Thunk Release Notes - March 2025
Overview
The March 2025 release brings significant enhancements to the platform with a focus on improved file handling capabilities, UI refinements, agent intelligence, conversational features, and expanded external integrations. Major updates include enhanced file picking support, advanced PDF processing, improved document collections, new workflow management tools, and more intuitive playbook visualization.
Feature Highlights
User Interface Improvements
Playbook & Data Visualization
- Redesigned playbook view with cleaner visualization of schema, step assignments, and status flow
- Added “force human approval” option at the per-step level
- Enhanced step navigation with streamlined controls and improved feedback display
- Improved draft mode view of data with better structure and validation state handling
- Added clear indication of step status in workflow visualization
Dialog & Input Refinements
- Unified bulk upload and add row dialogs for more consistent user experience
- Enhanced folder upload functionality with improved UI expansion logic
- Implemented rich text editor improvements, including inline editing behavior, saving indicators, and better state management
- Fixed issues with AI instructions display and handling
- Added flow stage handling for ‘request’ and ‘response’ stages in workflow management
Layout & Visual Enhancements
- Improved message input layout to be more reusable, particularly for conversation view
- Modified button states with ‘disableRipple’ property and updated theme colors
- Enhanced URL presentation with better display options and formatting
- Added support for displaying user avatars as badges within the AI avatar
- Fixed issues in inbox message positioning and layout
File Handling & Integration
Universal File Picker
- Updated file picker logic to support Google Drive, SharePoint, and OneDrive uniformly
- Added LocalFileSystem as a connection type for local file uploads
- Fixed presentation of file choices in the Data Import dialog
- Enhanced O365 file picker to work everywhere Google does
- Improved handling of file system connections and recognition
PDF & Document Processing
- Added inline image PDF viewer with OCR capabilities
- Enhanced OCR process with parallel conversion of PDF pages to text
- Improved handling of PDF files with embedded URLs
- Added support for extracting and appending external link URLs to OCR output
- Updated the default maximum number of pages per PDF from 3 to 10
Enhanced Document Collections
- Improved RAG indexing to include standard user columns alongside AI formula columns
- Enhanced structured query output format with better organization of document matches
- Added validation and retry logic to query type determination
- Improved accuracy of aggregate query generation
- Added functionality to index collections as part of test setup
Agent Intelligence & Tools
Model Support & Integration
- Added support for “Claude 3.5 Sonnet” as a selectable model in the UI
- Added “Gemini 1.5 Pro” to the list of model options
- Enhanced model selection for custom AI tools with refactored model references
- Updated handling of multiple clients in ThunkUserLlm for improved efficiency
Tools & Capabilities
- Added Maps integration with MapSearchTool and PlaceDetailsTool
- Created CopyDriveFileTool for file copying in Google Drive
- Introduced reflection module with “Note to Self” functionality
- Enhanced import sheet tool with grouping functionality
- Implemented batched execution for workflow request tools
- Added support for ‘mailto’ link format in email drafting tool
Agent Behavior
- Improved message routing efficiency with enhanced instructions for handling messages
- Updated step logic to include specific instructions for positive and negative examples
- Enhanced URL handling with link shortening and expansion capabilities
- Modified content trimming logic in Agent Content Viewer
- Fixed issues with invoking API and DB tools
Conversations & Messaging
Thread Management
- Introduced eaThreads for enhanced conversation management
- Added capabilities for starting and clearing conversations through the UI
- Updated the conversation UI components with new tab content and forms
- Enhanced message input with sendOnEnter functionality
- Added functionality to clear conversation nodes
Routing & Integration
- Fixed bugs in message routing with improved module matching logic
- Enhanced webhook routing efficiency
- Added REST API ‘bucket’ for requests
- Updated routing decision display and management
- Tied message routing to module enabled/disabled status
Security & Infrastructure
Authentication & Authorization
- Enhanced authorization logic with improved permission checks for entity updates
- Added functionality to change ownership of thunks
- Added browserContextId parameter to various components for better session management
- Updated the getAuthToken function with strict type-checking
Performance & Reliability
- Implemented batched tool execution with parallel processing capabilities
- Increased acquire timeout for BrowserBaseSemaphore from 60,000 ms to 300,000 ms
- Enhanced error handling in file uploads and fetches
- Added service.instance.id label to OpenTelemetry configuration
- Improved logging with added context for better debugging
Security Updates
- Updated multiple packages to address vulnerabilities:
- Updated cookie package to version 1.0.2
- Updated axios from versions 1.7.4 and 1.7.9 to 1.8.2
- Updated esbuild to version ^0.25.0
- Updated @babel/helpers from 7.26.0 to 7.26.10
- Updated @babel/runtime from 7.26.0 to 7.26.10
- Updated prismjs from 1.29.0 to 1.30.0
- Updated @mozilla/readability from 0.5.0 to 0.6.0
- Updated vite from 6.2.0 to 6.2.3