Skip to content

CHANGELOG.md

Latest commit

 

History

History
786 lines (643 loc) · 35.6 KB

CHANGELOG.md

File metadata and controls

786 lines (643 loc) · 35.6 KB

Changelog

All notable changes to Shello CLI will be documented in this file.

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

[0.7.1] - 2026-03-23

Fixed

  • Update command: replace running exe via rename-then-move on Windows to avoid [Errno 13] Permission denied on locked executables

[0.7.0] - 2026-03-23

Fixed

  • Update command: guard sys.argv[0] exe path lookup behind sys.frozen check

Changed

  • Tools: introduce tool registry and migrate to ShelloToolBase

[0.6.1] - 2026-03-21

Fixed

  • Windows Update Permission Error: Fixed /update failing with [Errno 13] Permission denied on Windows when installed to WindowsApps directory
    • Installer now redirects to ~/.shello_cli/shello.exe (already used for settings) instead of the protected stub path
    • Subsequent updates from that path work in-place with no permission issues
    • One-time PATH addition required: add %USERPROFILE%\.shello_cli to your user PATH
  • README Install Command: Updated Windows quick-install to use ~/.shello_cli/ instead of WindowsApps to prevent the issue for new installs

[0.6.0] - 2026-03-21

Added

  • Session History: Full session recording and resume system
    • Browse past conversations with interactive picker (/history)
    • Resume any previous session — conversation context fully restored
    • Delete individual sessions or clear all history (/history delete, /history clear)
    • Lazy recording: session files only created when user sends first message
    • Resumed sessions append to the original session file (no duplicates)
    • Background storage pruning with configurable size limit (session_history.max_storage_mb)
    • Session history can be disabled via session_history.enabled: false in settings
    • Bedrock tool-pair sanitization on restore to prevent ValidationException errors

Changed

  • Welcome Banner: Added /history command to the available commands list

0.5.2 - 2026-01-24

Fixed

  • Update Command in Frozen Executables: Fixed critical bug where /update command failed in PyInstaller-built executables with "Could not determine current version" error
    • Changed version detection from file-based reading to module import
    • Now uses import shello_cli; version = shello_cli.__version__ instead of reading __init__.py as a file
    • Works correctly in both development mode and frozen executables
    • Updated all related tests to match new implementation

Documentation

  • Community Guidelines: Added CODE_OF_CONDUCT.md for community standards and expectations

0.5.1 - 2026-01-22

Added

  • Self-Update System: Built-in update mechanism for keeping Shello CLI up-to-date

    • Version checker component with semantic version comparison
    • Platform detector for identifying OS and architecture
    • Executable updater with atomic file replacement and backup/rollback
    • Update manager orchestrating the complete update workflow
    • GitHub release integration for fetching latest versions
    • Automatic startup version check (configurable via settings)
    • /update command for manual update checks and installation
    • Progress indicators and user-friendly update flow

  • Update Configuration: New UpdateConfig in settings

    • check_on_startup - Enable/disable automatic version checks at startup (default: true)
    • auto_install - Automatically install updates without prompting (default: false)
    • Configurable via ~/.shello_cli/user-settings.yml

Changed

  • Agent Template: Enhanced system prompt with improved guidance

    • Added working directory awareness and cd command usage examples
    • Emphasized brevity and conciseness in responses
    • Better examples for avoiding verbose summaries

  • Help Display: Updated to include /update command in available commands list

  • Command Detection: Added /update to direct command detection for instant routing

Fixed

  • Documentation: Updated supported AI providers list to accurately reflect current capabilities

Dependencies

  • packaging>=23.0: Added for robust semantic version comparison

Testing

  • Update System Tests: Comprehensive test coverage for update functionality
    • Unit tests for version checker, platform detector, executable updater, and update manager
    • Integration tests for CLI update command and startup version check
    • Mock-based testing for GitHub API interactions
    • Edge case handling for network failures, permission errors, and invalid versions

