v0.1.4-alpha

Features

• Added privacy policy.
• Translation: Added support for Opus-MT HF models.
• Translation: Notebook to demo translation use case via SDK.
• Translation: Display COMET metric on UI.
• Translation: Enabled LLM as judge (G-Eval).
• Integrated DeepEvals Ollama support into evaluator.

Bugfixes

• Fixed incorrect information for adding API keys in experiments.
• Fixed bug where logs were not updated after first check.
• Corrected the job type when submitting annotation jobs.
• Fixed issue with download URLs being incorrect/404.
• Fixed logger adjustment in testing.
• Fixed async issues after moving to S3FileSystem from boto3 client.
• Fixed delete_workflow.
• Fixed _handle_workflow_failure async issue.
• Raised API auth errors from inference jobs.

Refactoring

• Removed unnecessary attributes from requests to create new workflows.
• Removed cache container.
• Removed dependency on boto and only use S3 file system.
• Refactored workflow service and MLflow tracking client and compile metrics method.
• Refactored integration tests and added e2e.

Documentation

• Added architecture diagram with secrets.
• Updated docs to reflect translation use case.