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

# Release Notes

## Karini AI Platform - Release Notes

This comprehensive documentation covers all releases for the Karini AI platform. Releases are organized by month with the most recent updates appearing first.

***

#### July 2026 <a href="#docs-internal-guid-4db0040d-7fff-a021-408e-f0860cfb1d0f" id="docs-internal-guid-4db0040d-7fff-a021-408e-f0860cfb1d0f"></a>

Release Date: Jul 10, 2026&#x20;

Sentinel Version: v1.9.70.19

CoreML Version: v1.10.25.13

## 🌐 Community Experience

We've introduced a streamlined onboarding experience for community users along with self-service knowledge management capabilities.

#### ✨ Community Onboarding

* New community sign-up and onboarding completion experience
* Improved verification flow with clearer guidance throughout the onboarding process
* Refined onboarding journey with tailored guidance and a smoother user experience

#### 📚 Community Knowledge Bases

Community users can now create and manage their own Knowledge Bases directly from the platform.

Key capabilities include:

* Upload documents through a dedicated Knowledge Base page
* Automatic validation of upload limits and supported file types
* Start ingestion directly from the interface with Save & Run
* View and manage Knowledge Bases from a centralized dashboard

***

## 💰 Budgeting & Usage Controls

Organizations now have greater visibility and control over platform usage and costs.

#### 📊 Organization Budget Management

* Configure daily and monthly budgets at the organization level
* Set spending limits for Recipes, Copilots, and user roles
* Override budget limits where needed while maintaining organization-wide governance
* Clear summaries and dashboards make monitoring usage easier

#### 🚦 Usage Limits & Rate Limiting

Expanded governance capabilities help ensure fair usage and platform stability.

* Configure request rate limits for users, recipes, and workflow nodes
* Control workflow execution frequency and resource consumption
* Clear notifications when usage or budget limits are reached

***

## ⚡ Semantic Cache Node

Introducing the new Semantic Cache node for workflow recipes, designed to improve performance by intelligently reusing previous responses.

#### ✨ Key Capabilities

* Reuse responses for repeated or semantically similar requests
* Deliver faster responses while reducing AI processing costs
* Configure cache hit and cache miss behavior within workflows
* Enable semantic caching only where it provides value for your workflows

***

## 📊 Copilot Improvements

Copilot continues to evolve with better analytics and conversation management.

#### 📈 Dashboard Enhancements

* Pin charts directly to personalized dashboards
* Refresh and rearrange dashboard widgets
* Build customized analytics views for your organization

## 🤖 Workflow & AI Enhancements

#### 🧠 Deep Agent Improvements

* New option to disable conversation history for stateless workflows
* Improved execution continuity for long-running and multi-step agents

#### 💻 Code Interpreter

Added support for VPC connectivity, enabling secure access to private network resources for enterprise deployments.

#### 🎨 Workflow Authoring

* Larger prompt variable and validation fields for improved readability
* Improved labels, descriptions, and tooltips across recipe configuration settings for a more intuitive authoring experience
* Added node icons to the configuration panel for better visual context and easier identification of selected workflow nodes
* Enhanced Prompt Playground and experiment workflows with support for multiple input variables
* Richer experiment results with full-screen viewing and easier result sharing

***

## 🔐 Security & Platform Reliability

* Continued security hardening through routine vulnerability remediation and dependency updates
* Added an additional email verification step for password-based sign-ins, requiring a verification code sent to the registered email to help protect accounts from unauthorized access.
* Improved platform stability, error handling, and operational reliability

***

#### June 2026 <a href="#docs-internal-guid-1d0908da-7fff-91c2-d5b9-738c60e1f3ba" id="docs-internal-guid-1d0908da-7fff-91c2-d5b9-738c60e1f3ba"></a>

Release Date: Jun 15, 2026&#x20;

Sentinel Version: v1.9.66.6

CoreML Version: v1.10.21.5

### ✨ New Features

#### 🗄️ Pinecone Vector Database Support

Pinecone is now supported as a vector database provider for enterprise search and retrieval use cases.

* Support for dense vector indexes and integrated embeddings
* Full-text search (FTS) support
* Advanced filtering using metadata, keywords, and access controls
* Namespace and index configuration support
* Dynamic metadata filtering for more accurate search results

#### 🤖 Expanded Model Support

Added support for the latest Anthropic models:

* Claude Opus 4.8
* Claude Fable 5 (Global)
* Claude Opus 4.5

Includes improved support for advanced reasoning and deployment flexibility.

#### 📄 Document Writer Tool

Introducing the new Document Writer Tool for creating and managing files within workflows and agents.

* Write files directly to S3
* Flexible file path configuration
* Multiple authentication options for storage access
* Improved validation and reliability for document operations

#### 🌐 New Web Search Tools

Added support for:

* Tavily
* Google
* DuckDuckGo

Allowing agents to retrieve and use information from external web sources.

***

### 🧠 NLA Agent Builder Enhancements

* Expanded and refined built-in skills for validation, tool orchestration, and workflow design
* Improved agent-building experience with updated skill configurations
* Enhanced consistency across agent templates and workflows

***

### 🔌 Connector Enhancements

#### 🗃️ MSSQL Improvements

* More secure query execution
* Improved schema discovery and validation
* Better reliability and error handling

***

### 🎨 Workflow & User Experience Improvements

* Simplified workflow canvas with a cleaner visual experience for Agent and Deep Agent recipes
* Improved workflow readability by removing unnecessary tool indicators
* Enhanced usability and consistency across workflow-building experiences

***

### ⚡ Platform Improvements

* Improved vector search configuration and observability
* Enhanced agent execution tracking and workflow visibility
* Optimized deployment performance and infrastructure efficiency
* Expanded support for Microsoft Office document formats and file detection

***

### 🛠 Enhancements & Fixes

* Improved reliability for document storage and retrieval workflows
* Enhanced search accuracy and metadata handling
* Better timeout validation and operational stability

***

Release Date: Jun 8, 2026&#x20;

Sentinel Version: v1.9.65.9&#x20;

CoreML Version: v1.10.20.7

### 🤖 Expanded Model Support

Added support for new global Anthropic model endpoints to provide greater deployment flexibility across regions:

* Claude Sonnet 4.6
* Claude Haiku 4.5

This enables organizations to deploy AI workloads more flexibly while maintaining access to the latest model capabilities.

