Skip to content

Latest commit

 

History

History
1057 lines (824 loc) · 38.3 KB

File metadata and controls

1057 lines (824 loc) · 38.3 KB

CoStrict Change Log Archive

This file contains historical version records. For the latest versions, please visit CHANGELOG.md

[2.4.4]

  • Add support for task history backup and restore, supporting export as .tar.gz backup files (PR #1049)
  • Fix: prevent mode reset when temporarily hidden by zgsm filter (PR #1048)

[2.4.3]

  • Implement streaming status tracking and improve task management
  • Optimize prompt caching by extracting static sections
  • Disable zgsmCodebaseIndexEnabled by default
  • Fix known issues

[2.4.2]

  • Optimize commit message generation
  • Update docs

[2.4.1]

  • Optimize streaming status tracking by moving state management from client to backend
  • Add OpenAI native provider support and comprehensive type definitions
  • Refactor settings UI organization (integrate Auto Cleanup into CheckpointSettings)
  • Sync roocode last commit
  • Fix known issues

[2.4.0]

  • Add new Plan Agent integration (new agents added: quick-explore, task-check, subcoding)
  • Agent Prompts i18n Support
  • Prevent parent task state loss during orchestrator delegation
  • Fix race condition in new_task delegation that loses parent task history
  • Serialize taskHistory writes and fix delegation state race condition
  • Optimize compression summary during task recovery
  • Optimize command-line output performance \ diagnostic cache \ UI interaction
  • Remove built-in deprecated browser integration features
  • Add batch tool calling
  • Support nested subtask tree navigation
  • Add file change summary logging
  • Fix AskFollowupQuestionTool parameter validation
  • Package size optimization
  • Documentation image updates
  • Update Gemini cli models
  • Render nested subtasks as recursive tree
  • Handling test case errors
  • Sync roocode last commit
  • Fix known issues

[2.3.10]

  • Update docs
  • Update openspec-init slash command
  • Optimize tools msg render
  • Fix known issues

[2.3.9]

  • Disable experimental features by default and adjust thresholds
  • Optimize the detection of dangerous commands
  • Optimize the selection of multi window models
  • Update CodeReview request timeout configuration (300s -> 600s)
  • Fix cancel backend auto-approval timeout when auto-approve is toggled off mid-countdown

[2.3.8]

  • UI experience improvements
  • Fix prevent parent task state loss during orchestrator delegation
  • Sync roocode last commit

[2.3.7]

  • UI experience improvements
  • Sync roocode last commit
  • Fix known issues

[2.3.6]

  • Add button to open annual summary with icon
  • Translate Chinese prompts to English

[2.3.5]

  • Fix empty-string baseURL in provider
  • Add support for .agents/skills directory
  • Sync roocode last commit
  • Fix known issues

[2.3.4]

  • Refactored Gemini provider with improved error handling and streaming
  • Enhanced Vertex provider with comprehensive model support
  • Support Qwen3-Coder-Next model for Zgsm provider
  • Optimize the behavior of the "resumeTask"
  • add fake reasoning chunk handling for response processing
  • Refactored custom instructions system for better mode customization
  • Update internationalization
  • Sync roocode last commit
  • Fix known issues

[2.3.3]

  • Enhanced gemini-cli prompt
  • Enhanced getClaudeCodeModels function to support multiple model IDs separated (ANTHROPIC_MODEL)
  • Simplified task command parameters: CancelTask and CloseTask commands no longer require data parameter
  • Improve Skills and Slash Commands settings
  • Fix worktree detection for Git repositories (Windows)
  • Refactored lite tool prompt system with improved text format and structure
  • Added QueuedMessagesUpdated event for better message queue tracking
  • Fixed type definitions in IPC.ts by removing incorrect type exports
  • Removed taskMap from API to simplify task management
  • Added @ai-sdk/mistral dependency to src/package.json
  • Cleaned up unused imports and variables in Task.ts, ClineProvider, and webviewMessageHandler
  • Improved error clipboard copy format with bracket labels
  • Updated extension tests to correctly handle .env file detection logic
  • Sync roocode last commit
  • Fix known issues

[2.3.2]

  • Improve string parsing logic and add parameter safety

[2.3.1]

  • Optimize message text handling in ChatRow

[2.3.0]

