Release Notes

📢 Deprecations:

May 29, 2026

• Context Studio Overview Documentation: Added an overview page for Context Studio in the Docs & Guides section. The guide explains how Context Studio translates advertiser briefs into contextual ad packages, including content hierarchy, Dynamic Tags, aggregation methods, inclusion/exclusion tags, and match percentage thresholds.
• Context Studio Ad Segments API: Added new endpoints under Context Studio for managing ad segments. The POST endpoint creates ad segments for videos by specifying start/end times and frame numbers, while the GET endpoint retrieves all ad segments stored for a video. These endpoints support scene-level targeting.
• Dynamic Tags - List Groups Endpoint: Added a paginated endpoint (GET /api/v3/dynamic-tags/groups) to retrieve dynamic tag groups. Supports filtering by name, version name, status, visual encoder, transcript encoder, training dataset, creator, and creation date range. Results can be sorted and paginated with configurable page size.

May 14, 2026

• Add via S3: Added a new “Add via S3” option to the dataset detail page dropdown menu. Users can now add images or videos to existing datasets by providing an S3 path prefix, with options to select a Connection and set a Max depth for subfolder recursion.

April 21, 2026

• Dynamic Tags for Nova Datasets: Extended Dynamic Tags to support Nova datasets. Tag groups can now be created on Nova-ingested video content with segment-level previews. Results display 30-second video segments with relevance scores, allowing tags to be refined before publishing.

March 6, 2026

• Context Studio: Build targeted video packages using dynamic tags and metadata filters. Features include LLM-powered tag suggestions, metadata filtering (genre, rating, series, title), and package building with up to 10 inclusion/exclusion tags. Packages can be exported as summary PDFs or sent directly to ad servers like Freewheel and GAM. A timeline view is also available for shot-level analysis.
• Video Level Aggregation Improvements: Added video-level tag aggregation to the Dynamic Tags experience. Users can now sort results by highest or lowest scoring videos, view representative thumbnails in result lists, navigate qualifying shots in video previews, and see total shot counts. Dynamic Tag Groups now support filtering across all columns.
• NOVA Search: Added support for AWS Nova-based asset ingestion and semantic search. Videos are processed in 30-second chunks for optimal search results.
• Agentic Search: Added a natural language interface for searching and refining results through conversation. Users can copy/paste queries, drill down by video ID, and upload images to find similar content. Results return in under 30 seconds.
• Concepts Support on V1 Ingestion Pipeline: Extended Concepts to support TX-Large 4 (Perception Encoder) via the V1 Ingestion Pipeline. Works with all V1 Pipeline datasets, including existing ones, with no reingestion required. Backwards compatible with TX-Large2 and TX-Large3 encoders. Note: Concept names must be unique within a dataset.

February 11, 2026

• Deprecated Search Endpoints Removed: The following legacy search endpoints have been removed:
  • GET /api/v1/search — replaced by POST /api/v1/search/text-to-image
  • GET /api/v1/search/similar-images — replaced by POST /api/v1/search/image-to-image

August 12, 2025

• The Embedding dropdown on the Queries page will be removed: This information has now been moved to the Dataset UI page and can also be retrieved via API.

June 16, 2025

• New Dynamic Tags Experience (Beta): We have a new dynamic tags experience within the Coactive UI that we are excited to share with our users. If you would like to try this out, reach out to Coactive support.
• Old Dynamic Tags Tables will no longer be accessible (Deprecation Notice): We are deprecating the old Dynamic Tags tables category_*. You will no longer be able to access these tables after June 20th, please recreate your tags on the New Dynamic Tags experience.
• Ingestion Observability (Beta): Track the status of ingested assets directly in the dataset view or via API, including error reporting, processing, and ready to search states. For full details, see the Ingestion Observability Guide.
• SQL UI Row Limit: The SQL tab in the platform now displays a maximum of 10,000 rows in the UI. If your result set exceeds this limit, you can still:
  • Use SELECT COUNT(*) to retrieve the full row count
  • Download the full result set as CSV or Parquet via S3 Export (New)
  Note: Users will see the total number of results, but will only be able to browse the first 10,000 rows within the UI.

June 2, 2025

• Connections API (Beta): Introduced endpoints to securely connect external cloud storage to Coactive for automated ingestion workflows. The API currently supports S3 buckets and allows programmatic creation, testing, updating, and deletion of connections. Future support for additional cloud storage providers is planned.

  For more details, see the Connections API reference and the Cloud Storage Access guide.

• Bucket-Based Ingestion: Added a new endpoint (POST /api/v1/ingestion/buckets) that scans a specified path within a connected S3 bucket and ingests all discovered assets. Ideal for bulk ingestion workflows that do not require individually listing files.

May 23, 2025

• System Credentials API (Beta): Added support for long-lived, programmatic credentials not tied to specific users. These credentials can be scoped to datasets or system-wide roles and are ideal for backend integrations.

  See the System Credentials reference for more information.

• User Management API (Beta): Introduced endpoints to invite, list, update, and remove users, as well as manage assignable roles. Role scoping supports system- and dataset-level access control.

  Learn more in the User Management reference and User Roles & RBAC guide.

March 13, 2025

• Ingestion API (v1): Released a modern ingestion pipeline that supports asset uploads via public URLs, S3 buckets, and batch asset lists. A separate endpoint also allows metadata to be applied post-upload. These new APIs replace earlier dataset-scoped routes and improve ingestion scalability, performance, and compatibility with newer ML workflows.

• As part of this release, we’ve begun deprecating the legacy multimodal encoder previously used in dataset creation and search. Customers should migrate to multimodal-tx-large2 or multimodal-tx-large3, which offer improved performance and semantic accuracy.

  See the Ingestion API reference and Update Metadata documentation.

  For compatibility information, see the Accepted Media Formats guide.

Febuary 20, 2025

• Coactive is transitioning to a unified and versioned API domain structure:
  • New API domain: https://api.coactive.ai
  • Legacy API domain: https://app.coactive.ai
• All new APIs — including ingestion, user management, connections, and system credentials — are hosted on the new domain and should be used for all future integrations. The legacy domain for existing APIs will remain available until December 31, 2025, but it will not receive further updates or new features.
  • Action recommended:
    • When reading our API docs, please verify the domain used.
    • Existing Legacy APIs will gradually be placed by newer APIs, please check this page frequently for our newly published APIs