Chapter 9

Taxonomies

Define structured workflow labels, relationships, and conditional branching pathways that drive dynamic task routing across workflow nodes.

Overview

Taxonomy is used to define structured workflow labels, such as escalation-related options, that can be used to guide task routing or workflow decisions. Taxonomy labels cannot be used for data annotation; annotation labels are defined separately within the project template.

Taxonomy Components

  • Labels (Classes): The categorisation labels (e.g. "Vehicle", "Pedestrian", "Requires Escalation").
  • Relationships: Named semantic interactions between different labels (e.g. "has-part", "associated-with").

Guide: Creating a Taxonomy

Step-by-Step Guide

1

Navigate to Taxonomies

Click Taxonomies in the left sidebar and click the blue Create Taxonomy button.

Figure 46: Create Taxonomy Page

Figure 1: Create Taxonomy Page

2

Enter Name and Description

Provide a clear taxonomy title (e.g. Escalation Required) and optional usage notes.

3

Add Labels

Under the Labels section, type a label name and click Add Label.

4

Configure Multiple Labels

After creating one label, you can add more. Enter another label name and click Add Label again to add multiple labels. For each label, set: Name, Color (hex code), optional Alias, and Default (optional).

Figure 47: Taxonomy - Add Label

Figure 2: Taxonomy - Add Label

5

Set Default Label

The first added label is automatically marked as default. You can select another radio button to make it the default, or click the selected radio button again to deselect it (leaving no default).

6

(Optional) Add Relationships

Specify a Name, Color, and optional Description for each relationship. (Note: Relationships currently do not serve a specific functional routing purpose).

7

Click Create

Click Create to finalize and publish the taxonomy.

Managing Taxonomies

View, search, filter, and paginate through all company taxonomies from the Taxonomies dashboard.

Deleting Taxonomies

From the Taxonomies list, select the taxonomy you wish to delete and click the Delete icon, then confirm in the prompt.

Figure 48: Taxonomy - Delete Action

Figure 3: Taxonomy - Delete Action

Figure 49: Taxonomy - Delete Confirmation

Figure 4: Taxonomy - Delete Confirmation

Active Project Attachment
You cannot delete a taxonomy if it is currently attached to an active project. You must remove it from the project settings first.

Viewing Taxonomy Details

When a user clicks on any taxonomy in the list, an interactive Details Panel slides open automatically on the right side of the page, displaying:

  • Taxonomy Name: Displays the name of the selected taxonomy.
  • Labels: Displays all labels that have been added to the selected taxonomy
  • Relationships: Displays any relationships configured for the selected taxonomy
  • No Relationships Found: If no relationships have been added, an appropriate message is displayed
  • Automatic Panel Opening: The details panel opens automatically as soon as the user clicks on a taxonomy from the list.
  • Close Panel: The user can close the details panel using the Close (X) button.
Figure 50: Taxonomy - Details Panel

Figure 5: Taxonomy - Details Panel

Taxonomies in the Annotation Task

  • When an annotator opens a task, the taxonomy is shown in a collapsible panel above the file viewer.
  • If a Default label is set, it is pre-selected when the task opens. Annotators can change the selection while annotating a task.
  • The taxonomy section is independent from task templates.

Workflow Routing via Taxonomies

Taxonomy values can control a task's journey through the workflow. When a task is submitted, downstream Plugin nodes receive the taxonomy selection, enabling conditional routing.

Example: A taxonomy named “Escalation Required” contains the labels “Yes” and “No.” During the annotation process, the annotator selects one of these labels based on whether the task requires escalation. Once the task moves to the downstream Plugin node, the external agent connected to that node can access the taxonomy selection made by the annotator. Based on the selected value, the external agent can route the task through the appropriate output path of the Plugin node — for example, selecting “Yes” routes the task to the escalation pathway, while selecting “No” routes it to the standard completion pathway.

Note
Currently, taxonomy labels are used only for workflow-based task routing and cannot be used for annotating data. Annotation labels are defined separately within the project template. Taxonomy selections made by users can be accessed by an external agent through a downstream Plugin node, which can then use the selected value to route the task to the appropriate workflow path.

Release notes - 2026

What’s New in 2026 — new capabilities, refinements, and fixes, in the order they landed.

v0.9 October 6, 2026

Dataset File View

Location: 5. Dataset Setup and Management

We’ve introduced a new Dataset File View in TensorAct, making it easier to browse, preview, and manage files within a dataset. Users can now switch between the existing List view and the new View tab, which displays files as thumbnails with file previews, file names, and quick actions. The new File Viewer allows users to preview files and navigate between them using Previous and Next controls, while supporting file-specific viewing for Images, PDF, PDFOCR, TEXT, CSV, Audio, and Video files. Pagination and items-per-page controls make it easier to navigate large datasets, while file selection allows users to select and delete multiple files directly from the View tab. Additional controls such as file-name copying, media playback, progress tracking, and audio controls provide a more convenient and efficient dataset file management experience.