***

### 🔌 LiteLLM Integration

Added support for LiteLLM, enabling organizations to simplify model management and routing across multiple AI providers through a unified interface.

Key benefits include:

* Simplified model access and management
* Greater deployment flexibility
* Improved support for enterprise AI architectures

***

### 🗄️ Knowledge Base Enhancements

Improved knowledge base management within the Prompt Playground by introducing a unified Knowledge Base experience across tools.

* Consistent configuration and selection experience
* Simplified setup for knowledge-driven prompts and workflows
* Better organization of enterprise data sources

#### 🔐 ACL Group Filtering & Mapping

Enhanced access control capabilities with support for ACL group filtering and mapping, helping ensure users only access content they are authorized to view.

***

### 📧 Prompt Playground Improvements

Added a new Email Draft Generation tag to simplify the creation of email-focused prompts and workflows.

This makes it easier to generate structured email content directly from prompts and agents.

***

### 📊 Dashboard & Cost Visibility Enhancements

Expanded Copilot and dashboard analytics with improved visibility into usage and cost trends.

#### 📈 Monthly Rollups

* Added monthly rollup views across Copilot and dashboard history pages
* Improved long-term usage and trend analysis

#### 💰 Cost Insights

* Added Cost per Request visibility in Copilot history
* Added cost-based filtering capabilities
* Improved cost tracking and operational transparency

***

### 🧩 Workflow & Scheduler Improvements

Improved webhook recipe scheduling by removing the EventBridge dependency, resulting in a more streamlined and reliable scheduling experience.

***

### 📘 API & Integration Enhancements

Expanded and improved Swagger API documentation with additional endpoints and enhanced API coverage.

This makes it easier for teams to integrate, automate, and build on top of the platform.

***

#### May 2026 <a href="#docs-internal-guid-cb0fd363-7fff-0b2a-fd36-8286e7bd987b" id="docs-internal-guid-cb0fd363-7fff-0b2a-fd36-8286e7bd987b"></a>

Release Date: May 22, 2026&#x20;

Sentinel Version: v1.9.63.2&#x20;

CoreML Version: v1.10.19.3

## 🔐 Security & Reliability Improvements <a href="#docs-internal-guid-b31cad32-7fff-99e4-e2d1-d5bb1c3878f7" id="docs-internal-guid-b31cad32-7fff-99e4-e2d1-d5bb1c3878f7"></a>

We’ve implemented major security, infrastructure, and platform reliability enhancements across the product.

#### 🛡️ Security Enhancements

* Fixed multiple security vulnerabilities including [CVE-2026-25896](https://github.com/advisories/GHSA-m7jm-9gc2-mpf2), [CVE-2025-7783](https://github.com/advisories/GHSA-fjxv-7rqg-78g4), [CVE-2025-29927](https://github.com/advisories/GHSA-f82v-jwr5-mffw), [CVE-2026-31938](https://github.com/advisories/GHSA-wfv2-pwc8-crg5), [CVE-2026-40393](https://github.com/advisories/GHSA-w54f-pw7x-c532), [CVE-2026-31789](https://github.com/advisories/GHSA-j79m-9jxq-788r), [CVE-2026-34000](https://github.com/advisories/GHSA-x7p4-8jjf-qcp4), and [CVE-2023-45853](https://github.com/advisories/GHSA-mq29-j5xf-cjwr)
* Improved webhook security, dependency management, and secure configuration handling
* Upgraded platform packages and infrastructure components to address known vulnerabilities and improve overall platform security

#### ⚙️ Reliability Improvements

* Enhanced error handling, validation, and timeout management across services
* Improved database connector stability and recovery handling
* Better concurrency handling and operational reliability for workflow updates and background processes

***

## 🤖 Agent & Workflow Enhancements

#### 🧠 NLA Improvements

* Enhanced NLA workflows and overall agent orchestration experience with improved usability and reliability.

#### 📧 Enhanced Email Tool

Expanded email capabilities with support for:

* CC and BCC recipients
* Attachments
* HTML content support

#### 🧩 Prompt Template Enhancements

Added new template insertion options for rich content workflows:

* Image tags for S3 references
* Citation tags for external web sources
* S3 document citations
* Sitecore CMS citations
* Reference tags for linked documents
* Video tags for S3-hosted video content

***

## 🔗 MCP Improvements

* Faster MCP page loading with improved pagination and search
* Better handling of large MCP server lists
* Improved remote MCP server connectivity and usability

***

## 📊 Observability

* Added clickable links within Copilot traces and responses
* Enhanced monitoring and debugging capabilities across workflows

***

## 🏢 License & Organization Management

* Added multi tenancy support for licenses
* Organization limits are now managed based on license tiers

***

## 🗄️ PostgreSQL Knowledge Base Support

Introducing PostgreSQL as a fully supported Knowledge Base with advanced ingestion, semantic search, and enterprise access control capabilities.

