Skip to content

CHANGELOG.md

Latest commit

 

History

History
184 lines (141 loc) · 6.85 KB

CHANGELOG.md

File metadata and controls

184 lines (141 loc) · 6.85 KB

Changelog

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

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

[0.6.4] - 2025-11-18

Fixed

  • Path validator now correctly handles heredoc operators (<< and <<<)
    • Fixed PathValidationError for valid commands like cat > /path/file << EOF
    • Heredoc delimiters are now properly skipped during path validation
    • Supports both quoted and unquoted heredoc delimiters

[0.6.3] - 2025-11-17

Added

  • BashToolMessage widget for improved bash command execution display
    • Shows command in header line (truncated at 160 characters)
    • Cleaner, more compact presentation compared to generic ToolMessage
    • Consistent with other specialized tool message widgets

[0.6.2] - 2025-11-11

Changed

  • Updated project dependencies for improved compatibility

Fixed

  • Context handling in CLI mode for file tool operations
  • Failing unit tests after recent changes

Performance

  • Optimized message streaming with smart update throttling for better responsiveness

[0.6.1] - 2025-10-18

Added

  • Automatic clipboard copy on text selection - selected text is automatically copied to clipboard when user releases mouse after selection

[0.6.0] - 2025-10-14

Changed

  • BREAKING: Simplified workflow API with direct user_message parameter
    • Removed context and session parameters from workflow functions
    • Workflows now receive only runner parameter with user_message available directly
    • Made built-in tools accept protocol contexts for better flexibility
  • Architecture Improvements:
    • Introduced runner pattern for better type safety and separation of concerns
    • Consolidated VibecoreApp initialization and session loading
    • Migrated CLI to flow mode architecture for consistency
    • Improved MCP manager lifecycle with async context management
    • Removed agent parameter from VibecoreApp and Vibecore for cleaner API
    • Cleanup of unused context properties and code
  • Removed Features:
    • Removed /clear command
    • Removed detect_reasoning_effort feature

Added

  • Increased ctrl+d prompt time for better user experience

Fixed

  • Context usage now correctly derived from last response
  • User inputs now processed in FIFO order
  • Fixed flaky feedback widget test
  • Fixed context usage calculation
  • Added .git to ignore for pyright

Documentation

  • Updated Flow Mode documentation for v0.5.0+ API changes
  • Updated README to reflect API improvements

[0.5.0] - 2025-10-12

Changed

  • BREAKING: Refactored Flow Mode into flexible multi-mode framework
    • Replaced flow() function with Vibecore class and @vibecore.workflow() decorator
    • Simplified workflow signature: removed explicit app, ctx, user_input parameters
    • Workflow functions now return typed results instead of void
    • Unified interface: user_input(), print(), run_agent() work across all modes

Added

  • Multi-Mode Execution Support: Run the same workflow in different modes
    • run_textual(): Full TUI with streaming (original behavior)
    • run_cli(): Simple stdin/stdout interaction
    • run(): Programmatic execution with predefined inputs (perfect for testing)
  • Runner Architecture: Extensible pattern for adding new execution modes
    • VibecoreRunnerBase: Base class for all runners
    • VibecoreTextualRunner: TUI mode implementation
    • VibecoreCliRunner: CLI mode implementation
    • VibecoreStaticRunner: Static/testing mode implementation
  • New example: basic_cli.py demonstrating CLI and static modes

Benefits

  • Cleaner, more Pythonic API with less boilerplate
  • Type-safe workflow return values with generics
  • Easy to test with static mode and predefined inputs
  • Separation of concerns between runners and workflow logic
  • Extensible: add new modes (HTTP, Discord, etc.) by implementing runner interface

[0.4.2] - 2025-09-30

Added

  • FeedbackWidget: New widget for collecting user feedback

Changed

  • Remove input textarea autofocus on click for improved user experience

Fixed

  • Fix snapshot flakiness for executing message headers in tests
  • Handle list_directory permission test when running as root

Refactoring

  • Consolidate todo item model for better code organization

Documentation

  • Add AGENTS.md and update CLAUDE.md

[0.4.1] - 2025-09-19

Fixed

  • Prevent duplicate message processing during agent execution by ensuring mutually exclusive handling of user input vs queued messages

[0.4.0] - 2025-09-05

Added

  • Path Confinement System: Enhanced security with comprehensive path validation
    • Configurable allowed and blocked paths via settings
    • Prevents access to sensitive system files and directories
    • Validates both file and shell operations
    • Supports pattern-based path restrictions
  • Copy Button for Agent Messages: Added copy functionality to AgentMessage widget for improved user experience

Security

  • Implemented path confinement for file and shell tools to prevent unauthorized access
  • Added comprehensive test coverage for path validation security boundaries

[0.3.2] - 2025-08-25

Fixed

  • Disabled dotenv settings loading since vibecore is a framework library

[0.3.1] - 2025-08-22

Added

  • Copy button for query arguments view to improve user experience

[0.3.0] - 2025-08-22

Added

  • Flow Mode (Experimental): New programmatic conversation control system for building structured agent-based applications
    • flow() entry point for creating custom conversation flows
    • Support for multi-agent orchestration with handoffs
    • Examples adapted from OpenAI Agents SDK (basic.py, customer_service.py)
  • Web Tools: New tools for web interaction
    • websearch tool with extensible backend architecture
    • webfetch tool for fetching and converting web content
  • Anthropic Pro/Max OAuth authentication support: Native integration with Claude Pro/Max accounts via OAuth flow
  • Rich tool call rendering: New rich_tool_names setting for enhanced tool call display
  • Configurable Features:
    • Make reasoning summary configurable via settings
    • Optional welcome message in VibecoreApp
    • Default model settings support

Changed

  • Default model updated to gpt-5
  • Improved Flow API with type-safe user input
  • Enhanced settings system with default_model_settings

Fixed

  • Tool call functionality with Claude Pro/Max authentication
  • Tool call results not displaying properly in the UI
  • Tool execution status handling during agent handoffs and response completion
  • Early exit handling in flow prototype
  • GitHub workflow conditions for stable PyPI releases
  • Version detection using github.ref_name

Documentation

  • Flow Mode documented as key differentiator
  • Added uvx as primary quick-test method

[0.2.0] - Previous Release

[Previous release notes...]