DocWire SDK Files

This release introduces a significant new capability by adding support for OpenAI text embeddings. It also features a major refactoring that modularizes the HTTP client into its own library, and enhances the robustness of AI components by switching from file extension checks to MIME type-based content identification.

A whisper of meaning, in numbers cast,
New dimensions open, built to last.
For search and reason, a powerful key,
Unlocking knowledge for all to see.
✨🔢🧠🔍

• Features
• OpenAI Embeddings Support: Added a new openai::embed chain element to generate text embeddings using OpenAI's models (text-embedding-3-small, text-embedding-3-large, text-embedding-ada-002). This includes corresponding CLI options (--openai-embed, --openai-embed-model) and extensive documentation with examples.

• MIME Type to File Extension Conversion: Introduced a new utility function, content_type::by_file_extension::to_extension, to convert MIME types back to their common file extensions, improving interoperability.

• Improvements

• Robust Content Handling in AI Components: AI components like openai::Chat and local_ai::model_chain_element now rely on MIME types (text/plain, image/*) instead of file extensions for more accurate content identification.

• Enhanced http::Post Robustness: The http::Post component can now automatically determine the file extension from the MIME type if it's not explicitly provided in the data_source.

• Refactor

• Modular HTTP Client: The http::Post class has been moved from docwire_core to a new, dedicated docwire_http library to improve modularity.

• Fixes

• Build (Windows): Fixed a build failure on Windows by explicitly adding libxml2 as a direct dependency in the vcpkg port.
• Documentation: Corrected a minor paragraph formatting issue in README.md.