New Features

Deep Research Mode

Introducing deep research capabilities for complex analytical workflows:

• Multi-Step Research Plans - Agents can now autonomously plan and execute comprehensive research workflows with multiple transition types
• Structured Research Flow - Define research workflows with start, transitions, and end configurations for query, visualization, and insight generation
• Block-Based Message Architecture - Refactored message storage to support rich, structured content blocks for better research output organization
• Integrated Thread Display - Seamlessly view research progress and results within conversation threads

Deep research mode enables agents to conduct thorough analysis by combining data queries, visualizations, and insights into cohesive research outputs.

Example transition types:

• query - Execute analytical queries against databases
• visualize - Generate data visualizations
• insight - Synthesize findings and generate insights

An example of deep research can be found in our repo 

Enhanced Workflow Output Controls

Improved interaction with workflow execution logs:

• Expand/Collapse All - Toggle visibility of all log items simultaneously with centralized state management
• Copy Functionality - Copy individual log items or all logs at once with dedicated copy buttons
• Better Log Management - Enhanced user control over large workflow execution outputs

Platform Improvements

• GitHub App Webhook Fix - Resolved onboarding issue in cloud mode where users were blocked if GitHub app was installed before deployment
• Workflow Output UX - Added header controls for batch operations on workflow logs
• Rust and Dependencies Update - Updated to latest Rust version and upgraded all dependencies for improved performance and security