  • Enhanced JetBrains platform support with updated environment configuration
  • Optimized IPC connection handling and reduced content limits for better performance
  • Centralized excluded file extensions management for improved maintainability
  • Added comprehensive mock configurations for vscode and child_process testing
  • Migrated multiple AI providers to AI SDK (DeepSeek, Cerebras, Groq, Fireworks)
  • Added OpenAI Codex provider with OAuth subscription authentication
  • Refactored read_file tool with Codex-inspired improvements
  • Added Skills system with built-in skills support (replacing fetch_instructions)
  • Enhanced Kimi 2.5 reasoning detection and empty reasoning handling
  • Improved terminal output buffer management to prevent memory leaks
  • Fixed Gemini thought signature validation errors
  • Fixed tool result ID sanitization to prevent API 400 errors
  • Added support for image content in MCP tool responses
  • Added settings tab titles to search index
  • Enhanced parallel tool calls support (enabled by default)
  • Added React Compiler integration to webview-ui
  • Fixed empty text blocks filtering for Gemini compatibility
  • Improved MCP tool schema handling with anyOf/oneOf/allOf flattening
  • Updated test cases and improved code coverage
  • Sync roocode last commit
  • Fix known issues

[2.2.5]

  • Support Kimi 2.5
  • Add Skills Management UI
  • Disable send button when no input and not streaming for better UX
  • Fix error when switching between reasoning and non-reasoning models in conversation
  • Sync roocode last commit
  • Fix known issues

[2.2.4]

  • Added Smart Code Folding
  • Enabled Smart Mistake Detection by default with auto-switch model enabled
  • Enhanced Zgsm streaming response parsing with improved format handling
  • Fixed file scanner to use relative paths for ignore directory checks
  • Removed deprecated POWER_STEERING experimental feature
  • Updated test cases and improved code coverage
  • Sync roocode last commit
  • Fix known issues

[2.2.3]

  • Compatible with zgsm function call parameter parsing
  • Update internationalization
  • Sync roocode last commit

[2.2.2]

  • Updated Worktrees UI
  • Updated suggestedBranch
  • Updated MCP configuration
  • Fix missing translations
  • Fix test cases
  • Sync roocode last commit

[2.2.1]

  • Added custom condense prompt editing in Context Management settings.
  • Re-architected intelligent context condensation algorithm (V2).
  • Introduced zgsmCodeMode filtering for mode selection.
  • Updated Worktrees description for generality.
  • Optimized subtask display in chat history with count and expand/collapse.
  • Improved Token progress display text and labels.
  • Removed experimental "Power Steering" mode.
  • Removed experimental "Multi-File Apply Diff" feature.
  • Updated English, Simplified Chinese, and Traditional Chinese localization texts.
  • Fix gemini-cli response parsing.
  • Sync roocode last commit

[2.2.0]

  • Added worktree functionality, supporting switching between multiple branches
  • Added review history feature, supporting viewing and managing historical review records
  • Added concise prompt mode to reduce token consumption
  • Improved intelligent error detection and automatic model switching logic
  • Improved task mode processing and mode selection logic
  • Improved IPC authentication error handling and retry logic
  • Removed all XML protocol code
  • Provided lightweight XML compatibility for Costrict Provider
  • UI experience improvements
  • Fixed known issues
  • Sync roocode last commit

[2.1.19]

  • Added GBK encoding support for Windows system commands, including tasklist, ping, netstat, etc.
  • Added OpenAI Codex rate limit monitoring and usage dashboard
  • Optimized task history update mechanism, supporting incremental updates to reduce communication overhead
  • Fixed issue where context compression state gets stuck when task is cancelled
  • Improved error messages when read_file tool is used on directories
  • Added MCP tool count statistics and warning UI components (WarningRow, TooManyToolsWarning)
  • Fixed known issues
  • Sync roocode last commit

[2.1.18]

  • Fix markdown collapse flickers
  • Sync roocode last commit

[2.1.17]

  • Add smart mistake detection and automatic model switching capability
  • Add auto_switch_model type support for quota display improvements
  • Add support for global Costrict directory in skills system
  • Enable markdown collapse by default for better readability
  • Add setting to toggle long markdown collapse without scroll
  • Add current task display with multilingual support (English, Simplified Chinese, Traditional Chinese)
  • Refactor chat display to show last user feedback instead of current task
  • Replace hardcoded zgsm base URL with COSTRICT_BASE_URL environment variable
  • Improve error handling and display in chat components
  • Sync roocode last commit
  • Fix known issues