Technical Details

  • Atomic Updates: Safe update process with backup and rollback on failure
  • Cross-Platform: Supports Windows (.exe), Linux, and macOS executables
  • GitHub Integration: Fetches releases from GitHub API with proper error handling
  • Semantic Versioning: Uses packaging library for accurate version comparison
  • Non-Blocking: Startup check doesn't delay application launch

0.4.3 - 2026-01-16

Added

  • Python 3.13 Support: Added official support for Python 3.13

Changed

  • Tool Naming: Renamed bash tool to run_shell_command for better clarity and consistency
  • Version Display: Enhanced version display with improved formatting
  • Command Detection: Enhanced input detection with comprehensive heuristics for better natural language vs command differentiation

Documentation

  • License: Updated license information
  • Security: Added SECURITY.md policy file
  • README: Enhanced readme with release and download badges

0.4.2 - 2026-01-14

Fixed

  • Stream Lifecycle Management: Handle missing messageStop events and improve stream lifecycle management
  • Message Content Display: Improve message content display ordering and formatting

Changed

  • Direct Command Execution: Remove trust evaluation from direct commands for faster execution
  • Welcome Message: Remove AI-powered CLI banner text from welcome message for cleaner UI
  • Live Display: Improve live display lifecycle management with comprehensive message processor tests

Added

  • Bash Tool Safety: Add explanation requirement to bash tool and simplify safety documentation
  • Tool Execution: Add is_safe parameter to bash tool execution for better safety control

0.4.1 - 2026-01-07

Fixed

  • Stream Lifecycle Management: Handle missing messageStop events and improve stream lifecycle management
  • Message Content Display: Improve message content display ordering and formatting

Changed

  • Direct Command Execution: Remove trust evaluation from direct commands for faster execution
  • Welcome Message: Remove AI-powered CLI banner text from welcome message for cleaner UI
  • Live Display: Improve live display lifecycle management with comprehensive message processor tests

Added

  • Bash Tool Safety: Add explanation requirement to bash tool and simplify safety documentation
  • Tool Execution: Add is_safe parameter to bash tool execution for better safety control

0.4.0 - 2026-01-06

Added

Settings Management System

  • Modular Settings Architecture: Separated internal patterns from user-changeable settings

    • patterns.py - Internal regex patterns and templates (NOT user-changeable)
    • defaults.py - Default values for user-changeable settings
    • settings/ module - Organized settings management with public API
    • Clean separation between code constants and user configuration

  • Settings Module Structure: New shello_cli/settings/ package

    • __init__.py - Public API exports (SettingsManager, get_settings, get_api_key, etc.)
    • models.py - Dataclasses for settings (ProviderConfig, OutputManagementConfig, CommandTrustConfig, UserSettings)
    • serializers.py - YAML generation with helpful comments and documentation
    • manager.py - SettingsManager class with load/save/merge logic

  • YAML Configuration Format: User-friendly YAML with inline documentation

    • Replaced JSON with YAML for better readability
    • Inline comments explaining each setting
    • Section headers for organization
    • Examples for all optional settings
    • Only configured values saved (rest use defaults)

  • Enhanced Setup Wizard: Generates well-documented settings file

    • Creates ~/.shello_cli/user-settings.yml with all options as comments
    • Only saves values user explicitly configured
    • Shows examples for optional settings (output_management, command_trust)
    • Includes helpful documentation in generated file

  • Configuration Management Commands: New CLI commands for settings

    • shello config - Display current configuration
    • shello config --edit - Open settings in default editor ($EDITOR)
    • shello config get <key> - Get specific setting value (supports dot notation)
    • shello config set <key> <value> - Set specific setting value (supports dot notation)
    • shello config reset - Reset settings to defaults with confirmation

  • Default Merging Strategy: Smart merging of user settings with defaults

    • User only specifies values they want to override
    • Missing values automatically filled from defaults.py
    • Denylist is always additive (user patterns added to defaults for safety)
    • Environment variables override file settings

