The following are the release notes for TetraScience AI Services versions 1.2.x. TetraScience AI Services empower developers and scientists to transform data into decisions through seamless, secure APIs, and enable intelligent automation across scientific, lab, and enterprise workflows.

📘

NOTE

TetraScience AI Services are currently part of a limited availability release and must be activated in coordination with TetraScience along with related consuming applications. For more information, contact your customer account leader.

v1.2.1

Release date: TBD

TetraScience has released TetraScience AI Services version 1.2.1. This release enhances vector store lifecycle management with update and deletion improvements, introduces vector store observability, and includes bug fixes for vector store creation and querying.

Here are the details for what's new in TetraScience AI Services v1.2.1.

Prerequisites

TetraScience AI Services v1.2.1 requires the following:

• Tetra Data Platform (TDP) v4.4.1 or later

Enhancements

Enhancements are modifications to existing functionality that improve performance or usability, but don't alter the function or intended use of the system.

Vector Store Update and Index Synchronization

The vector store PUT endpoint now triggers the vectorization notebook, allowing users to update a vector store's content by re-uploading knowledge base files and triggering re-vectorization. After chunking is complete, the system automatically issues a synchronization operation to update the vector search index, ensuring query results reflect the latest content.

Vector Store Deletion Cleanup

Deleting a vector store now properly removes the associated Databricks vector search resources in addition to the internal database record. Previously, deletion only removed the internal record, leaving orphaned resources in Databricks.

Knowledge Base File Type Support in AI Asset Files API

The AI Asset Files API now supports a type parameter to distinguish between AI Workflow assets and knowledge base files. When uploading knowledge base files, users specify type=knowledge-base and provide a knowledgeBase parameter instead of the namespace and aiWorkflow parameters used for workflow assets. This enhancement also adds path traversal protection and blocks dangerous file extensions for multipart uploads.

Vector Store Observability

Vector store operations now include key metrics and monitoring dashboards in Amazon CloudWatch. This provides operators with visibility into vector store performance, enabling troubleshooting, root cause analysis, and tuning.

Bug Fixes

The following bugs are now fixed.

• Vector store creation no longer fails with a 403 PERMISSION_DENIED error when the vector search endpoint requires explicit permissions for the service principal running the creation job.
• Querying a vector store immediately after it reaches ready status no longer returns a transient 502 Databricks Model Serving error.

Deprecated Features

There are no deprecated features in this release. For more information about TDP deprecations, see Tetra Product Deprecation Notices.

Known and Possible Issues

📘

NOTE

No new known issues were introduced in this release. The known issues listed below were also present in previous versions.

The following are known limitations of TetraScience AI Services v1.2.1:

• Task Script README Parsing: The platform uses task script README files to determine input configurations. Parsing might be inconsistent due to varying README file formats.
• AI Agent Accuracy: AI-generated information cannot be guaranteed to be accurate. Agents may hallucinate or provide incorrect information.
• Databricks Workspace Mapping: Initially, one TDP organization maps to one Databricks workspace only.
• Analyst Policy Install Workflow UI Access: Users assigned a role with an Analyst policy can view, but not use, the Install Workflow button even though the policy doesn't allow them to install an AI workflow or run an inference. A fix for this issue is scheduled for a future AI Services UI release.

Upgrade Considerations

To upgrade to the latest AI Services version, see Update the AI Services UI version in the TetraScience AI Services User Guide (v1.2.x).

During the upgrade, there might be a brief downtime when users won't be able to access AI Services functionality.

After the upgrade, verify the following:

• Existing AI Workflow installations continue to function correctly
• Inference requests complete successfully
• Knowledge base and vector store operations function as expected
• Vector store update and deletion operations complete without errors
• The AI Services UI displays correctly

📘

Security

TetraScience continually monitors and tests the codebase to identify potential security issues. Security updates are applied on an ongoing basis.

📘

Quality Management

TetraScience is committed to creating quality software. Software is developed and tested following the ISO 9001-certified TetraScience Quality Management System.

For instructions on how to validate core AI Services functionality programmatically, see Test TetraScience AI Services Programmatically in the TetraConnect Hub. For access, see Access the TetraConnect Hub.

For more information, see the TetraScience AI Services documentation.

v1.2.0

Release date: 30 April 2026

TetraScience has released its first general availability version of TetraScience AI Services version 1.2.0. This release introduces an Invoke a Training Notebook API, Knowledge Base and Vector Store support, a centralized model and artifact repository, cross-environment model promotion, model aliases for lifecycle management, enhanced LLM endpoint controls, and improved System Log coverage.

Here are the details for what's new in TetraScience AI Services v1.2.0.

Prerequisites

TetraScience AI Services v1.2.0 requires the following:

• Tetra Data Platform (TDP) v4.4.1 or later

New Functionality

New functionality includes features not previously available in TetraScience AI Services.

Invoke a Training Notebook API

TetraScience AI Services now provides an Invoke a Training Notebook API for model training that allows users to invoke any training notebook within their Scientific AI Workflow directly through the AI Services API. The new POST /v1/inference/invoke/* endpoint accepts arbitrary JSON payloads and Amazon Simple Storage Service (Amazon S3) input files (using file IDs), allowing users to trigger model training, data prefetching, and other custom notebook tasks without leaving the platform.