[2.1.16]

  • Fix parallel_tool_calls parameter handling for LiteLLM/Bedrock compatibility
  • Add Stop button with queue message functionality for task management
  • Improve RandomLoadingMessage with static display support and 28 new emoji-based loading phrases
  • Optimize OpenAI format transformation with enhanced test coverage
  • Optimize ChatRow/ChatTextArea/ChatView components for better user experience
  • Add tool ID validation utilities
  • Sync roocode last commit
  • Fix known issues

[2.1.15]

  • Fix zgsm provider model flush to include baseUrl from provider configuration
  • Performance optimization: Use shallow copy instead of deep clone in message processing (80-90% reduction in cloning overhead)
  • Add settings search functionality with fuzzy matching and highlight animation
  • Add RandomLoadingMessage component for varied loading messages
  • Refactor settings view with improved tab management and search integration
  • Add multilingual support for reasoning status messages (English, Simplified Chinese, Traditional Chinese)
  • Improve ChatRow component rendering and state management
  • Enhance ReasoningBlock with better message formatting
  • Sync roocode last commit
  • Fix known issues

[2.1.14]

  • Optimize JetBrains response rendering configuration for better streaming performance
  • Adjust JetBrains platform default rendering mode
  • Update debug state management, add defaultDebug state tracking
  • Sync roocode last commit
  • Fix known issues

[2.1.13]

  • Fix Claude Code OAuth authentication compatibility
  • Add disk usage display to history page

[2.1.12]

  • Add path stabilization handling for file editing tools to prevent truncated paths during streaming
  • Optimize model selection logic in ModelPicker and ProviderRenderer components
  • Refactor terminal output tracking with improved count logic
  • Fix toolname parsing to handle special characters (tool_call tags)
  • Preserve DeepSeek reasoning_content during tool call sequences
  • Sync roocode last commit
  • Fix known issues

[2.1.11]

  • Add image file @mentions support - Reference images directly in chat using @ mentions
  • Add evals-context skill for evaluation system infrastructure
  • Add filtering to file search results
  • Add chat search feature to UI settings
  • Update environment variables: ZGMSM** prefixes migrated to COSTRICT** (with backward compatibility)
  • Update auto-approve mode label from "YOLO" to "BRRR"
  • Remove deprecated Bedrock Claude models (2.1, 2.0, instant)
  • Improve follow-up suggestion UI with better edit button positioning
  • Sync roocode last commit
  • Fix known issues

[2.1.10]

  • Improve terminal process error handling and abort operation
  • Optimize chat UI performance and message handling
  • Extend ZgsmAi message format to support more models (glm, claude, minimax)
  • Refactor code review module, simplify issue structure and remove repo review command
  • Add tool protocol auto rollback
  • Add tool alias support for model-specific tool customization
  • Add native tool calling support for multiple providers (Anthropic, Z.ai, OpenAI compatible, Vertex AI, etc.)
  • Add downloadable error diagnostics feature
  • Improve graceful retry for "no tools used" and "no assistant messages" errors
  • Sanitize MCP server/tool names for API compatibility
  • Add CRLF line ending normalization to search_replace and search_and_replace tools
  • Fix default info model id for zgsmAiCustomModel
  • Add debug mode toggle to control zgsmAiCustomModelInfo usage
  • Sync roocode last commit
  • Fix known issues

[2.1.9]

  • Fix update documentation links to use direct URLs
  • Add auto cleanup service
  • Enhance stream handling for MiniMax models with flexible tag matching
  • Enhance ZgsmAi provider for native protocol support
  • Add tool protocol auto roolback
  • Support AI assistant chat message deletion
  • Sync roocode last commit
  • Fix known issues

[2.1.8]

  • Ensure default model info is merged before custom overrides
  • Sync roocode last commit

[2.1.7]

  • Add concurrent file reads limit to read_file tool (default max 5 files)
  • Remove extended thinking signature capture from Anthropic provider
  • Sync roocode last commit
  • Fix known issues

[2.1.6]

  • Improve code review types and performance
  • Add optional mode field to slash command front matter
  • Capture extended thinking signatures for tool use continuations
  • Add support for npm packages and .env files to custom tools
  • Sync roocode last commit
  • Fix known issues