  • Settings Validation: Graceful handling of invalid values

    • Validates provider values (openai, bedrock)
    • Validates approval_mode (user_driven, ai_driven)
    • Falls back to defaults on invalid values
    • Logs warnings for invalid configuration

  • File Security: Automatic secure file permissions

    • Sets 0o600 (user-only read/write) on settings files
    • Protects API keys and credentials
    • Creates parent directories if needed

Multi-Provider Support

  • AWS Bedrock Integration: Full support for AWS Bedrock as an AI provider

    • Anthropic Claude models (3.5 Sonnet, 3 Opus, 3 Sonnet)
    • Amazon Nova models (Pro, Lite, Micro)
    • Other Bedrock foundation models
    • Multiple AWS credential methods (profile, explicit credentials, default chain)
    • Environment variable support (AWS_REGION, AWS_PROFILE, AWS_ACCESS_KEY_ID, AWS_SECRET_ACCESS_KEY)

  • Provider Configuration System: Modern provider configuration using ProviderConfig dataclass

Fixed

Direct Command Execution

  • Natural Language Detection: Fixed issue where natural language questions starting with command words (e.g., "which model are you using") were incorrectly executed as shell commands

    • Added intelligent heuristics to distinguish between shell commands and natural language
    • Commands like "which python" still execute directly as shell commands
    • Questions like "which model are you using" now correctly route to AI
    • Handles question patterns: "are you", "do you", "can you", "what is", etc.
    • Special handling for ambiguous commands like "which" and "find"
    • Question marks at the end with natural language context route to AI
    • Validates Requirement 1.3: Natural language context routes to AI processing
    • Separate configurations for OpenAI-compatible APIs and AWS Bedrock
    • Support for multiple configured providers simultaneously
    • Provider-specific settings (API keys, base URLs, AWS credentials, regions)
    • Per-provider model lists and default models

  • Client Factory Pattern: Factory function for creating appropriate client based on provider

    • create_client() function in shello_cli/api/client_factory.py
    • Automatic client selection based on provider configuration
    • Graceful error handling with helpful troubleshooting messages
    • Support for boto3 import error detection

  • Runtime Provider Switching: Switch between providers during chat sessions

    • New /switch command in chat interface
    • Interactive provider selection menu
    • Conversation history preservation across provider switches
    • Automatic cache clearing and agent recreation
    • Display confirmation with new provider and model

  • Enhanced Setup Wizard: Interactive provider selection and configuration

    • Provider selection menu (OpenAI-compatible API vs AWS Bedrock)
    • OpenAI-compatible setup flow with API provider presets (OpenAI, OpenRouter, Custom)
    • AWS Bedrock setup flow with credential method selection
    • Support for configuring multiple providers
    • Model suggestions based on selected provider

  • Environment Variable Support: All credentials support environment variable overrides

    • OpenAI: OPENAI_API_KEY
    • Bedrock: AWS_REGION, AWS_PROFILE, AWS_ACCESS_KEY_ID, AWS_SECRET_ACCESS_KEY
    • Environment variables take precedence over configuration files

Command Trust and Safety System

  • Trust Manager: Evaluates commands before execution through 5-step flow (Denylist → YOLO → Allowlist → AI Safety → Approval)
  • Pattern Matcher: Supports exact match, wildcards (git *), and regex (^git (status|log)$)
  • Approval Dialog: Interactive UI for command approval with keyboard controls (A/D)
  • Default Allowlist: Safe commands execute automatically (ls, pwd, git status, cat, grep, etc.)
  • Default Denylist: Dangerous commands require approval (rm -rf /, dd, mkfs, format, etc.)
  • YOLO Mode: Bypass approval checks via config or --yolo flag (still respects denylist)
  • AI Safety Integration: Required is_safe parameter in bash tool for AI command evaluation
  • Configuration: Flexible trust settings in ~/.shello_cli/user-settings.json
    • approval_mode: "user_driven" (default) or "ai_driven"
    • allowlist: User patterns replace defaults
    • denylist: User patterns added to defaults (additive for safety)

