Skip to main content

PDF to markdown using Azure OpenAI batch processing

Project description

Parallex

What it does

  • Converts PDF into images
  • Makes requests to Azure OpenAI to convert the images to markdown using Batch API
  • Polls for batch completion and then converts AI responses in structured output based on the page of the corresponding PDF
  • Post batch processing to do what you wish with the resulting markdown

Requirements

Parallex uses graphicsmagick for the conversion of PDF to images.

brew install graphicsmagick

Installation

pip install parallex

Example usage

import os
from parallex.models.parallex_callable_output import ParallexCallableOutput
from parallex.parallex import parallex

os.environ["AZURE_API_KEY"] = "key"
os.environ["AZURE_API_BASE"] = "your-endpoint.com"
os.environ["AZURE_API_VERSION"] = "deployment_version"
os.environ["AZURE_API_DEPLOYMENT"] = "deployment_name"

model = "gpt-4o"

async def some_operation(file_url: str) -> None:
  response_data: ParallexCallableOutput = await parallex(
    model=model,
    pdf_source_url=file_url,
    post_process_callable=example_post_process, # Optional
    concurrency=2, # Optional
    prompt_text="Turn images into markdown", # Optional
    log_level="ERROR" # Optional
  )
  pages = response_data.pages

def example_post_process(output: ParallexCallableOutput) -> None:
    file_name = output.file_name
    pages = output.pages
    for page in pages:
        markdown_for_page = page.output_content
        pdf_page_number = page.page_number
        

Responses have the following structure;

class ParallexCallableOutput(BaseModel):
    file_name: str = Field(description="Name of file that is processed")
    pdf_source_url: str = Field(description="Given URL of the source of output")
    trace_id: UUID = Field(description="Unique trace for each file")
    pages: list[PageResponse] = Field(description="List of PageResponse objects")

class PageResponse(BaseModel):
    output_content: str = Field(description="Markdown generated for the page")
    page_number: int = Field(description="Page number of the associated PDF")

Default prompt is

"""
    Convert the following PDF page to markdown.
    Return only the markdown with no explanation text.
    Leave out any page numbers and redundant headers or footers.
    Do not include any code blocks (e.g. "```markdown" or "```") in the response.
    If unable to parse, return an empty string.
"""

Project details


Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

parallex-0.2.0.tar.gz (9.5 kB view details)

Uploaded Source

Built Distribution

parallex-0.2.0-py3-none-any.whl (13.8 kB view details)

Uploaded Python 3

File details

Details for the file parallex-0.2.0.tar.gz.

File metadata

  • Download URL: parallex-0.2.0.tar.gz
  • Upload date:
  • Size: 9.5 kB
  • Tags: Source
  • Uploaded using Trusted Publishing? No
  • Uploaded via: poetry/1.8.4 CPython/3.12.6 Darwin/21.6.0

File hashes

Hashes for parallex-0.2.0.tar.gz
Algorithm Hash digest
SHA256 8d9096f58b94bd5eb2e730bd6a53332506ca3386389373f2880de4ff9515213e
MD5 33c2078ac0071f4ecbba1ae0e5a39c48
BLAKE2b-256 27052e9b9b79d2e8632b8e5be60b923bfdf36f8a716860f0380851827a545cdb

See more details on using hashes here.

File details

Details for the file parallex-0.2.0-py3-none-any.whl.

File metadata

  • Download URL: parallex-0.2.0-py3-none-any.whl
  • Upload date:
  • Size: 13.8 kB
  • Tags: Python 3
  • Uploaded using Trusted Publishing? No
  • Uploaded via: poetry/1.8.4 CPython/3.12.6 Darwin/21.6.0

File hashes

Hashes for parallex-0.2.0-py3-none-any.whl
Algorithm Hash digest
SHA256 e3561f19c16555d654c933a507719e330dc37841a5d7cd22d10c81bcb07a57ff
MD5 1ef68a711b15863b9692bac267575b8e
BLAKE2b-256 da67efd2c8778eeaddc884bcb24940cfc1e1e31ed0c2d172426de1949b626c55

See more details on using hashes here.

Supported by

AWS AWS Cloud computing and Security Sponsor Datadog Datadog Monitoring Fastly Fastly CDN Google Google Download Analytics Microsoft Microsoft PSF Sponsor Pingdom Pingdom Monitoring Sentry Sentry Error logging StatusPage StatusPage Status page