Change Task Stage from the Project Task List

Location: 8. Projects and Task Management

Project managers and administrators can now quickly change the stage of multiple tasks directly from the Project Task List. By selecting one or more tasks, users can transition tasks to different stages without needing to open each task individually. This enhancement significantly speeds up project management workflows and bulk task tracking.

Azure Blob Storage Integration

Location: 7. Integrations

TensorAct now supports direct integration with Azure Blob Storage. Users can securely connect their Azure storage accounts using either a Shared Access Signature (SAS) Token or a Service Principal. Once connected, users can browse their Azure storage containers directly from the TensorAct interface, select specific folders, and seamlessly import large batches of files into their datasets.

Reviewer Feedback and Pinned Feedback Management

Location: 8. Projects and Task Management

We’ve enhanced the feedback loop between annotators and reviewers. Reviewers can now add detailed feedback and pin specific issues directly to task items. The new issue management panel allows annotators to see all pinned feedback at a glance, reply to comments, and resolve issues systematically, ensuring higher data quality and clearer communication during the review process.

Email Notification Triggers

Location: 17. Email Templates

Administrators can now configure automated email notification templates and triggers. Located in the Settings menu, this feature allows you to customize the content of emails sent to users based on specific system events, keeping project teams instantly informed about task assignments, stage changes, and critical updates.

Uploading and Importing Files

Location: 6. File Management

We have introduced clearer guidelines and limits for file uploads to ensure optimal performance. Users uploading files from their local device can select and upload up to 100 files at a time. For larger batch imports, users leveraging our cloud storage integrations (AWS S3, Google Cloud Storage, or the new Azure Blob Storage) can import up to 25,000 files in a single action.

v0.8 September 8, 2026

Release Summary

Change Type Sections
Added
Updated
Removed No sections removed in this release

Sign-Up & Onboarding

Location: 1. Sign-Up & Onboarding

We’ve introduced a comprehensive Sign-Up and Onboarding system in TensorAct, establishing standardized self-registration and administrator-guided onboarding workflows. Users with verified business email domains can now self-register directly into their company’s workspace, receiving an automated temporary password with mandatory password change enforcement upon first sign-in. For users from unregistered companies, the system detects the domain and provides a streamlined modal to submit a new company registration request to our team. Additionally, the release introduces administrator-initiated user creation with welcome notifications, secure password rotation for pending onboarding invites, and strict email validation rules that reject personal email providers to safeguard platform security.

Deleting Users

Location: 4. User Management

Super Users and Company Admins can now permanently delete user accounts directly from the Users list. The feature supports bulk operations via multi-select checkboxes and incorporates a safety confirmation modal requiring administrators to type DELETE in uppercase to prevent inadvertent deletions. Deletion immediately invalidates active user sessions, presenting an access revocation message upon sign-in. To maintain workflow integrity, in-progress tasks are handled automatically: unsubmitted annotation and review tasks claimed by deleted users are released back to the pool for other team members to claim, while all previously submitted annotations and review decisions remain fully preserved.

Google Cloud Storage (GCS) Integration

Location: 7. Integrations

TensorAct has expanded its cloud storage integrations to include native support for Google Cloud Storage (GCS) alongside AWS S3. Accessible under the Integrations menu, administrators can connect GCS buckets using a secure Google Cloud Service Account permission model without handling static access keys. The platform provides pre-generated CORS setup commands for Google Cloud Shell and includes a built-in integration testing tool to verify bucket connectivity against object URLs. Once configured, users can browse GCS bucket directories, search files, and import large datasets directly into TensorAct from the cloud file browser.

Email Templates

Location: 17. Email Templates

Super Users can now customize and manage automated platform notification emails via the Email Templates interface located in Settings. The platform maps distinct email templates to critical system events, including Password Reset, MFA/OTP verification, new account Welcome messages, company access invitations, project team assignments, and account activation or deactivation alerts. This centralized notification management ensures project teams and users receive clear, automated, and timely communication across the entire user and project lifecycle.

Exporting Results

Location: 8. Projects and Task Management

Users can now export task annotation results directly from the Project Tasks dashboard. The export process packages output data into a structured ZIP archive named after the project (<Project Name>.zip), containing individual JSON files for each task (<Task Name>_<Project Name>.json) to guarantee modular data management. The export scope adapts dynamically based on user selection: teams can export specific manually checked tasks, all displayed tasks using the master checkbox, or filter exports by workflow stage (such as exporting only completed tasks under the Complete node filter).

Code copied to clipboard!