Changelog

All notable changes to this project will be documented in this file.

The format is based on Keep a Changelog, and this project adheres to Semantic Versioning.

[0.1.4] - 2025-12-26

Zero-Configuration Launcher: New deepagent-lab command that automatically configures Jupyter server settings
- Auto-detects available ports using socket programming
- Generates secure authentication tokens with secrets.token_urlsafe(32)
- Sets DEEPAGENT_JUPYTER_SERVER_URL and DEEPAGENT_JUPYTER_TOKEN environment variables automatically
- Supports all jupyter lab arguments (e.g., --no-browser, --port)
- See JUPYTER_AUTO_CONFIG.md for details
Dynamic Agent Name Display: Chat interface now displays custom agent names
- Reads the name attribute from agent objects
- Updates dynamically when agents are switched via DEEPAGENT_AGENT_SPEC
- Falls back to “Deep Agents” if no name is set
Custom Logo Integration: Extension now uses custom DeepAgent Lab logo
- Theme-aware SVG icon in sidebar and command palette
- Centralized icon definitions in src/icons.ts
- Professional branding throughout the interface

Improved README: Completely restructured documentation
- Launcher command featured as recommended approach
- Manual configuration shown as alternative method
- Dedicated “Using Custom Agents” section with clear examples
- Simplified Quick Start instructions
- Enhanced environment variables reference table
Icon-Only Sidebar Tab: Cleaner sidebar appearance
- Removed label text from sidebar tab (icon only)
- Moved extension to bottom of sidebar for better organization
- Agent name still displayed in chat window header
- Tooltip shows “Deep Agents” on hover

LangGraph Utilities Module: New langgraph_utils.py providing helper functions for LangGraph integration
- Utility functions for agent state management and graph operations
- Enhanced agent execution framework with better LangGraph support

Improved Chat Message Styling: Enhanced message content styling for better readability
- Clearer visual hierarchy and structure
- More responsive and polished UI design
- Better spacing and typography for message content
Refactored Interrupt Handling: Improved interrupt mechanism for human-in-the-loop interactions
- More reliable interrupt processing
- Better clarity in interrupt flow
- Enhanced responsiveness during agent execution
Enhanced Notebook Cell Operations: Updated to use Jupyter Server API for saving changes
- Direct integration with Jupyter Server API for cell modifications
- More reliable cell save operations
- Better notebook state management
Agent Execution Refactoring: Streamlined agent execution logic in agent_wrapper.py
- Cleaner execution flow leveraging LangGraph utilities
- Improved error handling and state management
- Better integration with LangGraph graph structures

Centralized Configuration System: New config.py module with get_config() function for hierarchical environment variable support
- All configuration now uses DEEPAGENT_ prefix (not DEEPAGENT_LAB_)
- Full compatibility with deepagent-dash
- Agents can be shared between deepagent-lab and deepagent-dash seamlessly
File Path Agent Loading: Enhanced agent loading to support both module paths and file paths
- DEEPAGENT_AGENT_SPEC environment variable in format "module_or_file:variable"
- Support for relative paths (./my_agent.py:agent)
- Support for absolute paths (/path/to/agent.py:graph)
- Automatic detection of file vs module paths
Dynamic Workspace Configuration: Workspace root now configurable via DEEPAGENT_WORKSPACE_ROOT
- Automatic workspace discovery for agents
- Environment variable set by extension for agent access
- Dynamic workspace path resolution

Standardized Environment Variables: All variables now use DEEPAGENT_ prefix for cross-library compatibility
- DEEPAGENT_AGENT_SPEC replaces previous agent configuration
- DEEPAGENT_JUPYTER_SERVER_URL for Jupyter server connection
- DEEPAGENT_JUPYTER_TOKEN for authentication
- DEEPAGENT_MODEL_NAME and DEEPAGENT_MODEL_TEMPERATURE for model configuration
- DEEPAGENT_VIRTUAL_MODE for FilesystemBackend safety
- DEEPAGENT_WORKSPACE_ROOT for dynamic workspace paths
Enhanced Security: Updated default Jupyter token from "12345" to cryptographically secure random value
- .env.example includes command to generate secure tokens: python3 -c "import secrets; print(secrets.token_hex(16))"
- Default token: 8e2121e58cd3f9e13fc05fc020955c6e
Streamlined Documentation: Updated README.md with clearer, more concise instructions
- Emphasized critical Jupyter server configuration requirements
- Added environment variables reference table
- Highlighted agent portability between deepagent-lab and deepagent-dash
- Removed verbose sections to focus on essential information
Agent Initialization: Updated agent_wrapper.py with improved loading mechanisms
- Smart detection of file paths vs module paths
- Support for importlib.util for file-based loading
- Workspace root environment variable propagation

Unused Environment Variables: Removed MODEL_MAX_TOKENS and LOG_LEVEL (were defined but never used)

Jupyter Configuration: Corrected Quick Start documentation to use hardcoded values matching .env.example
- Fixed incorrect reference to non-existent DEEPAGENT_JUPYTER_PORT variable
- Updated jupyter lab command with correct port and token values

Stop Execution Button: Added ability to cancel ongoing agent execution with a red stop button that replaces the send button during processing
- Backend cancellation endpoint (/cancel) with thread-safe execution tracking
- Graceful cancellation between streaming chunks
- User feedback when execution is cancelled
Multi-line Input Support: Input box now supports multiple lines for longer messages
- Enter key sends message
- Shift+Enter creates new line
- Manual vertical resizing with drag handle
- Auto-constrained between 40px and 200px height
Non-blocking Execution: Agent operations now run in thread pool to keep Jupyter responsive
- Notebooks remain interactive while agent is working
- No UI freezing during long-running operations
- Thread pool with 4 max workers for concurrent operations

Professional UI redesign with light color palette
- Blue gradient send button with hover effects
- Red gradient stop button with professional styling
- Refined shadows, borders, and spacing throughout
- System messages now left-aligned with subtle gray styling
Improved markdown rendering with compact spacing
- Reduced paragraph padding to 0.1em vertical
- Tighter line height for efficient reading
- Better typography consistency

Fixed raw tool call dictionaries appearing in chat output
Fixed todo list parsing to handle Python-style single quotes using ast.literal_eval()
Fixed todo list display issues after content filtering