* Connect your own PostgreSQL database as a knowledge base with high-performance vector search powered by pgvector
* Ingest structured data directly from CSV and Excel files with automatic column type detection and simplified configuration
* Combine multiple columns into a single searchable field for richer semantic search and improved retrieval quality
* Support for flexible formatting options including comma-separated, space-separated, key-value, and custom formats
* Dynamic metadata filtering automatically narrows search results based on user queries using LLM-powered filtering
* Row-level ACL-based access control ensures users only see documents they are authorized to access
* Automatic schema change detection rebuilds the knowledge base when table mappings or configurations are updated

<br>

***

## 🎨 UI & Workflow Builder Improvements

* Improved charts, tooltips, and workflow editing experience
* Better validation and stability for recipe node connections
* Getting Started pages now display only relevant quick actions for power users, including Create Prompt and Create Recipe

***

## ⚡ Infrastructure & Performance Improvements

* Optimized Docker multi-stage builds for faster deployments and smaller image sizes
* Improved Python 3.12 environment consistency support
* Better connection pooling and resource management

***

## 🛠 Enhancements & Fixes

* Improved error handling and operational stability across workflows and integrations
* Corrected actor naming consistency in Ray jobs
* Improved access controls on evaluation runs
* Fixed form validation for prompt publishing workflows
* Various backend performance optimizations and bug fixes across the platform

***

Release Date: May 18, 2026&#x20;

Sentinel Version: v1.9.62.1&#x20;

CoreML Version: v1.10.18.3

\
🎨 Workflow Builder UI Enhancements

Improved the Recipe node experience with a more modern and intuitive interface.

#### ✨ What’s New

* Modernized node icons and visual styling
* Manual node resizing support
* Improved connector line rendering and workflow readability

These updates provide a cleaner and more flexible workflow-building experience.

***

## 🤖 Deep Agent Improvements

Enhanced Deep Agent capabilities with better summarization, observability, and sub-agent handling.

#### 🧠 Enhancements

* Improved summarization middleware behavior
* Added custom read\_file token limit enforcement
* Enhanced sub-agent descriptions for better workflow clarity

#### 📝 Trace Logging Improvements

* Added detailed token counter logging for Deep Agent summarization middleware
* Improved trace visibility and debugging support

***

## 🗄️ Knowledge Base & Data Integration Enhancements

#### 🐘 PostgreSQL Knowledge Base Support

Added PostgreSQL as a supported knowledge base source.

#### 📊 Improved Data Mapping

* Combine multiple columns during configuration
* Apply custom formatting options for better indexing and retrieval quality

***

## 🔌 MCP Support for NLA Agent Builder

Added MCP integration support for the NLA Agent Builder, enabling expanded interoperability and external tool orchestration capabilities.

***

## ❄️ Snowflake Connector Enhancements

Improved Snowflake integration with enhanced query and warehouse management support.

#### ✨ Enhancements

* Warehouse selection support
* Database and schema context configuration for unqualified SQL queries
* Improved query execution flexibility

***

## 🌐 URL Connector Improvements

Extended web crawling capabilities with iframe source support for broader and more comprehensive content extraction.

***

## 📄 Excel Processing Optimization

Refactored the Excel extraction pipeline for improved efficiency and metadata handling.

#### 🚀 Improvements

* Better token counting accuracy
* Enhanced metadata processing
* Improved overall extraction performance

***

## 🛠 Enhancements & Fixes

* Minor UI refinements across workflow and connector experiences
* Improved backend stability and processing reliability
* Performance optimizations for data ingestion and extraction workflows
* Various bug fixes and operational improvements

***

Release Date: May 8, 2026&#x20;

Sentinel Version: v1.9.61.11&#x20;

CoreML Version: v1.10.16.4

## 🤖 NLA Agent Builder <a href="#docs-internal-guid-201f8904-7fff-0604-2c75-8f60425948c1" id="docs-internal-guid-201f8904-7fff-0604-2c75-8f60425948c1"></a>

Introducing the new NLA Agent Builder for creating intelligent agent workflows with a more streamlined and scalable orchestration experience.

* Faster agent creation and configuration
* Improved workflow composition
* Enhanced extensibility for enterprise use cases

***

## 🧠 Deep Agent Enhancements

Deep Agent continues to evolve with major upgrades focused on usability, scalability, and observability.

#### ✨ New Skills Support

* Added Skills capability to Deep Agent
* Dynamic skill loading and resolution
* Support for both native and custom skills
* Improved skill caching and middleware-based serving

#### 📡 Real-Time Event Streaming

* Enhanced live streaming support with improved event visibility
* Better nested workflow tracking for sub-agents
* Improved event summaries and UI-friendly activity descriptions

#### 🔀 Improved Model Fallback & Timeout Handling

* Enhanced fallback behavior across models
* Better timeout propagation and retry handling
* Improved resiliency for long-running agent tasks

***

## 🔧 Workflow Builder Improvements

#### 🧑‍💻 Human-in-the-Loop (HITL) Node Support

Added a new HITL node to workflow recipes.

* Multiple incoming and outgoing edge support
* User decision-based resume handling
* Multi-path routing support
* Improved state management and rollback handling

#### 🛡️ Guardrail Node Support

Introduced Guardrail Nodes in workflow recipes.

* Integrated testing APIs
* UI integration support
* Safer workflow execution and validation

***

## 🎨 Prompt Playground & UX Improvements

#### ✨ Prompt Playground Updates

* Improved UI experience with expandable text areas in popups
* Better usability for large prompts and configurations

#### 📊 Copilot Dashboard Enhancements

Expanded analytics and dashboard capabilities:

* Multi-select metrics support
* Counter cards for:
  * Unique users
  * Unique sessions
  * Requests
  * Tags

#### 👀 File Upload Support

Added file upload capability for:

* Summarization prompts
* Classification prompts

***

## 🔐 Enterprise Governance & Licensing

#### 📜 License Manager

Introduced a new License Management System.

* Package-based licensing support
* Dedicated license dashboard
* Usage visibility and tracking

#### 💰 AWS Partner Revenue Measurement Support

Added support for AWS Partner Revenue Measurement (PRM) integrations and reporting.