For more information, see Invoke a Training Notebook in the Run an Inference (AI Services v1.2.x) guide.

Knowledge Base and Vector Store Support

TetraScience AI Services now supports the creation and management of vectorized knowledge bases, powered by Databricks Vector Search. This capability enables AI use cases that require retrieval-augmented generation (RAG) and semantic search over enterprise knowledge bases.

Key capabilities include:

• Vector store creation and management: Create, update, and delete vector stores scoped to your organization with role-based access control.
• Customizable vectorization: Configure how knowledge base content is vectorized at creation time, including the embedding model, chunking strategy, chunk size, and chunk overlap, to optimize retrieval accuracy for your data and use cases.
• Text-based and vector-based querying: Query vector stores using natural language text directly, without needing to manually generate embeddings.
• Knowledge base file uploads: Upload and manage knowledge base files through the AI Asset Files API, with support for multipart uploads for large files.
• Permission model: Organization-level vector stores are accessible to all users within the organization, with role-based permissions for create, read, update, and delete operations.

The following new API endpoints support this functionality:

Model Promotion (Beta)

Customers can now promote AI assets (models and tables) between TDP environments (for example, from a development environment to a production environment) using Databricks Delta Sharing. This new beta release feature provides controlled, auditable model deployment across your organization's environments without manual data copying.

Model Promotion is available as part of a beta release and must be enabled in coordination with TetraScience. For more information, see Promote Assets Between Environments in the TetraScience AI Services User Guide (v1.2.x).

Model Aliases

AI Workflows can now use model aliases to reference model versions by lifecycle stage instead of by version number. Supported aliases include dev, staging, canary, champion, prod, and rollback. Updating an alias automatically updates endpoint routing without requiring redeployment, enabling fast and safe model promotions and rollbacks.

For more information, see Model Aliases in the TetraScience AI Services User Guide (v1.2.x).

Enhancements

Enhancements are modifications to existing functionality that improve performance or usability, but don't alter the function or intended use of the system.

Multipart Upload Support for AI Assets

The AI Asset Files API now supports multipart uploads with pre-signed URLs for large model weights and knowledge base files. This enables reliable upload of files that exceed single-request size limits, with automatic recovery from orphaned upload sessions.

System Log Improvements

Activating, cancelling, and changing the version of an AI Workflow are now captured in the System Log, providing a complete compliance record for GxP validation.

For more information, see System Log.

Bug Fixes

The following bugs are now fixed.

• The AI Asset upload API no longer returns an incorrect S3 path containing an extra /schemas segment, which previously caused uploaded assets to be unreachable from Databricks.
• The concurrent upload slot counter now automatically recovers from orphaned upload sessions, preventing organizations from being permanently blocked from uploading after sessions expire without being completed or aborted.
• The real-time inference curl command now correctly includes the namespace when an AI Workflow is installed in the common namespace.
• The AI Services UI now displays the correct version number.

Deprecated Features

There are no deprecated features in this release. For more information about TDP deprecations, see Tetra Product Deprecation Notices.

Known and Possible Issues

The following are known limitations of TetraScience AI Services v1.2.0:

• Task Script README Parsing: The platform uses task script README files to determine input configurations. Parsing might be inconsistent due to varying README file formats.
• AI Agent Accuracy: AI-generated information cannot be guaranteed to be accurate. Agents may hallucinate or provide incorrect information.
• Databricks Workspace Mapping: Initially, one TDP organization maps to one Databricks workspace only.
• Analyst Policy Install Workflow UI Access: Users assigned a role with an Analyst policy can view, but not use, the Install Workflow button even though the policy doesn't allow them to install an AI workflow or run an inference. A fix for this issue is scheduled for a future AI Services UI release.
• Vector Store Creation Permission Error: Vector store creation may fail with a 403 PERMISSION_DENIED error when the service principal running the creation job does not have explicit permissions on the vector search endpoint. Status: Known since v1.2.0. Fixed in v1.2.1.
• Transient Vector Store Query Errors: Querying a vector store immediately after it reaches ready status may return a transient 502 Databricks Model Serving error. The error typically resolves within approximately 30 minutes. Status: Known since v1.2.0. Fixed in v1.2.1.

Upgrade Considerations

To upgrade to the latest AI Services version, see Update the AI Services UI version in the TetraScience AI Services User Guide (v1.2.x).

During the upgrade, there might be a brief downtime when users won't be able to access AI Services functionality.

After the upgrade, verify the following:

• Existing AI Workflow installations continue to function correctly
• Inference requests complete successfully
• Knowledge base and vector store operations function as expected
• Model training API endpoints respond correctly
• The AI Services UI displays correctly

📘

Security

TetraScience continually monitors and tests the codebase to identify potential security issues. Security updates are applied on an ongoing basis.

📘

Quality Management

TetraScience is committed to creating quality software. Software is developed and tested following the ISO 9001-certified TetraScience Quality Management System.

For instructions on how to validate core AI Services functionality programmatically, see Test TetraScience AI Services Programmatically in the TetraConnect Hub. For access, see Access the TetraConnect Hub.

For more information, see the TetraScience AI Services documentation.

Other Release Notes

To view other TetraScience AI Services release notes, see TetraScience AI Services Release Notes.