# Release Overview

This release delivers a robust suite of capabilities to unify tool management, enhance knowledge management, and strengthen benchmarking, governance, and platform reliability.

Knowledge Garden Updates: Google Drive integration, new file summary dashboards, and enhanced retrieval settings with reranking for improved knowledge access and flexibility.
Benchmarking & Governance Improvements: Support for soft/hard error differentiation, mandatory file validation, document access from results, and trace retention for deleted agents to strengthen auditability and reliability.
Credential & Tool Automation: Unified credential management across tools and benchmarks with auto-handling during execution for improved security and reduced manual effort.
Flow Designer & Agent Enhancements: Out-of-the-box, prebuilt, and custom tools now usable in workflows, with a Prime Agent Template for quick setup and expanded connector capabilities for Confluence and Jira.
Enhancements & Bug Fixes: Fixes across Knowledge Garden (upload, sync, reranking), Flow Designer (execution errors, schema) and Governance (trace visibility, metrics).

# 2. Detailed Feature Updates

Below are the key features and enhancements introduced in this release, each designed to address a specific area of your operational transformation journey.

# 1. Unified Tools Management

Overview: A comprehensive centralized interface that consolidates all platform tools, including pre-built connectors, custom functions, APIs, and account credentials, accessible through an intuitive tabbed navigation system.

Benefits:

Single access point for all tool management needs
Streamlined workflow for discovering and configuring automation tools
Real-time search capabilities across all tool categories
Intuitive visual organization with clear status indicators
Quick access to frequently used functions and connectors

Highlights:

Four main sections accessible via tabs:
1. Directory: Pre-built connector catalog with visual cards
2. Custom Functions: Manageable list of custom function that can be built in python
3. APIs: Custom endpoint management
4. Accounts: Credential management interface
- Key Features per Section:
  1. Directory: Visual connector cards with logos and descriptions
  2. Custom Functions: Added standardized input/output schema definition, secure credential attachment, improved testing and debugging and version tracking.
  3. APIs: Endpoint management with status indicators
  4. Accounts: Secure credential management with visibility controls - private and public
- Universal Features:
  1. Runtime credential injection for agents and workflows
  2. Real-time AJAX search functionality
  3. Consistent action menus across all sections
  4. Last modified tracking in user’s local timezone
  5. Quick-access creation buttons for each category
- Enhanced Navigation:
  1. Detailed view access for connectors
  2. Direct editing capabilities for custom functions
  3. Streamlined account credential management
How to Use: Navigate to Expert Agent Studio -> Benchmark -> Benchmark settings -> Run Prompt

# 2. Tool Integration in Flow Designer

Overview:
The Flow Designer now supports seamless integration of OOTB tools, prebuilt tools, and custom functions as components on the canvas, enabling users to create dynamic workflows with easy configuration of inputs and credentials.

Benefits:

Enhanced workflow creation with a wide range of integrated tools
Simplified credential management within the workflow design process
Flexible credential handling for both design-time and runtime scenarios
Improved user experience with intuitive component categorization

Highlights:

Comprehensive tool availability under “Integration” and “Custom” categories
Support for using existing credentials or creating new ones on-the-fly
Validation checks to ensure proper credential association before testing
Ability to publish workflows with or without predefined tool credentials
Seamless output integration with downstream components for complex workflow design

# 3. Credential Management for Tools in Automation/Conversational Agents

Overview:
Enhanced credential management for tools in Automation and Conversational Agents, allowing users to link credentials only when required for authentication, ensuring secure access to external services while maintaining agent functionality.

Benefits:

Improved tool management with enable/disable toggles and removal options
Runtime injection of credentials
Streamlined process for adding and configuring tools within agents
Clear visibility of authentication requirements for each tool
Flexible credential linking options, including using existing or creating new credentials
Enhanced control over credential override permissions for agent creators/admins

Highlights:

Redesigned tools section with improved UX for better tool management
Option to create new credentials directly within the agent configuration flow
Ability to remove linked credentials easily with an ‘X’ icon
Support for publishing agents with or without predefined tool credentials
Credential override controls for enhanced security and management

# 4. Retention of Deleted Agents and Transactions in Governance

Overview:
The Governance module now retains all deleted agents, conversations, and transactions, ensuring full auditability and maintaining a comprehensive governance history without loss of critical tracking information.