[2.1.5]

  • Add skills support
  • Add detection for Windows CMD special escape characters in dangerous substitutions
  • Remove deprecated simple read file tool implementation
  • Update environment variable prefixes from PKG** to COSTRICT_PKG** (#844)
  • Sync roocode last commit
  • Fix known issues

[2.1.4]

  • Fix Suppress 'ask promise was ignored' error in handleError (#9914)
  • Optimize response render configuration, adjust streaming limit parameters to improve performance
  • Enhance ReadFileTool file reading functionality
  • Add tool protocol display feature, showing current tool protocol (Native/XML) in task header
  • Add showSpeedInfo setting option to control speed info display in UI settings (CoStrict provider only)
  • Update resolveToolProtocol function to simplify decision tree
  • Update test cases to reflect new tool protocol behavior
  • Update protocol resolution test logic
  • Add multilingual support for tool protocol, performance metrics, and speed info display (English, Simplified Chinese, Traditional Chinese)
  • Fix known issues

[2.1.3]

  • Update error detail copy button with visual indicator (red dot) and improved instructions
  • Enhance error messages for copying error details in English, Chinese (Simplified), and Chinese (Traditional)
  • Fix code review race condition in completion handling by adding async delay to cleanup operations
  • Add partial message check and completion flag to prevent duplicate completion handling in code review
  • Add ask_multiple_choice native tool support for structured user input collection
  • Update gemini-cli model configurations with tool protocol support
  • Update tool protocol resolution logic to prioritize user preferences (XML protocol now takes precedence over native tools)
  • Refactor resolveToolProtocol function with simplified decision tree
  • Update test cases to reflect new tool protocol behavior
  • Add ask_multiple_choice native tool support
  • Sync roocode last commit
  • Fix known issues

[2.1.2]

  • Enhanced code review mode with improved UI behavior and provider support

[2.1.1]

  • Fix model response handling for "no tools used" scenarios
  • Add comprehensive error handling utilities for different API providers
  • Update UI components for better error display and user experience
  • Fix telemetry client to send events through vscode.postMessage instead of posthog-js
  • Update system prompts to use "Native fallback" instead of "XML fallback" as default
  • Add internationalization updates for error messages in multiple languages
  • Fix ModeSwitch component display issues
  • Sync roocode last commit
  • Fix known issues

[2.1.0]

  • Refactor: Convert claude-code models to dynamic loading with getClaudeCodeModels()
  • Fix: Race condition in new_task tool for native protocol
  • Add model-specific tool customization via excludedTools and includedTools
  • Add customTools for opt-in only tools
  • Add search_and_replace tool for batch text replacements
  • Enable native tool support for DeepSeek, Doubao, Requesty, and multiple other providers
  • Add native tool support to OpenAI-compatible providers, Vertex Gemini, Grok, Bedrock, and more
  • Refactor: Terminal simplification (Vscode/JetBrains)
  • Update i18n: Improve skipHint text clarity in multiple languages
  • Refactor: Update scope prefix from roo_cline to costrict
  • Update terminal test mocks and shell path handling
  • Add ModeSwitch component to manage display modes in chat
  • Optimize directory scanning by caching gitignore checks
  • Enhance error messages and documentation links
  • Remove line_count parameter from write_to_file tool
  • Handle malformed native tool calls to prevent hanging
  • Fix Vercel AI Gateway model fetching
  • Add search_replace native tool for single-replacement operations
  • Improve auto-approve timer visibility in follow-up suggestions
  • Cancel auto-approval timeout when user starts typing
  • Sync roocode last commit
  • Fix known issues

[2.0.27]

  • Add tool result ID validation and fix mechanism to prevent ID mismatch issues in API requests
  • Enhance maxTokens handling in ZgsmAiHandler and SettingsView components
  • Improve code review event handling to prevent duplicate completions
  • Optimize code review mode handling using reset functions for task completion and error states
  • Enhance task error handling by including request IDs in streaming failure messages
  • Add /dotest shortcut command and update mode configuration
  • Reduce maximum non-busy terminal count from 5 to 3
  • Fix disable handling logic in reasoning budget component
  • Unify webview panel identifiers using consistent tabPanelId
  • Add minimum and medium reasoning effort levels for Gemini
  • Filter out non-Anthropic content blocks sent to Vertex API
  • Fix test cases in terminal registry to adapt to new terminal limits
  • Add rehype-sanitize and markdown error handling
  • Sync roocode last commit
  • Fix known issues

