> For the complete documentation index, see [llms.txt](https://docs.unitlab.ai/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.unitlab.ai/release-notes/2025-releases.md).

# 2025 RELEASES

{% updates format="full" %}
{% update date="2025-11-28" %}

## v3.6.0 — NIfTI, medical releases, and reliable augmentation

**Features**

* Added NIfTI support, medical thumbnails and releases, medical usage quotas, frame metadata, and medical result handling.
* Added resumable and idempotent dataset augmentation with RabbitMQ-backed workers and safer task recovery.
* Expanded HubSpot lead and meeting integration.

**Fixes**

* Corrected rotation, crop, flip, and resize transforms across boxes, polygons, lines, points, skeletons, and segmentation masks.
* Reduced worker crashes and memory pressure, and corrected augmented release thumbnails, split progress, and image-extension handling.
  {% endupdate %}

{% update date="2025-10-20" %}

## v3.5.0 — Dataset augmentation, Gemini, and dynamic uploads

**Features**

* Added image dataset augmentation, Gemini model support, dynamic SDK upload validation, dataset class/status/history fields, and augmentation version progress.

**Fixes**

* Improved N+1 queries, subscription quota checks, split-ratio handling, thumbnails, task progress, invalid text encoding, and SDK context handling.
  {% endupdate %}

{% update date="2025-09-26" %}

## v3.4.0 — Audio models, dynamic assignment, and dataset splits

**Features**

* Added audio AI-model integration, dynamic reviewer assignment, batch upload counts, data tags, dataset splits, SDK split support, and tunnel-managed model endpoints.
* Expanded review of rejected data and dataset filtering by generic type and workflow status.

**Fixes**

* Corrected review and annotation status transitions, dataset clone inclusion, empty-result checks, text filtering, action validation, and history-preview queries.
  {% endupdate %}

{% update date="2025-08-25" %}

## v3.3.0 — Medical annotation and workspace-wide automation

**Features**

* Added DICOM medical annotation, multi-frame extraction, medical SAM encoding, medical upload limits, and dynamic medical scaling.
* Redesigned automation flows for workspace use, added reviewer flow updates, audio AI/STT integration, and generic-type filtering.
* Added HubSpot CRM integration and improved model suitability by annotation type.

**Fixes**

* Corrected audio review, text assignment, attribute errors, duplicate querysets, timeouts, and Sentry diagnostics.
  {% endupdate %}

{% update date="2025-07-21" %}

## v3.2.0 — Text annotation and collaborative notifications

**Features**

* Added the start-annotating flow for text data and allowed text releases without class requirements.
* Added comment notifications for project creators, labelers, and reviewers, with source-file context.

**Fixes**

* Improved dynamic AI-model upload logging and removed obsolete dataset-creation checks.
  {% endupdate %}

{% update date="2025-04-02" %}

## v3.1.0 — Multi-AI flows and class lifecycle management

**Features**

* Added multi-model AI flows, flow assignment to existing projects, dynamic model parameters, dataset tags, batch auto-labeling, and richer SDK/CLI behavior.
* Added class deletion/restoration, history-result updates, centralized result merging, and model-class lifecycle controls.

**Fixes**

* Corrected release creation, dataset cloning, class IDs, project deletion, result merging, date filters, and N+1 queries.
  {% endupdate %}

{% update date="2025-01-27" %}

## v3.0.0 — Tags, reassignment, and safer class management

**Features**

* Added class edit/delete/bulk-restore, datasource tags and tag filters, selected-item reassignment, pending-upload counts, and class occurrence counts.
* Simplified project creation and dataset cloning while preserving source tags and metadata.

**Fixes**

* Corrected class defaults and clone behavior, member project visibility, authentication paths, deleted-class rendering, and Auto Crop error reporting.
  {% endupdate %}
  {% endupdates %}