Benefits:

Complete audit trail for all agents and transactions, including deleted items
Enhanced compliance and governance capabilities
Improved transparency in agent and transaction management
Consistent data representation across all views and dashboards

Highlights:

Permanent retention of deleted automation and conversation agents, including all associated data
Immutable data storage in the Governance module, preventing unauthorized deletions
Advanced filtering options including status-based and view mode filters
Clear visual indicators for deleted agents and transactions in all relevant screens
Disabled actions for deleted agents with informative hover messages

# 5. Prime Agent Template for a Agent & Team of Agents

Overview:
The Prime template is now the default selection when creating a digital asset, offering users a comprehensive, modular structure for building scalable conversational agents with integrated system instructions, tools, child agents, and knowledge components.

Benefits:

Streamlined agent creation process with a pre-selected, powerful template
Enhanced modularity and scalability in agent design
Prime Template unifies everything into a single agent. You can define perspective, attach knowledge, attach sub-agents, and connect tools - all in one place..
Achieve lower cost with fewer input tokens, faster speed with parallel processing, and higher accuracy

Highlights:

Comprehensive configuration options including Perspective, Knowledge Garden, Tools, and Child Agents
Clear separation between external tools and sub-agents in the UI
Full support for add-ons, benchmarking, traces, and test & debug interfaces
Sequential execution of tool calls based on system instruction logic
Real-time chat and trace functionality in the Test and Debug panel

How to Use: Navigate to Agent -> Prompt template -> Prime

# 6. Enhanced Knowledge Garden Creation Experience

Overview: We’ve significantly improved the Knowledge Garden creation process with updated file upload instructions, a comprehensive File Summary section, and clearer text for advanced settings. These enhancements provide users with better guidance, real-time insights, and improved clarity throughout the Knowledge Garden creation workflow.

Benefits:

Clearer guidance and reduced errors during file uploads
Improved visibility into Knowledge Garden composition and processing status
Enhanced user understanding of advanced features and settings
Streamlined Enterprise Knowledge Garden management experience

Highlights:

Context-specific file upload instructions across the platform, including file size and count limits
New File Summary card view displaying total storage used, file count, and real-time processing status
“Import Data” tab renamed to “Manage Data”
Updated Chunk Settings descriptions for Table Extraction and Image Processing
Consistent implementation across Data Library, EKG, Expert Agent Studio, and governance module
Automatic refresh of file statuses every 5 seconds in the File Summary view

How to Use: Navigate to Knowledge Garden → Manage Data → Files Summary to access the new File Summary view and updated instructions.

# 7. Google Drive Integration for Knowledge Garden Creation

Overview:
Implement a seamless integration between Google Drive and the Knowledge Garden creation process, allowing users to import and manage documents directly from their Google Drive for efficient information retrieval and management.

Benefits:

Streamlined document import process from Google Drive
Expanded source options for Knowledge Garden creation
Secure and permission-based access to Google Drive files

Highlights:

OAuth-based authentication for secure Google Drive connection
Intuitive file/folder selection interface with bulk upload support
Support for various file types including PDF, DOC, DOCX, JPEG, PNG, TIFF
Automatic chunking and indexing of imported documents
Read-only access to Google Drive files, ensuring data integrity

How to Use: Navigate to Knowledge Garden -> Manage Data -> Import -> Drive

# 8. Configurable Retrieval Settings for Knowledge Garden

Overview:
Introduce advanced retrieval configuration options for Knowledge Garden creators, allowing fine-tuning of similarity thresholds, TOP-K results, reranking options, and TOP-N filtering to optimize the relevance and quality of retrieved results for AI agents.

Benefits:

Enhanced control over knowledge retrieval process
Improved relevance and accuracy of AI agent responses
Flexibility to adapt retrieval settings to specific use cases
Optimized performance through customizable reranking options

Highlights:

Adjustable similarity threshold using a slider (range: 1-100)
Configurable TOP-K value for initial result retrieval (range: 1-100)
Dropdown selection for reranker models (when enabled)
Conditional TOP-N setting for filtering reranked results (max: 50)
Immediate application of setting changes