Changed

Settings Management Changes

  • Settings File Format: Migrated from JSON to YAML

    • ~/.shello_cli/user-settings.json → ~/.shello_cli/user-settings.yml
    • Backward compatibility maintained (old JSON files still work)
    • Automatic migration on first run with new version

  • Constants Organization: Split constants.py into focused modules

    • patterns.py - Internal patterns (COMMAND_PATTERNS, CONTENT_PATTERNS, etc.)
    • defaults.py - User-changeable defaults (DEFAULT_CHAR_LIMITS, DEFAULT_STRATEGIES, etc.)
    • Updated all imports across codebase

  • Settings Manager Location: Moved to dedicated module

    • utils/settings_manager.py → settings/manager.py
    • Backward compatibility via re-exports from utils/settings_manager.py
    • No breaking changes for existing code

Multi-Provider Support Changes

  • Settings Manager: Extended with provider configuration support

    • New ProviderConfig dataclass for provider-specific settings
    • openai_config and bedrock_config fields in UserSettings
    • Helper methods: get_provider(), set_provider(), get_provider_config(), is_provider_configured(), get_available_providers()
    • Backward compatibility maintained for existing get_api_key() and get_base_url() methods

  • ShelloAgent: Refactored to use dependency injection

    • Constructor now accepts client parameter instead of creating its own
    • Removed api_key, base_url, model parameters from constructor
    • Supports both ShelloClient and ShelloBedrockClient via Union type
    • Cleaner separation of concerns

  • CLI Integration: Updated to use factory pattern and support provider switching

    • create_new_session() uses client factory
    • New switch_provider() function for runtime switching
    • Enhanced error handling with provider-specific messages
    • /switch command added to chat loop

  • UI Renderer: Added /switch command to help display

Command Trust and Safety Changes

  • Bash Tool: Added required is_safe parameter, trust evaluation before execution
  • Direct Executor: Integrated with trust system for direct commands
  • Settings Manager: New CommandTrustConfig with validation and safe defaults
  • Tool Definitions: Updated bash tool description with SAFETY section

Testing

Settings Management Tests

  • Comprehensive Test Coverage: 400+ tests for settings system

    • Settings loading and merging with defaults
    • YAML serialization with comments
    • Configuration validation and fallback
    • File permissions and security
    • Environment variable overrides
    • Dot notation for get/set commands

  • Property-Based Tests: Formal correctness validation

    • Property 1: Settings Round-Trip (serialize → deserialize produces equivalent settings)
    • Property 2: Default Merging (partial settings filled with defaults)
    • Property 3: Denylist Immutability (user denylist always includes defaults)
    • Property 4: Environment Variable Fallback (env vars used when config missing)
    • Property 5: Singleton Identity (all get_instance() calls return same object)

Multi-Provider Support Tests

  • Client Factory Tests: 542 lines of comprehensive tests

    • Test creating ShelloClient for OpenAI provider
    • Test creating ShelloBedrockClient for Bedrock provider
    • Test error handling for invalid providers and missing configuration
    • Test boto3 import error handling

  • Settings Manager Tests: Extended with provider configuration tests

    • Test loading settings with different providers
    • Test provider config resolution with environment variables
    • Test available providers detection

Command Trust and Safety Tests

  • 550+ tests covering pattern matching, trust evaluation, configuration, and integration
  • 8 property-based tests validating correctness properties

Documentation

Settings Management Documentation

  • README Updates: Configuration section rewritten

    • New "Configuration Management" section with CLI commands
    • YAML configuration examples with inline comments
    • Explanation of default merging strategy
    • Updated all JSON examples to YAML

  • DEVELOPMENT_SETUP.md Updates: Comprehensive developer documentation

    • New CLI commands section (config --edit, get, set, reset)
    • YAML configuration format with examples
    • Updated configuration hierarchy with defaults.py
    • Project structure showing new settings/ module
    • Updated troubleshooting for YAML validation

