Release Notes
ZEBLOK COMPUTATIONAL AI-MICROCLOUD RELEASE NOTES
Gen-AI – Enhanced User Interaction
Implemented streaming responses to enable real-time reply generation for a faster and smoother chat experience..
Each agent now retains its own chat history, maintaining conversation context for improved personalization.
Displayed the active model name in the chat interface to enhance transparency and user understanding.
Added notifications for agent pods triggered by key lifecycle events, enhancing monitoring and awareness.
Fixed a restart issue in Chat Playground caused by a missing resource key in the deployment YAML.
Infrastructure Flexibility and Control
Introduced support for secondary clusters under existing primary clusters, enhancing scalability and management flexibility.
Enabled edit and delete functionalities for additional datacenters, simplifying infrastructure operations.
Updated the BareMetal Cluster table UI to visually align with AKS and EKS layouts for a consistent design experience.
Fixed cluster deletion issues for AKS and EKS clusters to ensure reliable infrastructure management.
AI/ML Development Improvements
Added AI inference sharing, allowing users and teams to securely share inference results across the platform.
Simplified plan creation by removing the redundant Type field, enhancing the user experience.
Decoupled storage configuration from the plan creation process, providing greater flexibility and clarity for developers.
Improved UI responsiveness for Spawn buttons and disabled the Info Panel by default for a cleaner interface.
Resolved plan visibility issues in the AI-API section to ensure all active plans display correctly.
Added SDK support to attach datasets directly to selected datacenters, enhancing data management and mapping.
Platform Security Enhancements
Upgraded Harbor to the latest version to improve container image security and stability.
Enhanced the Kubectl microservice with complete role-based access control (Developer, Admin, User) for secure operations.
Fixed workstation event triggers to ensure accurate execution of start and stop events.
Corrected KD log mapping between models and pipelines for accurate tracking and debugging.
Consolidated client and server logs for improved traceability and faster issue resolution.
AI Agents – Agentic Framework
Deploy, configure, and manage agents effortlessly within the platform.
Agents can now connect directly with Large Language Models for intelligent decision-making and automation.
Agents work seamlessly across multiple services, ensuring reliability and scalability in real-world use cases.
A dedicated interface allows intuitive interaction with agents, making integration and communicating with the agent smoother than ever.
SDK & AIMC Improvements
Resolved orchestration issues during deployments to ensure smoother operations.
Addressed user management issues, ensuring consistent org-level user creation and assignment.
Launcher & Infrastructure Flexibility
The Launcher now supports image uploads during provisioning, enabling greater customization.
Expanded validation ensures both cloud-based and bare-metal clusters operate seamlessly within the platform.
Improved failover handling with clearer issue resolution for uninterrupted continuity.
CaaS (Container-as-a-Service) Reliability
Resolved critical issues that caused ML-flow saved models to fail during containerization.
Smarter Infrastructure, Greater Control
Multiple datacenter support is now native, whether it’s a cluster or a single-node instance.
Object stores are linked to individual datacenters, enabling efficient storage management.
Bucket mounting is datacenter-specific for better isolation and performance.
Select your target data center when spawning workstations, microservices, creating datasets, and other managed services.
Launch datacenter-specific Plans to optimize workloads per region or deployment strategy.
Monitor all datacenters in real-time, directly from MicroCloud.
Gain a complete view of system health, GPU, CPU, memory usage, active users, services, and resource availability.
Gen-AI Workspace, Now Sharper
Knowledge Distillation now supports embedded model references and multi-modular datasets, including PDFs.
Reuse running model instances from the Chat Playground for distillation workflows.
A new organizational dropdown makes reusable model selection fast and intuitive.
Launcher, Now with More Power
Deploy reliable/remote clusters directly through the Launcher.
Seamlessly bring your own clusters, such as EKS or AKS, and register them as full-fledged data centers.
Once integrated, these data centers are fully accessible and manageable within the platform.