***

## ⚡ Reliability & Resilience

#### ⏱️ Invocation Timeout Configuration

Added configurable timeout handling with per-attempt controls, graceful cleanup, and improved retry budgeting.

#### 🔄 Enhanced Retry & Error Handling

Improved resiliency with smart retry classification, adaptive exponential backoff, and better handling for AWS, Bedrock, network, and streaming failures.

***

## 📂 Data Processing & Ingestion Improvements

#### 📄 Auto Chunking Strategy

Added automatic chunking strategies for different file types to improve ingestion quality and retrieval performance.

#### 📊 Data Ingestion Enhancements

* Improved paused run cancellation support
* Better archive mutation rollback handling
* Enhanced reranker timeout configuration and performance optimization

***

## 🔗 Connector & Media Enhancements

#### 🎙️ Amazon Transcribe Enhancements

Added diarization support to Amazon Transcribe output formats for improved speaker separation.

#### 📺 YouTube Connector Improvements

Enhanced YouTube ingestion and metadata processing:

* Batch video metadata fetching
* Privacy status filtering
* Improved playlist/channel handling
* Enhanced metadata enrichment

#### 🌐 Sitecore Connector Improvements

* Image URL database parameter support
* Metadata handling improvements
* Checksum exclusion enhancements

***

## 📈 Observability & Logging

#### 📝 Improved Logging & Monitoring

Enhanced debugging and observability capabilities:

* MongoDB connection timing metrics
* Structured event logging
* Improved context preservation
* Better event formatting for parsing and analytics

***

## ⚙️ Infrastructure Improvements

#### 🚀 Connection & Performance Optimizations

* Reduced connection overhead
* Improved connection pooling behavior
* TCP keepalive support for long-lived connections
* Better adaptive timeout defaults

***

## 🛠 Enhancements & Fixes

* Minor UI refinements across platform experiences
* Improved backend reliability and workflow stability
* Enhanced error handling and callback URL construction
* Performance optimizations across agent execution and integrations
* Various bug fixes and operational improvements

***

#### April 2026 <a href="#docs-internal-guid-9fb07d56-7fff-82be-5a98-603a640d3233" id="docs-internal-guid-9fb07d56-7fff-82be-5a98-603a640d3233"></a>

Release Date: April 13, 2026&#x20;

Sentinel Version: V1.9.56.3&#x20;

CoreML Version: v1.10.9.6

### 🔒 Security & Performance <a href="#docs-internal-guid-bf96e98b-7fff-d48a-b9e8-c16cb7b197cd" id="docs-internal-guid-bf96e98b-7fff-d48a-b9e8-c16cb7b197cd"></a>

Security Hardening

* Enhanced password security with intelligent rate limiting for incorrect attempts
* OTP attempt limits for forgot password flow
* Upgraded deprecated packages to address vulnerabilities
* Automatic session validation after critical account activities

Performance Optimizations

* Optimized Observability queries with native app-level query generation, eliminating third-party dependencies
* Memory monitoring endpoints with automated garbage collection and cleanup
* Ray framework integration for webhook recipes, enabling parallel processing in isolated environments

### 📊 Copilot Enhancements

Advanced Analytics Dashboard

* New metrics-focused dashboards in copilot history: latency, TTFT (Time to First Token), request routing patterns
* Real-time metric rollups with historical comparison

Enhanced Filtering & Data

* New time-based filters in copilot history table
* Copilot history filters for external users
* Guardrail and feedback filtering options
* Support for CSV, XLS, XLSX, and WAV file uploads

Recipe Management

* Enhanced recipe import functionality
* Prompt version preservation across imports

### 🤖 Deep Agent Platform

Core Agent Capabilities

* Complex workflow orchestration with enhanced event logging
* Sequential event streaming for real-time monitoring
* Improved interruption handling UI

New Agent Tools

* 🧠 Think Tool: Generates and streams agent reasoning process
* 💻 Code Interpreter: Execute code within agent workflows
* 🔌 MCP Support: Full integration with traces in recipes and prompts

Runtime & Execution

* Abort capabilities for Karini Runtime and Agentcore recipes
* Webhook request status polling
* Enhanced error handling and recovery

### 🎯 New Nodes & Tools

Workflow Nodes

* Pass Through Node: Routes responses to next node without modification
* Guardrail Node: Content filtering and safety controls with comprehensive filter support

KnowledgeBase Node

* Teradata EVS integration support
* Azure AI Search integration support (parity with OpenSearch)

Browser Use Tool

* File upload capability
* Client ID and request ID propagation in browser configuration

### 🔗 Integrations & Connectors

SSO & Authentication

* Ping Identity SSO provider support

Data Sources

* Sitecore connector enhancements: video support, caching optimization

AI Services

* Custom SageMaker reranker endpoints with configurable response parsing
* Amazon Transcribe diarization output format option
* OCR Textract output format selection (JSON or plain text)

### 🧠 Model Hub Updates

* Claude Sonnet 4.6 and Claude Opus 4.6 now available
* Service tier selection for Amazon Bedrock models
* User unable to add Deprecated models in model hub

### 🎨 Agent Hub

Agent Creation & Management

* Create custom agents from pre-built agent templates
* Workflow recipe validates all Required fields so user can navigate to appropriate resources
* Agent selection from curated library

### 🛠️ Developer Experience

Knowledge Base Management

* Pause/unpause KB recipes when dataset changes based on user configured limits for adds/deletes/change events
* Admin confirmation workflow before unpausing recipe

Infrastructure

* Automated migration scripts with integrated change logs
* Streamlined deployment process

Prompt Engineering

* Image upload support in summarization and classification prompt nodes
* Consistent experience across agent and standalone prompts

### 🐛 Bug Fixes & UI Enhancements

* UI/UX improvements across platform components
* Enhanced error messaging and user feedback
* Performance optimizations for smoother user experience

***

#### February 2026

**Release Date:** February 10, 2026&#x20;

**Sentinel Version:** v1.9.46.4&#x20;

**CoreML Version:** v1.9.99.3

### 🧠 Deep Agent (Beta) Launch

✨ Deep Agent (Beta) is now live!