  • Design Documentation: Complete design document

    • Architecture overview with file responsibilities
    • Data flow diagrams (loading and saving)
    • Example generated user-settings.yml
    • Correctness properties with validation strategy
    • Error handling and testing strategy

Multi-Provider Support Documentation

  • README Updates: Comprehensive multi-provider documentation

    • New "AI Provider Support" section with provider overview
    • Runtime provider switching examples and workflow
    • Environment variable documentation for all providers
    • Configuration examples for OpenAI, Bedrock, and multiple providers
    • Updated commands section with /switch command

  • DEVELOPMENT_SETUP.md Updates: Developer-focused provider documentation

    • Provider-specific configuration examples
    • Multiple provider setup instructions
    • Environment variable configuration for all providers
    • Updated troubleshooting section with provider-specific errors
    • Updated project structure showing new files

  • requirements.txt: Added boto3 with optional dependency comment

    • Marked as optional dependency for AWS Bedrock support
    • Installation instructions for users who only need OpenAI-compatible APIs

Command Trust and Safety Documentation

  • Comprehensive README section on Command Trust and Safety
  • Enhanced inline documentation with examples
  • Complete design document with architecture and evaluation flow

Technical Details

Settings Management Technical Details

  • Modular Architecture: Clean separation of concerns

    • Settings models (dataclasses)
    • Serialization logic (YAML with comments)
    • Management logic (load/save/merge)
    • Public API (convenience functions)

  • Type Safety: Full type hints throughout

    • Dataclasses for all settings structures
    • Optional fields with proper defaults
    • Type-safe merging and validation

  • Backward Compatibility: Zero breaking changes

    • Old JSON files still work
    • Existing imports continue to work via re-exports
    • Automatic migration path

  • Security First: Secure by default

    • Automatic file permissions (0o600)
    • No credentials in error messages
    • Environment variable support for sensitive data

Multi-Provider Support Technical Details

  • Modular Provider System: Clean separation with client factory pattern
  • Type Safety: Union types for client interfaces, full type hints
  • Backward Compatibility: Existing OpenAI configurations continue to work
  • Graceful Degradation: Helpful error messages when boto3 is not installed
  • Zero Breaking Changes: Existing users can continue using OpenAI without changes

Security

  • Protection against dangerous commands with denylist
  • Audit logging of all trust decisions
  • Safe defaults on errors or invalid configuration

0.3.0 - 2026-01-04

Added

Direct Command Execution

  • Command Detection System: Automatically detects and routes direct shell commands vs AI queries

    • Recognizes common Unix commands (ls, pwd, cd, cat, grep, find, etc.)
    • Recognizes Windows commands (dir, cls, type, copy, del, etc.)
    • Routes recognized commands directly to shell without AI processing
    • Falls back to AI for natural language queries and complex requests

  • Direct Executor: Fast command execution without AI overhead

    • Executes commands directly in current shell environment
    • Maintains directory state across commands (cd persistence)
    • Automatic shell detection (bash, PowerShell, cmd)
    • Real-time output streaming
    • Integrated with output caching system
    • 30-second timeout for safety

  • Context Manager: Tracks direct command history for AI awareness

    • Records last 10 direct commands with output
    • Provides context to AI when switching from direct to AI mode
    • Includes cache IDs for AI to retrieve full output
    • Marks commands as sent to avoid duplicate context
    • Clears on /new command

  • Enhanced Prompt: Directory-aware prompt display

    • Shows current working directory in prompt
    • Abbreviates home directory as ~
    • Truncates long paths for readability (40 char max)
    • Color-coded path display (orange)
    • Updates dynamically as directory changes

  • Direct Command UI: Clean rendering for direct command execution

    • Terminal-style header without AI branding
    • Shows user@hostname and current directory
    • Command prompt with $ indicator
    • Distinguishes direct execution from AI-processed commands

