v3.4.0
6 days ago by Ethan Lin
New Features
- Model Context Protocol (MCP) Support: Complete MCP client implementation with multiple transport types (StreamableHTTP, SSE, Stdio, WebSocket), flexible authentication methods, rate limiting, execution logging, and AI agent integration for consuming external MCP-enabled tools
- OAuth 2.1 Proxy with Scope Authorization: Added OAuth 2.1 authorization server with dynamic client registration, hierarchical scope-based access control, and proxy-signed JWTs for consistent upstream provider authentication
- Fireworks.ai LLM Provider: New AI provider package supporting Fireworks.ai API with Kimi K2.5 model integration for enhanced AI capabilities
- Field-Level Credential Encryption: Advanced security with automatic encryption/decryption for credential values and new Box.com OAuth credential type support
- Multi-Schema Database Support: Enhanced MJCLI migrate command with configurable schema placeholders and improved diagnostic capabilities for complex database deployments
Improvements
- Dashboard Redesigns: Major UX overhauls including Communication Dashboard with MD3 design tokens, redesigned Testing Dashboard with 5-tab Explorer view, Component Studio with tabbed code editor and collapsible panels, and streamlined Scheduling Dashboard
- Enhanced CLI Tooling: Added configurable entity package names in CodeGen, improved schemaPlaceholders support, and better node-flyway diagnostic reporting
- Jest Testing Infrastructure: Comprehensive test suite implementation across AI packages with improved testing capabilities and better error handling
- Metadata Management: New metadata-driven AI model configuration system replacing SQL migrations with declarative JSON files for better maintainability
- Cross-Platform Improvements: Enhanced environment variable parsing for boolean values and more robust cross-platform compatibility
Bug Fixes
- Large Text Field Corruption: Resolved critical data corruption issues affecting large text field storage and retrieval
- AI Prompts Failover: Fixed failover mechanism in AI prompt execution to ensure reliable AI service availability
- Metadata Cleanup: Corrected 84 mismatched AIModelVendor and AIModelCost primary key IDs and removed orphaned API application scope records
- Tree-Shaking Issues: Eliminated Angular build problems with better module exports and dependency management
- Node.js Compatibility: Resolved engine compatibility warnings and dependency conflicts across packages