v0.0.5

TL;DR

Performance Improvements

• Pages are now evaluated for visual content before being passed to the LLM.
• Non-visual pages are processed with traditional text parsing.
• Override with use_llm_for_all=True to use LLM for all pages.

Dependency Cleanup

• Replaced pdf2image with PyMuPDF for image conversion.
• Removes the need for poppler, simplifying installation.
• Dependencies are cleaned up and version-pinned to avoid conflicts (e.g., with MindsDB).

Enhanced Configurability

• Added support for additional OpenAI parameters (e.g., temperature).
• API key can now be set via the AIPDF_API_KEY environment variable.

Async OCR Support

• Introduced ocr_async() to make asynchronous OpenAI API calls.
• Complements the existing multi-threaded ocr() function.

Code Quality & Testing

• Refactored functions for better readability and maintainability.
• Added unit and integration tests to ensure reliability.

---

What's Changed

• Updated Content Extraction Flow for Better Performance by @MinuraPunchihewa in #8
• Added a Asynchronous Processing Option by @MinuraPunchihewa in #9
• Release 0.5.5 by @MinuraPunchihewa in #10

Full Changelog: v0.0.4...v0.0.5