This release adds access to Azure OpenAI models and adds missing caches for documentation notebooks.
- API: Added :class:`.AzureOpenAIChat` to grant users access to models deployed on Azure OpenAI.
- API: Update :class:`.TitanBedrockChat().get_response` to accept a plain message string instead of a JSON string.
- DOC: Added missing caches in the Connecting to GenAI Models documentation notebook.
- DOC: Fixed some outdated and unclear instructions in the documentation.
This is a bugfix release containing updates for a failing unit test and enhancements to historical release notes.
- TEST: Fix test_huggingface_retry where the session was not mocked correctly
- DOC: Retroactively updated release notes for more consistent quality and detail e.g. hyperlinking class definitions
This release adds access to Google's Vertex AI model and fixes the links in the documentation.
- API: Added :class:`.VertexAIChat` to grant users access to Gemini models deployed on Google Vertex AI.
- DOC: Fix broken links on the sphinx homepage
This release expands ARTKIT's connectivity to include any GenAI application with an exposed HTTP endpoint enabling evaluations of virtually any custom target system.
- API: Added the :class:`.HTTPXChatConnector` class which requires further customization by a user for implementation but considerably expands ARTKIT's connectivity with custom target systems.
- DOC: Added "Calling custom endpoints via HTTP" section in :doc:`user_guide/advanced_tutorials/creating_new_model_classes` to guide implementation of the HTTPX connector
This release adds access to the Titan diffusion model in AWS bedrock and improves the documentation.
- BUILD: Add prefix to veracode scan number
- DOC: Add clarifying documentation including links, badges, and extra details on dev dependencies
- API: Add :class:`.TitanBedrockChat` to enable users to access Titan diffusion models.
This release provides direct ARTKIT access to AWS Bedrock models.
- API: Added :class:`.TitanBedrockDiffusionModel` to enable users to access models deployed on AWS Bedrock.
This release exposes key base classes so that users can access them for typing purposes.
- API: Key base classes are now exposed through the :mod:`artkit.api` module: :class:`.ChatModel`, :class:`.CompletionModel`, :class:`.DiffusionModel`, and :class:`.VisionModel`. These classes are frequently used in type hints, and this change makes it easier to import them without having to know the exact module they are defined in.
This is a maintenance release that introduces backward compatibility with Python 3.10 along with a few minor API improvements.
- BUILD: artkit can now be used with Python versions back to 3.10, allowing use with popular services such as Google Colab.
- API: Method :meth:`~.CachedGenAIModel.clear_cache` can now clear cache entries
after a given create or access time using the new arguments
created_afterandaccessed_after. - DOC: Minor documentation cleanups.
- FIX: :class:`.CachedDiffusionModel` and :class:`.CachedVisionModel` are now also available through the :mod:`artkit.api` module. Bot classes had been defined in the :mod:`artkit.diffusion.base` and :mod:`artkit.vision.base` modules, respectively, even though they are not abstract base classes. The fix moves both classes one level up to the :mod:`artkit.diffusion` and :mod:`artkit.vision` modules, which also exposes then through the :mod:`artkit.api` module.
Initial release of artkit.