Changed

  • Cache System Improvements:

    • Removed TTL expiration - cache persists for entire conversation
    • Increased cache size from 10MB to 100MB
    • Cache cleared only on /new command or app exit
    • Counter resets on new conversation for clean cache IDs
    • Updated all references to remove expiration mentions

  • PowerShell Output Handling:

    • Strips trailing whitespace from each line (removes column padding)
    • Reduces character count by 2-3x for PowerShell commands
    • Preserves output structure and readability
    • Applied consistently across bash_tool and direct_executor

  • Tool Result Metadata:

    • Added truncation metadata to tool results sent to AI
    • Includes cache_id, total_chars, shown_chars, total_lines, shown_lines
    • AI can see truncation status and make informed decisions
    • Enables smarter use of get_cached_output tool

  • API Debug Logging:

    • Added detailed HTTP request/response logging for OpenAI API
    • Logs full message content, tool calls, and parameters
    • Shows token usage and model information
    • Helps debug tool calling and prompt issues
    • Enabled via debug flag in ShelloClient

  • Agent Cache Management:

    • Added clear_cache() method to ShelloAgent
    • Automatically clears cache on /new command
    • Clears cache on Ctrl+C exit
    • Provides get_bash_tool() for shared caching with direct executor

Testing

  • Direct Command Execution Tests: 233 tests for direct executor

    • Command execution across different shells
    • Directory change handling
    • Error handling and timeouts
    • Output caching integration

  • Command Detection Tests: 201 tests for command detector

    • Direct command recognition
    • AI query detection
    • Edge cases and empty input

  • Context Manager Tests: 241 tests for context management

    • Command recording and history
    • AI context generation
    • History limits and clearing

  • Cache Sequential ID Tests: 193 tests for cache ID generation

    • Sequential ID generation
    • Counter reset on clear
    • LRU eviction behavior

  • CLI Integration Tests: 119 tests for end-to-end workflows

    • Direct command execution flow
    • AI query routing
    • Context switching

  • User Input Tests: 235 tests for prompt display

    • Directory abbreviation
    • Path truncation
    • Prompt formatting

  • Output Utils Tests: 106 tests for PowerShell padding removal

    • Line padding stripping
    • Structure preservation
    • Edge cases

Technical Details

  • Modular Command System: New commands/ submodule with clean separation
    • command_detector.py - Input classification
    • direct_executor.py - Direct command execution
    • context_manager.py - History and context tracking
  • Shared Caching: Direct executor and bash_tool share same cache instance
  • Zero Latency: Direct commands execute immediately without API calls
  • Seamless Integration: Transparent switching between direct and AI modes

0.2.0 - 2026-01-03

Added

Smart Output Management v2

  • Character-Based Truncation: Intelligent output limits based on character count (5K-20K) instead of arbitrary line counts

    • List commands: 5,000 chars (~1.2K tokens)
    • Search results: 10,000 chars (~2.5K tokens)
    • Log files: 15,000 chars (~3.7K tokens)
    • JSON output: 20,000 chars (~5K tokens)
    • Install/Build commands: 8,000 chars (~2K tokens)
    • Test commands: 15,000 chars (~3.7K tokens)
    • Safety limit: 50,000 chars (hard maximum)

  • Smart Truncation Strategies: Different strategies for different command types

    • FIRST_ONLY: List and search commands (show beginning where results start)
    • LAST_ONLY: Log commands (show end with most recent entries)
    • FIRST_LAST: Install/build/test commands (20% first + 80% last to see both setup and results)
    • Automatic JSON analysis for large JSON outputs

  • Output Caching System:

    • Sequential cache IDs (cmd_001, cmd_002, cmd_003...)
    • 5-minute TTL with LRU eviction
    • 10MB maximum cache size
    • Full output always cached for later retrieval

  • Get Cached Output Tool: New AI tool for retrieving specific sections from cached output

