Bug fixes

• Fixed a bug where the response changed the case of the rubric title and caused an error.

Features

• Added 2025-02-24 model which adds support for sources and allows up 5 sources to be added to the evaluation of essays.
• Added responsive design accordion layout for the rubric when the available display container is narrow.

Bug fixes

• Fixed a bug where empty JSON parsing failed due to LaTeX formatting in AI responses.

Polish

• Change: Reduced the maximum size of the stimulus from 10,000 to 5,000 characters now that sources are available.

Features

• Added new models: standard-essay-ISCED1-2025-02-05, advanced-essay-ISCED1-2025-02-05, standard-essay-ISCED3-2025-02-05, and advanced-essay-ISCED3-2025-02-05. The ISCED1 variants are optimised for elementary school student levels and the ISCED3 variants for high school students.
• Added feedback in 10 languages other than English via Model options for 2025-02-05 and newer models.

Polish

• Improvement: improved the reliability of model responses with adversarial content.
• Improvement: improved error handling for docx_to_html endpoint. It now consistently returns 415 Unsupported Media Type errors when non-DOCX files are sent.

Features

• Added the isReadyForReview method, which determines whether a feedback session has been marked as ready for review.

Bug fixes

• Fixed a bug where a feedback session marked as insufficient_to_grade (i.e. not scorable) incorrectly returned a score when the method getFeedback was called.
• Fixed a bug in the holistic rubric feedback that would show "n" characters instead of newlines.

Polish

• Improvement: enhanced the "Submit to student" flow by disabling the save button with a clear explanation once feedback is marked as ready for review, instead of removing the button.

Features

• Added support for model options in the frontend code that was released in v0.17.0.

Bug fixes

• Fixed a bug where empty paragraphs were counted in the structurecheck. They are now correctly ignored.
• Fixed a bug where moderation was not running on older models.

Features

• Added support for model options to be passed to essay models to control behavior.
• Added support for structure check via the options check for a minimum number of paragraphs in an essay.
• Added support for moderation check via the options.
• Added support for topic check configuration via the options and the ability to tune the relatedness from Slightly related through to Completely related.
• Added models preview-standard-essay-ISCED3-2024-12-16 and preview-advanced-essay-ISCED3-2024-12-16 which work with ISCED level 3 (high school learners).

Polish

• Improvement: updated RESTful API documentation for the Essay Evaluate payload.

Bug fixes

• Internal logging improvements.

Bug fixes

• Added models preview-standard-essay-ISCED1-2024-12-16 and preview-advanced-essay-ISCED1-2024-12-16 which work with ISCED level 1 (elementary/primary learners).

Features

• Added a new method getEvaluateResponse, to retrieve server-generated feedback data.
• Introduced request rate-limiting support to enhance API performance and prevent misuse.

Bug fixes

• Fixed an issue where Docx-to-HTML conversion would fail if no style was applied to a paragraph.

Polish

• Improvement: internal tooling upgrades to improve the stability of the API.

Bug fixes

• Fixed a bug in the topic detection that was causing false positives.

Features

• Added preview models for structured essay formats: preview-standard-structured-2024-11-28 and preview-advanced-structured-2024-11-28
• Added a notification if an essay is not scorable including if it is too short, too long, or off topic.

Bug fixes

• Fixed a bug to validate the session_uuid at the request parameter parsing stage.

Polish

• Improvement: simplified oAuth authentication.
• Improvement: updated models for standard-essay and advanced-essay to point to 2024-11-22 which include topic detection and improved accuracy.
• Improvement: updated the documentation of the REST endpoints and RESTful mode documentation.

Polish

• Improvement: enhanced overall system stability.

Features

• Added a description for each rubric criterion, improving clarity and usability.
• Added a new endpoint to expose all internationalization labels used within the API for easier localization.

Bug fixes

• Fixed a bug where the API would use a different AI model when feedback was re-generated after an error during the previous feedback generation.

Polish

• Improvement: updated internal tooling to enhance the accuracy of our AI models.
• Improvement: sanitized rubric title and description fields to improve API security.
• Improvement: removed vague, low-use internalization labels to streamline localization.

Features

• [BETA] Added a new endpoint to convert .docx files into .html files.

Bug fixes

• Fixed a bug that caused the generateFeedback() progress bar to remain on the screen after grading complex essays.

Polish

• Improvement: improve the feedback accuracy for holistic rubric.
• Improvement: increase the maximum size of the comment box.
• Improvement: various internal tooling upgrades to improve the stability of the API.
• Improvement: various UI/UX clean up to remove unnecessary spacing.
• Improvement: show an error message when an unauthorized or invalid access_token is being used to generate feedback.

Features

• Added support for holistic rubric.

Bug fixes

• Fixed a bug that prevented feedback sessions with a short response from being initialized.

Polish

• Improvement: various internal tooling upgrades to improve the stability of the API.
• Improvement: update the moderation messages to be more informative.

Bug fixes

• Fixed a bug where -moderation models were not working.

Bug fixes

• Fixed a bug where feedback comments would not render after resuming an unsaved session in grading mode and calling generateFeedback.

Features

• Added models for standard-essay-2024-08-05 and advanced-essay-2024-08-05
• Added preview versions of standard-essay-moderation and advanced-essay-moderation, including AI moderation flagging for violence, sexual, and hate content in essays.
• Added a confirmation dialog if AI moderation flags a potential issue.
• Added a progress bar to show grading progress.
• Added input sanitization for untrusted content.
• Added a new setFeedback method.

Bug fixes

• Fixed a bug with rubric levels without titles.
• Fixed a bug with rendering zero-score rubrics.

Polish

• Improvement: deprecated the standard-essay-2024-06-30 and advanced-essay-2024-06-30 models, which have been superseded.
• Improvement: tightened rubric table styles to reduce space used by larger rubrics.

Features

• Added skipped score functionality to currently supported rubrics.

Polish

• Improvement: display unreviewed state instead of rejecting the promise when we create a new feedback session in the review state of a not_ready_for_review / non-existent session.
• Improvement: changed spacing around the stimulus and response areas.

Features

• Added support for rubrics with different score ranges.
• Added ready_for_review support for the feedback session.

Bug fixes

• Fixed a bug causing score and max_score to be intermittently calculated incorrectly during the generateFeedback process.

Features

• Added a new /essays/models endpoint to return a list of available models for essay evaluation.
• Added a new getter method session_uuid to the FeedbackSession instances.

Bug fixes

• Fixed a bug to increase the timeout limit to allow for longer model requests.
• Fixed a bug where an error in customer callbacks could stop all processing on the page.

Polish

• Improvement: increased the scalability and the reliability of the system.
• Improvement: updated the CSS build process for all styles.

Features

• Added credits and tag the models used for metrics.
• Added standard-essay-2024-08-05 and advanced-essay-2024-08-05 models

Bug fixes

• Fixed a bug where the initial auto generated comment was not being saved.
• Fixed a bug to make the feedbackSession request gracefully fail when it is not a 404.

Polish

• Improvement: updated error messages and handling to be consistent.
• Improvement: internal AI engine improvements.

Bug fixes

• Fixed a bug where we calculate the max score based on the highest level rather than the last index.

Polish

• Improvement: removed unused level property.

Bug fixes

• Fixed a bug and added a missing insufficient_to_grade option when resuming.
• Fixed a bug where the wrong validation was applied to a new feedback session.
• Fixed a bug where the comments were not showing when resuming in grade or student mode.

Polish

• Improvement: fix layout alignment with the rubric table.

Polish

• Improvement: updated styling for various elements.