[2.0.26]

  • Add multiple-choice question tool for structured user input collection
  • Enhance CoStrict provider with parallel tool calls support
  • Add MCP support to workflow modes (requirements, design, tasks, testing, code)
  • Improve terminal running state detection with platform-specific delays
  • Update architect mode instructions to use /plans directory
  • Add plans directory to .gitignore
  • Fix Gemini transformer to gracefully skip unsupported content blocks
  • Update system prompts and localization for multiple-choice feature
  • Update apply_diff BUFFER_LINES
  • Sync roocode last commit

[2.0.25]

[2.0.24]

  • Enhance model cache flushing mechanism to support optional configuration parameters
  • Add "Plan" development mode
  • Fix model cache race condition by preserving memory cache during refresh operations
  • Remove deprecated insert_content tool logic and prompts
  • Optimize apply-related tool validation to reduce errors from outdated monitoring rules
  • Optimize bash.exe command rules for double-slash parameter cases
  • Optimize tool invocation prompts for more flexible model tool usage
  • Optimize context compression summary to prevent key information truncation and context corruption loops
  • Optimize file editing behavior: treat identical AI edits as normal rather than errors to proceed with next steps
  • Optimize inline terminal execution/termination logic for conversation stability
  • Fix JetBrains shell integration compatibility (sync default shell changes from IDEA to plugin to avoid mismatched shell commands)
  • Add conversation button to editor opentab top for user-friendly guidance
  • Update gemini-cli gemini-3-pro-preview model
  • Optimize JetBrains shell integration
  • Improve JetBrains workflow compatibility
  • Optimize statistics performance
  • Sync roocode last commit
  • Fix known issues

[2.0.23]

  • Fix welcome page login url

[2.0.22]

  • Add JetBrains platform detection and disable unsupported VS Code features (code completion, CodeLens)
  • Add editor type detection for API and Webview message handler
  • Add mode preservation in code review service
  • Add parameter support and proportional limits to file content reading
  • Rebrand diff view scheme to costrict and improve wechat image styling
  • Update rate limit default and optimize streaming-related components
  • Adjust content limits and simplify prompts
  • Sync roocode last commit
  • Fix known issues

[2.0.21]

  • Remove apiRequestBlockHide setting and related UI components

[2.0.20]

  • Update files prompt
  • Update getSafeOperatingSystemName call in safe fallback for OS detection
  • Sync roocode last commit
  • Fix known issues

[2.0.19]

  • Remove run_test action and simplify run_task action prompt

[2.0.18]

  • Add vibeplus built-in commands
  • Sync roocode last commit
  • Fix known issues

[2.0.17]

  • Fix WriteToFileTool.handlePartial
  • Update About page
  • Fix finishSubTask no taskId when canceling or deleting tasks
  • Fix user_feedback when readfile
  • Sync roocode last commit
  • Fix known issues

[2.0.16]

  • Added bidirectional navigation between parent and child tasks.
  • Adapted to Browser Control 2.0.
  • Added built-in openspec-init initialization command.
  • Hid hint-type warnings to reduce noise.
  • Strictly compatible with all attempt_completion formats to reduce unnecessary requests.
  • Added compatibility for browser_action tool parameters, enabling small models to use the built-in browser tool.
  • Fixed freeze issue when clicking Force Continue during command execution.
  • Enabled chatbox input to write directly into the command line (note: inline terminal does not support sudo).
  • Fixed deleted files appearing in context.
  • Moved the MCP icon out of the collapse menu.
  • Upgraded upstream repository dependencies to resolve file-encoding security issues.
  • Improved JetBrains platform compatibility.
  • Optimized duplicate shell terminal creation (limit up to 5 terminals).
  • Added native tool support for CoStrict provider.
  • Sync roocode last commit
  • Fix known issues

[2.0.15]

  • Add user ID tracking to API requests
  • Restore previous mode after review completion
  • Add signal parameter support to provider completePrompt methods and improve request cancellation
  • Enhance condition checks and error handling
  • Enhance model info schema and update model fetching logic
  • Update jetBrains code review config
  • Update Gemini tool config
  • Optimize modes defaultSelect
  • Optimize build size
  • Sync roocode last commit
  • Fix codereview (#638)