    • lines="+N" - First N lines
    • lines="-N" - Last N lines
    • lines="+N,-M" - First N + last M lines
    • lines="N-M" - Lines N through M (1-indexed)
    • Full output retrieval with safety limit

  • Semantic Truncation: Intelligent line importance classification

    • CRITICAL: Errors, failures, exceptions, fatal messages
    • HIGH: Warnings, success messages, summaries
    • MEDIUM: Test results, status indicators
    • Critical lines always visible regardless of position in output
    • Adjusts truncation budget to preserve important information

  • Progress Bar Compression: Automatic detection and compression of repetitive progress output

    • Detects percentage indicators, progress bars, spinner characters
    • Keeps only final state of each progress sequence
    • Saves hundreds of lines in install/build commands
    • Compression happens before character counting

  • Enhanced JSON Handling:

    • Automatic json_analyzer_tool invocation for JSON >20K chars
    • Returns jq paths showing structure instead of raw JSON
    • Raw JSON cached for retrieval via get_cached_output
    • Graceful fallback to text truncation if JSON parsing fails

  • Truncation Summary: Comprehensive summary appended to truncated output

    • Total vs shown characters and lines
    • Strategy used (FIRST_ONLY, LAST_ONLY, FIRST_LAST)
    • Optimizations applied (compression, semantic)
    • Cache ID with expiration time
    • Suggestion for get_cached_output with appropriate line range

Core Improvements

  • Type Detection System: Automatic detection of output type from command and content

    • Command pattern matching (npm install, pytest, docker logs, etc.)
    • Content pattern matching (JSON structure, test results, build output)
    • Content detection takes precedence over command detection

  • Streaming Output Enhancement:

    • User sees output streaming normally in real-time
    • AI receives truncated result with summary at end
    • Full output accumulated and cached in background
    • No interruption to user experience

  • Settings Manager Update: Enhanced configuration support for output management

    • Configurable character limits per output type
    • Configurable truncation strategies
    • Semantic analysis toggle
    • Progress bar compression toggle
    • Cache settings (TTL, max size)

Changed

  • Output Manager Architecture: Complete rewrite with modular design

    • Replaced monolithic output_manager.py with output/ submodule
    • Separate modules: cache.py, type_detector.py, truncator.py, semantic.py, compressor.py, manager.py
    • Clean separation of concerns and testability
    • Type-safe with comprehensive dataclasses

  • Bash Tool Integration: Updated to use new output management system

    • Shared OutputCache instance across tools
    • Automatic caching of all command output
    • Integration with truncation and summary generation

  • Tool Executor: Enhanced to share output cache across all tools

    • Single cache instance for entire session
    • Consistent cache IDs across tool invocations

  • System Prompt: Updated with concise output management guidance

    • Character limits and strategies
    • Cache retrieval examples
    • JSON handling behavior
    • Best practices for filtering at source

Removed

  • Deprecated Output Manager: Removed old line-based output_manager.py
  • Old Tests: Removed test_output_manager.py (replaced with comprehensive new test suite)

Testing

  • Comprehensive Test Suite: 105+ tests for output management
    • Unit tests for all components (cache, detector, truncator, semantic, compressor)
    • Property-based tests using Hypothesis library (100+ iterations per property)
    • Integration tests for end-to-end workflows
    • 8 formal correctness properties validated:
      1. Character Limit Enforcement
      2. Line Boundary Preservation
      3. Cache Round-Trip
      4. Semantic Critical Preservation
      5. Progress Bar Compression Idempotence
      6. Type Detection Consistency
      7. JSON Analyzer Fallback
      8. Summary Completeness

Technical Details

  • Modular Architecture: Clean separation with output/ submodule
  • Type Safety: Full type hints and dataclass models throughout
  • Property-Based Testing: Formal correctness guarantees via Hypothesis
  • Zero Data Loss: Full output always cached, retrieve any section on demand
  • Performance: LRU cache with efficient eviction, minimal overhead on streaming