Better UX, Tighter Feedback, and a Bug Bounty to the team!
Fixed service creation issues with additional port configurations during microservice edits.
Smoothed out form validations for a cleaner interaction.
Improved pop-up notifications for instant clarity.
Plans now have visibility control; share them as public, private, or organization-specific.
SDK, Enhanced for Gen-AI Workflows
The Gen-AI Toolkit supports seamless text generation using pretrained models.
Built-in SDK utilities allow the retrieval of historical chat data from object storage for smoother session management.
Bugs
Extra ports added via the edit feature don’t create a Service, but ports defined during microservice deployment do.
Pop-up alerts for unfilled form fields improve user experience.
Private plans can be set as Public, Private, or Organization-only, giving users flexible privacy controls.
Knowledge Distillation in Gen-AI Workspace
Our most significant addition in this release is the Gen-AI Workspace featuring Knowledge Distillation:
Automatically extracts key insights by generating question-answer pairs from uploaded documents
Computes and stores vector embeddings for intelligent document interactions
Enables fast, context-aware AI responses across large knowledge bases
This workspace lays the foundation for next-level productivity and intelligent automation.
Secure API Key Management (for Gen-AI Workspace)- (The Gen-AI Key)
To enable secure and flexible use of Gen-AI Workspace features — both internally and through SDKs or external access:
API keys can now be generated, validated, and revoked directly from the credentials tab
Built-in checks for expiry, revocation, and hash-based validation
Each key is linked to a user and tracked via audit logs for accountability
SDK improvement
We have integrated API key authentication into our SDK, and we have also implemented Chat interactions as mentioned below.
API Key-based Authentication
Completion API integration
LLM pretrained models list
LLM Chat integration: Only getting answers to a question. No history of chats
Monitoring Enhancements
Pod-Level GPU Monitoring
View real-time GPU metrics at the pod level for detailed usage insights
Storage Monitoring & Alerts
Track storage consumption with alerting support for high-availability environments
UI Improvements (Dark Mode Compatibility)
Dark mode visibility issues have been addressed across:
Checkboxes, inputs, selects, and form elements
Header, documentation icons, and section labels
Infrastructure & Reliability Updates
Improved system stability and compatibility across deployment environments
Better performance and flexibility for managing object storage and file operations
Bug Fixes
Resolved multi-GPU training issues for large model fine-tuning
Numerous improvements across logging, environment validation, and file handling
Support tickets answered at a high priority and resolved as requested.
GenAI Workspace - The Hottest New Feature!
Seamless AI Collaboration: Introducing the GenAI Workspace, a powerful environment for teams to interact with AI models in real time.
Enhanced AI-Assisted Development: Users can leverage AI-generated insights to accelerate coding, debugging, and problem-solving.
User-Friendly Interface: A streamlined UI designed for efficient interaction with AI-powered tools.
Automation for Workstation to Microservice Conversion
Effortless Transformation: Automates the process of converting Workstations to Microservices using ZBL-CLI.
Streamlined Workflow: Enhances efficiency by reducing manual intervention.
UI Enhancements & AIM Improvements
Improved Navigation & Usability: Streamlined UI for a smoother experience.
Enhanced File Management: Users can now observe uploaded files.
Environment Variable Validation: Ensures all variables are checked before execution.
Configurable Info Cards: Option to enable/disable the info card.
Custom Upload Support: Users can upload their own profile pictures.
Better Form Validation: Improved checks for file uploads.
Revamped Email Templates: Enhanced readability and engagement.
Edit Feature improvement:
A bucket can be attached to any already spawned/running workstation/microservice using the edit feature
The ability to attach an extra PVC attachment to the workstation/microservice has also been introduced
Infrastructure Enhancements
Terraform Script Updates: Improved automation for cloud infrastructure provisioning.
AWS Integration: Enhanced support and optimization for AWS deployments.
Azure Integration: Improved compatibility and performance for Azure environments.
Last updated
Was this helpful?