Data Science AI Quick Actions v 1.0.5 Fast Follow Release

Data Science AI Quick Actions v 1.0.5 now includes:

  • Expanded multiple inference endpoint support
    • You can now use the chat completion endpoint in the inference playground.
    • You can select Copy payload in the inference playground to get the model endpoint and get code sample to invoke the model with the CLI that corresponds to the completion or chat completion inference format.
  • Exclusion of files during model registration
    • You have the option to exclude files based on file format during the model registration process. This is helpful when you want to bring in a model from Hugging Face to AI Quick Actions and want to exclude certain files from being downloaded such as .txt and .pdf files.
  • Addition of copy model ocid feature in model details page
    • After registering the model, you can find and copy the model ocid in the model details page without navigating to the Console to find the model ocid.
  • Model details actions button UI changes
  • Tool calling support
    • AI Quick Actions now has a version of vllm container that supports tool calling capabilities. Tool calling lets a model invoke external functions during inference, enhancing its ability to provide context-specific answers.

For more information, see the Data Science documentation on AI Quick Actions.