* 🤝 Human-in-the-loop support for guided agent execution
* 📡 Live event streaming for real-time visibility into agent workflows
* ⚡ Designed for complex, multi-step orchestration scenarios

This marks a major step toward enterprise-grade agentic automation.

### 🧩 Recipe & Prompt Enhancements

* 🆕 Added new prompt type support for:
  * Deep Agent
  * Node-level prompts in Recipes
* 🔐 Introduced OAuth2 authentication method in:
  * REST API Tool (Prompt Playground)
  * Recipe-level integrations
* 🧵 Improved recipe processing with a detached actor implementation:
  * Optimized memory usage
  * Reduced throttling risks
  * Better isolation and stability

### 🔌 MCP & External Integrations

* 🌍 Remote MCP servers integration support
* 📂 Support for MCP recipes via external APIs, including file handling

### 🔐 Advanced Authorization Controls

* 👥 Added Access Control Groups at:
  * Node level
  * Agent level

Enables fine-grained permission control to restrict or allow authorized users securely.

### 🤖 Model Support Expansion (Bedrock)

Added support for new models:

* Claude Opus 4.6 – Advanced reasoning and long-context workflows
* GLM 4.7 & GLM 4.7 Flash – Fast multilingual performance
* Kimi K2.5 – Strong reasoning and long document handling
* MiniMax M2.1 – Efficient conversational AI
* Qwen 3 Coder Next – Optimized for coding tasks
* DeepSeek V3.2 – High-performance reasoning model

Expanding your options for performance, reasoning, coding, and multilingual use cases.

### 📺 New YouTube Connector&#x20;

* 🎥 Improved transcript extraction
* 🧠 Updated video embedding conversion logic
* More reliable processing and enhanced semantic search capability

### 📊 Evaluation & Observability

* 📈  Evaluation added for:
  * Agents
  * Webhooks

Improved evaluation metrics for response quality and performance benchmarking.

### 💬 Copilot Enhancements

* 🎨 UI improvements for better usability and responsiveness
* 👀 Admin preview for user conversation threads

### 🔎 OpenSearch Enhancements

* 📝 Added Form-based search
* 🔍 Integrated Vector search
* ⬇️ Dropdown selector for choosing search type

Provides more flexible and powerful search capabilities.

### 💰 Cost & Token Tracking Improvements

* 🔄 Separated Token Rollup into a dedicated background process
* 📊 Improved accuracy for:
  * Global dashboards
  * Cost analytics
  * Token consumption reporting

### 🛠 Enhancements & Fixes

* Minor UI refinements for improved usability and responsiveness
* Performance optimizations across agent and recipe execution
* Various bug fixes and backend reliability improvements

***

#### January 2026

**Release Date:** January 16, 2026\
**Sentinel Version:** v1.9.43\
**CoreML Version:** v1.9.97.1

### 👥 Admin & Permissions

* Mass UpVote/DownVote: Admin users can now bulk upvote/downvote items in Copilot history

### ⚙️ Configuration & Settings

* Organization Page Defaults: Added default model configuration for:
  * Dynamic Filtering
  * Query Rewriting
  * Evaluation Judge
  * Caching
  * Topic Classification

### 🔄 Workflow & Nodes

* Pass-through Node: New node type added to workflow system

### ⚡ Performance Optimizations

* Server-side Pagination: Implemented global server-side pagination with optimized key-based search for faster page loads and reduced memory usage
* Payload Serialization: Added compression to reduce payload size, improving bandwidth utilization and transfer speed

### 🎨 UI/UX Improvements

* Traces HTML Formatting: Enhanced rich text view for trace data

### 🔧 Tools & Integrations

* MCP Tool Type: Added new MCP (Model Context Protocol) tool type
* YouTube Connector: New connector supports fetching single videos or entire channel content

### 🤖 Copilot Enhancements

* File Upload Limit: Increased maximum file size from 30MB to 50MB (📤)
* Dataset Export: Added multi-select export functionality for:
* Fine-tuning datasets
* Node optimization data
* Node evaluation data from Copilot history

***

### November 2025

**Release Date:** November 21, 2025\
**Sentinel Version:** v1.9.8.2\
**CoreML Version:** v1.9.89.7

#### 🎯 Agent Evaluation System

* **Comprehensive Evaluation Framework**: Test agents against datasets with automated metrics for performance validation, per-sample context, and custom prompts for fine-grained testing
* **Agent-Specific Metrics**: Tool Accuracy, Goal Achievement, and Execution Efficiency metrics tailored for agentic workflows
* **Agent Iteration Tracing**: Full iteration-level tracing for debugging agentic workflows and understanding agent behavior
* **Evaluation Runs Tracking**: Monitor evaluation progress and results across multiple runs with detailed dashboards
* **Prompt Playground Evaluation**: Interactive evaluation directly in playground environment for rapid testing
* **Origin Tracking**: Evaluation run tracking with origin field in chatbot requests for audit trails

#### 📊 Prompt Evaluation Enhancements

* **Results Analytics**: Detailed statistical analysis with mean, median, standard deviation, and score distributions
* **Evaluation Dashboard**: Enhanced UI for viewing and managing evaluation runs with filtering and sorting

