons:

1. To offer a cost-effective solution.
2. To provide a model with a higher RPM (requests per minute).

If you're working with a PDF over 10 pages, you might hit Google's rate limit for the Gemini model under the free tier.

Here's a quick comparison of some models:

Model Name	Input Cost / Token	Output Cost / Token
"amazon.nova-lite-v1:0"	0.000000060	0.00000024
"gemini/gemini-1.5-flash"	0.000000075	0.00000030
"gemini/gemini-1.5-flash-latest"	0.000000075	0.00000030
"gpt-4o-mini"	0.000000150	0.00000060
"azure/gpt-4o-mini"	0.000000165	0.00000066
"groq/llama-3.2-11b-vision-preview"	0.000000180	0.00000018
"anthropic.claude-3-haiku-20240307-v1:0"	0.000000250	0.00000125
"amazon.nova-pro-v1:0"	0.000000800	0.00000320
"claude-3-opus-20240229"	0.000015000	0.00007500

The Amazon model is super competitive in pricing, making it a breeze to use with AWS Lambda functions.

By adding more options right in the README, we're making it easier for newcomers to find what they need.

Thanks!