How to Use: Navigate to Knowledge Garden -> Retrieval Test -> Retrieval Setting

# 9. Enhancements

Enhanced Custom Functions with Collaboration Tool Libraries: Pre-installed client libraries for Slack, Jira, ServiceNow, Confluence, AWS, Google, and SQL are now available within Custom Functions, enabling faster, seamless integration and workflow automation across collaboration platforms.
Streamlined Benchmark Configuration Interface: Simplified metric selection by removing read-only system prompts for out-of-the-box metrics, offering a cleaner, more intuitive setup experience.
Enhanced warning message for lenient toxicity settings, providing clearer guidance on potential risks and responsible use within the platform
Direct Document Access in Benchmarking Results: Users can now open full input documents directly from Benchmarking results for seamless comparison between LLM outputs and source content, enhancing validation and review efficiency.
Flow Designer Update Component Enhancements: The “Knowledge Base Update” component has been renamed to “Knowledge Garden Update” with an enhanced output schema, including a new doc_id field for improved clarity and consistency across the platform.
Enhanced error handling in benchmark runs: For a clear distinction between hard and soft errors, improved guidance for resolving issues, and continued metric calculation even with soft errors.
Enhanced multi-tab session handling: To ensure seamless re-authentication and prevent data loss, allowing users to maintain their work context across tabs.
Trace Visibility for Failed Transactions: Users can now view transaction traces for failed transactions for conversation and automation agents, enhancing visibility into error patterns and agent performance.
Automated Credential Management for Tool Execution: Tool execution by agents and flow designers is now fully autonomous, automatically and securely retrieving pre-defined credentials for tools, centralizing security, and eliminating runtime credential management.
Unified Credentials for Benchmarking Tools: Benchmarking now uses a unified credentials system that prompts for credentials only once (on the initial "Run Prompt"), then automatically applies them for all subsequent runs, ensuring efficient and consistent tool access.
Enhanced Conversation Benchmarking with Contextual File Management: Conversation benchmarking now features intelligent file management with multi-file support, automatic propagation, and cascading removal for enhanced control and accuracy.
Connector SDK Tool Integration with Agents: Agents can now seamlessly integrate tools built with the Connector SDK, enabling them to use custom functionalities and interact with external systems for enhanced versatility and capability
Enhanced Confluence & JIRA Connector for Agents: The Confluence & JIRA Connector now supports Read, Create, and Update operations, unifying content management across Flow Designer and Conversational/Automation Agents.
Unified Service-Level Connectors: All feature-based connectors are now consolidated into unified service-level connectors (e.g., AWS, Google), simplifying discovery, configuration, and ensuring a consistent user experience across integrations.
Dynamic Tool Type Display in Notification Traces: Execution traces now dynamically display the correct "Tool Type" next to component labels, offering immediate clarity for faster analysis and troubleshooting.
Old API Tool Support in Unified Tools Module: Existing API tools are integrated into the new Tools module, preserving functionality, ensuring backward compatibility, and preparing for future migration to the new API tool tab.
Deletion of Processing Files in EKG: Users can now delete files in the "Processing" state within the Knowledge Hub (EKG), providing immediate control to remove incorrectly uploaded files and supporting both individual and multi-select deletion.
Improved User Onboarding with Dedicated Workspace Access: User onboarding now restricts access to only specifically assigned workspaces and policies, resulting in a more focused, secure, and personalized platform experience.
Enhanced Search Functionality: The Purple Fabric landing page now features an advanced search bar with intelligent suggestions and fuzzy matching to efficiently locate agents, workflows, and assets across major modules.
Multi-Version Display for Digital Workers: Digital workers (agents/workflows) now default to displaying only the latest version with a collapsible "View versions" option, providing clearer version history management and tracking in both card and list views.
API and Component Metrics Capture for AI Query Engine: A new, configurable metrics capture system records overall request duration, LLM processing time, and internal API timings in the AI Query Engine to provide detailed performance insights and enhance debugging.
Guided Credential Management for Conversational Agents: Conversational Agents now provide a runtime, guided experience to securely select, create, or modify tool credentials for both custom functions and OOTB connectors.
Tenant Admin Assignment in Platform: The user creation process is enhanced to allow existing tenant admins to assign the 'Tenant Admin' role to new users, granting them full platform access and the ability to manage all workspaces and users within the tenant.
Flexible Field Occurrence Definition for Benchmarking: Benchmarking now allows users to define fields as single or multiple occurrences regardless of data type, enabling more precise metrics capture and accommodating complex data structures in LLM outputs.
Export options for Benchmark Results: The benchmarking results can now be exported in CSV and Excel formats, in addition to the already supported JSON format.

# 3. Feedback & Support

We’re committed to continuous improvement—your feedback drives our roadmap! For any assistance, please reach out to Purple Fabric support team at below address.

Email: purplefabric.support@intellectdesign.com