Documentation

  • Updated README with unique features and technical highlights
  • Comprehensive design document (.kiro/specs/output-management/design.md)
  • Detailed requirements document with EARS patterns
  • Implementation tasks with property-based testing strategy

0.1.0 - 2026-01-02

Added

Core Features

  • AI-Powered Chat Interface: Interactive conversational AI assistant with streaming responses
  • Command Execution: Execute shell commands directly through AI with real-time output streaming
  • Multi-Platform Support: Native support for Windows (cmd, PowerShell, Git Bash), Linux, and macOS
  • OpenAI-Compatible API Integration: Works with OpenAI, OpenRouter, LM Studio, Ollama, and any OpenAI-compatible endpoint

Tools & Capabilities

  • Bash Tool: Execute shell commands with automatic shell detection (bash, PowerShell, cmd)
    • Real-time command output streaming
    • Working directory persistence across commands
    • Support for cd command with state management
    • Cross-platform command execution
  • JSON Analyzer Tool: Analyze JSON output structure from commands
    • Automatic jq path generation
    • Prevents terminal flooding from large JSON responses
    • Helps discover data structure before filtering
  • Output Management: Smart output truncation to prevent terminal flooding
    • Configurable output size limits
    • Automatic truncation warnings
    • Streaming output support

User Interface

  • Rich Terminal UI: Beautiful terminal interface powered by Rich library
    • Syntax highlighting for code blocks
    • Custom markdown rendering
    • Colored output and formatting
    • Welcome banner with system information
  • Interactive Commands:
    • /quit or /exit - Exit the application
    • /new - Start a new conversation
    • /help - Display help and shortcuts
    • /about - Display about information
  • CLI Commands:
    • shello or shello chat - Start chat session
    • shello config - Show current configuration
    • shello --version - Display version information

Configuration & Settings

  • Flexible Configuration System:
    • Global user settings: ~/.shello_cli/user-settings.json
    • Project-specific settings: .shello/settings.json
    • Environment variable support
  • Custom Instructions: Support for custom AI instructions via .shello/SHELLO.md
  • Multi-Model Support: Configure and switch between different AI models
  • API Configuration: Customizable API endpoints and authentication

Session Management

  • Conversation History: Persistent chat history within sessions
  • Context Awareness: AI maintains context across multiple interactions
  • Tool Call Tracking: Complete tracking of tool executions and results
  • System Information Integration: Automatic detection of OS, shell, working directory, and timestamp

Developer Features

  • Modular Architecture:
    • Separate modules for agent, API, chat, commands, tools, UI, and utilities
    • Clean separation of concerns
    • Extensible tool system
  • Comprehensive Testing: Test suite with pytest
  • Build System: Automated build scripts for creating standalone executables
    • PyInstaller integration
    • Platform-specific build scripts (build.bat, build.sh)
    • Automated GitHub Actions workflow for releases

Distribution

  • Standalone Executables: Pre-built binaries for Windows, Linux, and macOS
  • Easy Installation: Multiple installation methods
    • Download pre-built executables
    • Install from source
    • Add to system PATH
  • GitHub Releases: Automated release workflow with CI/CD

Documentation

  • Comprehensive README with quick start guide
  • Detailed setup instructions (SETUP.md)
  • Build instructions (BUILD_INSTRUCTIONS.md)
  • Contributing guidelines (CONTRIBUTING.md)
  • Multiple technical documentation files in doc/ directory

Technical Details

  • Python 3.11+ required
  • Dependencies: Click (CLI), Rich (UI), OpenAI client library
  • Shell Detection: Automatic detection of bash, PowerShell, cmd, and Git Bash
  • Streaming Architecture: Real-time streaming for both AI responses and command output
  • Error Handling: Comprehensive error handling with user-friendly messages
  • Cross-Platform Paths: Proper path handling for Windows and Unix-like systems

Note

This is the initial release of Shello CLI. While fully functional, expect improvements and potential breaking changes before v1.0.0. Feedback and contributions are welcome!