#### 🔌 MCP Server Management

* **Trusted Agents Catalog**: Pre-configured MCP servers ready for one-click deployment
* **MCP Testing Tools**: Test functionality for MCP tools directly in the playground environment
* **Configuration Validation**: Automatic validation of MCP server configurations on upload to prevent errors
* **Server Lifecycle Management**: Enable, disable, test, edit, and remove MCP servers through intuitive UI

#### 🔍 Vector Search & Retrieval

* **Hybrid Search Dataset API**: RESTful API for hybrid search across vector datasets with programmatic access
* **Search Configuration UI**: Interactive search panel for testing and configuring search settings in real-time

#### 🔗 Data Connectors & Enterprise Integrations

* **Email Connector**: Integration for email data ingestion and processing workflows with document reader agent tools
* **URL Connector Optimization**: String metadata handling with improved performance
* **Sharepoint Enhancements**: Enhanced document extraction for knowledge base nodes
* **Sitecore Connector**: Optimized CMS integration for content management

#### ⚡ Semantic Caching Enhancements

* **Cache Setup Logging**: Detailed logs for semantic cache initialization and debugging

#### 📄 Document Processing & Preprocessing

* **Filename Context**: Document preprocessor includes filename in processed-data context
* **Audio File Ingestion**: Fixed audio file chunking in copilot interactions
* **Archive Processing**: Improved handling of ZIP, RAR, 7Z files with better extraction

#### 🤖 Model Support

* **Amazon Nova Act**: Browser automation integration with configurable timeout settings
* **Embeddings Optimization**: Batch embedding inference via AWS Bedrock for improved throughput

#### ⚙️ Performance & Infrastructure

* **Dependency Updates**: Updated core dependencies for security and performance improvements

***

### October 2025

**Release Date:** October 30-31, 2025\
**Sentinel Version:** v1.8.97\
**CoreML Version:** v1.9.88

#### 📊 Observability & Analytics

* **Grafana Integration**: Enterprise-grade observability with customizable Grafana dashboards
* **Agent Requests Table**: Detailed table view of all agent executions with advanced filtering
* **Performance Trends**: Line charts showing latency, success rates over time for trend analysis
* **Token Metrics**: Comprehensive token tracking across agents and time periods with bar graph visualizations
* **Tool Analytics**: Tool usage frequency, duration trends, execution counts, and performance metrics
* **Custom Queries**: Build custom visualizations and reports with SQL query support

#### 🤖 Model Support & Capabilities

* **Meta Llama 4 Series**: Support for Llama 4 70B and 8B variants for diverse use cases
* **Claude Sonnet 4.5**: Latest Anthropic models with extended context windows
* **Amazon Nova Multi-modal Embeddings**: Configurable dimensions (3072, 1024, 384, 256) for flexibility

#### 🔄 Agentic Workflows Framework

* **Code Interpreter**: Integrated code execution capabilities for dynamic programming tasks

#### 🔍 Vector Search & Retrieval

* **Reranker Integration**: Reranker node for improved search relevance with cross-encoder reranking and Cohere and AWS Bedrock reranker support for relevance optimization
* **Hybrid Search**: Combined keyword and semantic search capabilities for best results
* **Batch Embedding**: Optimized batch inference for embeddings to improve throughput

#### 📦 Batch Processing & Distribution

* **Bedrock Batch Inference**: AWS Bedrock batch job management with status tracking
* **Error Handling**: Graceful failure handling with file skip logic for resilience

#### 🔌 MCP (Model Context Protocol)

* **MCP Selector Component**: Dedicated component for selecting and managing MCP servers
* **Configuration Encryption**: Encrypted credential storage for sensitive MCP server environment variables
* **Server Management**: Enhanced enable/disable functionality for MCP servers
* **Tool Discovery**: Endpoint for discovering available MCP tools

#### 📅 Recipe & Workflow Management

* **Recipe Scheduling**: Schedule automated recipe executions with cron expression support
* **Recipe Versioning**: Version control for recipes with rollback capabilities and version comparison
* **Execution History**: Track all recipe runs with detailed logs and outputs
* **CRON Scheduling**: Job scheduling with cron expressions and timezone support
* **Recurring Tasks**: Automated recipe execution with schedule definitions

#### 📈 Evaluation Enhancements

* **Online Evaluation Matrix**: Dashboard for online evaluation metrics with filtering capabilities
* **Topics Classification**: Topic trends analysis for copilot interactions

***

### September 2025

**Release Date:** September 30, 2025\
**Sentinel Version:** v1.8.82\
**CoreML Version:** v1.9.7

#### 🏢 Enterprise Integrations

* **SAP ODATA Integration**: Native SAP ODATA connector support
* **Zendesk Integration**: Integrate with Zendesk for customer support ticket management with full API support

#### 🌐 Browser Automation Tools

* **Browser Use Integration**: Playwright-based web automation framework
* **Interactive Browser Control**: Chat interface for browser operations
* **Web Scraping**: Dynamic content extraction with anti-scraping bypass capabilities
* **Chromium Support**: Headless browser automation with full Chrome DevTools Protocol
* **Browser Use Tool**: Nova Act integration for browser automation tasks

#### 🎵 Audio & Video Processing

* **Speech-to-Text**: Audio transcription with intelligent chunking support
* **Video Processing**: Video file upload and processing with transcript generation

#### 📦 Recipe & Workflow Management

* **Recipe Import/Export**: Three import options with clone prompts functionality
* **Rollback Feature**: Quickly revert to previous recipe versions on failures
* **Recipe Auto Layout**: Automatic node arrangement for better visualization

#### 🤖 Agent Runtime & Orchestration

* **Agent 2.0 Recipe Enhancements**: Code interpreter tool and hybrid search in knowledgebase nodes
* **Recipe Versioning Improvements**: Version display in copilot history with published\_by tracking

#### ⚡ Performance & Operations

* **Scheduling Enhancements**: End date filters and validation improvements for scheduled tasks
* **Tracing UI Improvements**: Expanded view with collapse/expand all functionality
* **Token Rollup Optimization**: Increased window to 24 hours with frequency of 1 hour for better performance

***

### August 2025

**Release Date:** August 25, 2025\
**Sentinel Version:** v1.8.64\
**CoreML Version:** v1.9.2

#### 🏢 Enterprise Integrations

* **Sitecore Connector**: Integration with Sitecore CMS for content management
* **Box Integration**: Secure file storage and collaboration through Box

#### 🤖 Model Provider Expansion

* **Google Vertex Gemini**: Full Gemini model family support across all versions
* **Azure OpenAI**: Complete Azure OpenAI integration with latest GPT models
* **Context Length Updates**: Extended context windows for supported models

#### 🔍 Large File Context Management

* **Split & Merge Strategy**: Advanced document chunking strategies for optimal retrieval
* **Metadata Extraction**: Entity validation for metadata extraction

#### 🤖 Agent Runtime & Orchestration

* **Semantic Cache**: Initialize semantic cache button for Agent 2.0 for improved response times

#### ⚡ Performance & Infrastructure

* **Token Management**: Improved token rollup with chunking for large arrays
* **Token Optimization**: Automatic token counting and optimization suggestions
* **AgentCore Runtime**: Agentcore Runtime Support with ARM Containers

#### 🔌 MCP & Tools

* **MCP Organization Support**: Organization-level MCP server management
* **Recipe Tool Forms**: Enhanced forms for Box connector and presigned URL connector

#### 🔒 Security & Compliance

* **Okta Authentication**: Enhanced logging and debugging for SSO issues

***

### July 2025

**Release Date:** July 24, 2025\
**Sentinel Version:** v1.8.58\
**CoreML Version:** v1.8.98

#### 🤝 Collaboration Features

* **Draft Mode**: Collaborative editing mode for prompts and agents with real-time updates
* **Rich Text Editing**: Editor with table of contents, custom tables, image upload, hyperlinks, and font styling
* **Visual Context**: Citation visibility controls and image panel in draft mode
* **Share & Collaborate**: Share drafts with team members for review and feedback

#### 🎨 User Experience & Interface

* **Chat Widget Enhancements**: Seed questions, follow-up questions, and close button improvements
* **Visual Capabilities**: Process and analyze images, diagrams, and visual content with enhanced image panel

#### 💻 Developer Experience

* **WebSocket Support**: Real-time bidirectional communication for interactive applications
* **API Documentation**: Comprehensive API documentation with code examples
* **SDK Support**: Python, JavaScript, and TypeScript SDKs
* **Webhook Support**: Outbound webhooks for event notifications
* **Batch APIs**: Asynchronous batch processing endpoints

#### 🔒 Security & Access Control

* **API Key Management**: Organization-level API key validation and secure access control

#### ⚡ EventBridge Integration

* **AWS EventBridge**: Agent 2.0 scheduling with AWS EventBridge

***

### June 2025

**Release Date:** June 27, 2025\
**Sentinel Version:** v1.8.24

#### 📊 Observability & Analytics

* **MCP Server Count**: Dashboard statistics for MCP server usage and monitoring

#### 🔒 Security & Operations

* **Recipe Import Security**: Organization-level checks to prevent cross-org imports
* **Security Scanning**: Integrated security scanning in CI/CD pipeline for vulnerability detection

#### 🎨 User Experience & Interface

* **Full Screen Mode**: Enhanced UI for focused editing experience
* **File Upload Improvements**: Presigned URL support and HTML file upload option

#### 📦 Recipe & Workflow Management

* **JSONPath Validation**: Enhanced regex for jsonpath field validation
* **Scheduled Recipe Fixes**: Improved reliability for scheduled executions

#### 🏪 Catalog & Discovery

* **Agent Marketplace**: API changes for agent discovery and deployment

***

### May 2025

**Release Date:** May 28, 2025\
**Sentinel Version:** v1.8.8\
**CoreML Version:** v1.8.86

#### 🤖 Model Support & Capabilities

* **Thinking Mode Support**: Extended reasoning capabilities for compatible models
* **Reasoning Transparency**: Expose model reasoning process for better debugging and understanding
* **Model Versioning**: Pin to specific model versions for reproducibility
* **Amazon Nova Premier v1**: Latest Amazon Nova flagship model with advanced capabilities
* **Meta Llama 4**: Support for Llama 4 70B and 8B models
* **GPT 4.1 Series**: GPT 4.1, 4.1 Mini, 4.1 Nano with enhanced performance
* **OpenAI O-Series**: O4 Mini and O3 reasoning models for complex tasks
* **GPT 4.5 Preview**: Early access to GPT 4.5

#### 📄 Document Format Support

* **PowerPoint Processing**: PPTX file parsing and extraction
* **Excel Support**: XLSX spreadsheet processing with formulas
* **Legacy Word**: DOC file format support for older documents
* **File Type Detection**: Enhanced MIME type detection for accuracy

#### 🔗 Connector Enhancements

* **SharePoint Connector**: Enterprise SharePoint integration
* **Box Token Fixes**: Resolved authentication issues for stability
* **UTF-8 Handling**: Improved character encoding support for international content

#### 📦 Recipe & Workflow Management

* **Deployment Confirmation**: Enhanced modals for recipe deployment with clear confirmation flows
* **Recipe Tool Forms**: Improved tool configuration interfaces

#### 📈 Performance & Monitoring

* **Performance Monitoring**: Enhanced latency tracking (p50, p95, p99) and error rate monitoring
* **Latency Tracking**: Monitor latencies across all endpoints

#### 🎨 User Experience

* **Loading States**: Improved loading indicators and progress feedback

***

### April 2025

**Release Date:** April 24, 2025\
**Sentinel Version:** v1.6.223\
**CoreML Version:** v1.8.85

#### 🔄 Agentic Workflows

* **Tool Orchestration**: Multi-step tool execution and chaining
* **Agent State Management**: Persistent state across interactions
* **JSON Placeholder**: Template variable resolution in workflows

#### 🌐 Web Content Extraction

* **Web Scraping**: Enhanced content extraction capabilities
* **URL Processing**: Improved URL connector with metadata handling

#### 📦 Recipe Orchestration

* **Conditional Logic**: Add conditional steps and branching in recipes

#### 🤖 Model Support

* **Extended Model Library**: Support for latest models from OpenAI, Anthropic, AWS, and Google
* **Model Fallbacks**: Configure fallback models for improved reliability
* **Function Calling**: Native support for tool use and function calling across models

#### ⚙️ Infrastructure

* **Horizontal Scaling**: Auto-scaling infrastructure for handling load spikes
* **Load Balancing**: Intelligent request routing for optimal performance

#### 🧪 Testing & Debugging

* **Debug Mode**: Detailed execution traces and intermediate results
* **Performance Profiling**: Identify bottlenecks and optimization opportunities

***

### March 2025

**Release Date:** March 28, 2025\
**Sentinel Version:** v1.6.204\
**CoreML Version:** v1.8.70

#### 🛡️ Guardrails & Safety

* **Guardrails Integration**: Comprehensive guardrail system to ensure AI safety and compliance
* **Policy Enforcement**: Configure and enforce content policies, PII detection, and safety constraints
* **Real-time Monitoring**: Track guardrail violations and policy adherence in production

#### 🕸️ Neptune Graph Database

* **Amazon Neptune**: Production-grade graph database integration
* **Graph Data Processing**: Entity and relationship extraction

#### 🔍 Vector Search & Retrieval

* **Dynamic Metadata Filtering**: LLM-powered filter generation from natural language queries
* **ACL Prefiltering**: User and group-level access control for search results

#### 🏪 Catalog & Discovery

* **Catalog Export**: Export functionality for agent catalog

#### 📦 Recipe & Workflow Management

* **Preprocessing Enhancements**: Improved data preprocessing capabilities
* **Webhook Recipe Fixes**: Stability improvements for webhook-based recipes

#### 🔧 System Operations

* **Health Check Endpoint**: System status monitoring
* **Error Reporting**: Enhanced tracing and debugging capabilities
* **Database Optimization**: Query optimization and indexing

#### 🔒 Security & Encryption

* **Document Encryption**: Secure document storage capabilities
* **Credential Management**: Enhanced secret management
* **Access Control**: Authentication and authorization improvements

#### ⚙️ Infrastructure

* **Helm Chart Updates**: Enhanced Kubernetes deployment configurations

***

### February 2025

**Release Date:** February 26, 2025\
**Sentinel Version:** v1.6.183\
**CoreML Version:** v1.8.20

#### 🗄️ Data & Storage

* **PostgreSQL Connector**: Table name support in credentials for improved database connectivity

#### 🗄️ Data Management

* **PostgreSQL**: Knowledge base query integration
* **Filtering**: Sensitive information detection and filtering
* **Usage Reporting**: Enhanced metrics and reporting

#### 💻 Developer Experience

* **Presigned Upload URL**: Enhanced API for secure file uploads

#### ⚡ Performance & Infrastructure

* **Database Optimization**: Query optimization and indexing improvements
* **Ray Dockerfile**: Distributed computing optimization
* **Package Updates**: Security and performance updates

***

### January 2025

**Release Date:** January 29, 2025\
**Sentinel Version:** v1.6.152\
**CoreML Version:** v1.8.5

#### 👁️ Visual Language Model (VLM) Support

* **Image Processing**: Process and analyze images, diagrams, and visual content
* **Image Input**: Upload images directly in playground for multimodal interactions
* **Visual Context**: Provide visual context to agents for enhanced understanding
* **Multimodal Dataset**: Support for image and video understanding

#### 🔒 Security & Access Control

* **ACL Implementation**: User and group-level access control for search results

#### 📦 Recipe & Workflow Management

* **Recipe Action Improvements**: Enhanced recipe run execution

***

### December 2024

**Release Date:** December 26, 2024\
**Sentinel Version:** v1.6.103\
**CoreML Version:** v1.7.8

#### 🔍 Vector Search & Retrieval

* **Advanced RAG Techniques**: Query rewrite and multi-query generation for better retrieval
* **Top-K Configuration**: Flexible result count settings with prompt context options
* **Dataset Management UI**: Interactive search panel for testing and configuring search settings
* **Performance Monitoring**: Search latency tracking, relevance scores, and usage metrics

#### 🏗️ Platform Foundation

* **Document Ingestion**: Basic document processing pipeline

#### 🕸️ Knowledge Graph Features

* **Neptune Integration**: AWS Neptune graph database with Gremlin query support
* **Graph Ingestion**: Standard and Ray-distributed graph building for scalability

#### 🎨 User Experience & Interface

* **Copilot Thread Management**: Logical delete for copilot threads
* **Prompt Template Improvements**: Enhanced variable handling in prompt templates

#### 📦 Recipe & Workflow Management

* **Recipe Scheduling Foundation**: Initial scheduling capabilities for automated execution

#### 🛡️ Guardrails & Safety

* **PII Detection**: Sensitive data identification and filtering for compliance
* **Toxicity Checking**: Content safety validation to prevent harmful outputs
* **Custom Guardrails**: User-defined safety rules with output filtering
* **Guardrail Tracing**: Complete visibility into safety filter execution

***

**Questions or Feedback?**

Contact us at <support@karini.ai> or visit our documentation at <https://karini-ai.gitbook.io/karini-ai-documentation/>

**Last Updated**: November 23, 2025