[2.0.14]

  • Add api provider filtering for modes and workflow features
  • Add commit review
  • Add abort signal handling for streaming responses
  • Optimize model display information calculation
  • Optimize codereview
  • Optimize UI identification for all models
  • Optimize performance
  • Update mcp dependencies
  • Refactor follow-up auto-approval
  • Remove unused TelemetryEventName import
  • Update disclaimer section
  • Sync roocode last commit
  • Fix known issues

[2.0.13]

  • Add additional notes on Windows taskkill argument behavior in bash
  • Add API provider validation for code review features
  • Optimize settings for batch saving
  • Update branding and improve debounce timing
  • Add notification service with banner
  • Update UI style
  • Updated internationalization
  • Optimized performance
  • Optimize PDF file processing
  • Update default modelID to Auto
  • Fix mcp timeout setting
  • Fixed zgsmAiCustomModelInfo being overwritten
  • Fix openAiHeaders being overwritten
  • Sync roocode last commit
  • Fix known issues

[2.0.12]

  • Fix codereview (#638)

[2.0.11]

  • Update docs
  • Update error message
  • Updated internationalization
  • Optimize commands parsing
  • Add purchase quota
  • Fix message queue status inconsistency
  • Sync roocode last commit
  • Fix known issues

[2.0.10]

  • Update todolist ui
  • Enhance attempt_completion parsing
  • Optimize message retrieval
  • Update error message
  • Optimized tools prompts
  • Sync roocode last commit
  • Fix known issues

[2.0.9]

  • Update userinfo
  • Optimize tools prompts
  • Add 413 error handle
  • Update mcp handle
  • Optimize auto completion
  • Optimize log output
  • Sync roocode last commit
  • Fix known issues

[2.0.8]

  • Add mode parameter to ZgsmAiHandler
  • Optimized codereview error handle
  • Update wiki to v2
  • Update file detail
  • Optimized tools prompts
  • Optimized shell config
  • Fix eliminate UI flicker during task cancellation
  • Update internationalization of agents
  • Sync roocode last commit
  • Fix known issues

[2.0.7]

  • Delete legacy api and add timeout handling
  • Add auto discount
  • Sync roocode last commit
  • Fix known issues

[2.0.6]

  • Optimized auto completion
  • simplify custom instructions
  • Updated zgsm provider default temperature
  • Optimized logging
  • Sync roocode last commit
  • Fix known issues

[2.0.5]

  • Optimized request sending speed
  • Reduced file read/write handle usage
  • Optimized shell/non-shell integration compatibility and command execution efficiency
  • Updated brand name to CoStrict
  • Compatible with jetbrains codereview
  • Added review mode request headers
  • Optimized static file size
  • Optimized conversation memory usage
  • Fixed background silent editing Chinese encoding issues
  • Optimized shell integration prompts
  • Optimized readfile prompts
  • Optimized tool logs
  • Sync roocode last commit
  • Fix known issues

[2.0.4]

  • Add support for project spec commands from .cospec directory (#583)
  • Sync roocode last commit
  • Fix known issues

[2.0.3]

  • Add mode toggle with vibe/strict options in ModeSelector (#564)
  • Fix: update isInvalidId function to include additional invalid machine ID (#567)
  • Add error tracking with raw error storage enhancement (#569)
  • Add client ID header to API requests for completion (#570)
  • Refactor code-review to simplify startReview method (#572)
  • Add cursor positioning to first change after file save (#573)
  • Update default limits and delays, refactor follow-up countdown handling (#575)
  • Fix K2 model output truncation
  • Optimized CodeReview
  • Updated workflow prompts
  • Optimized api request performance and memory usage
  • Sync roocode last commit
  • Fix known issues

[2.0.2]

  • Optimized CodeReview
  • Sync roocode last commit
  • Fix known issues

[2.0.1]

  • Optimized performance
  • Optimized CodeReview
  • Add TestGuide agent
  • Refactored project wiki command
  • Optimized project wiki prompts
  • Added codebase fix functionality
  • Updated workflow prompts
  • Updated internationalization
  • Updated default model ID(GLM-4.5)
  • Increase context window size for ZGSM provider
  • Fix known issues

[2.0.0]

  • Add workflow with requirements, design, tasks, and testing support
  • Add test guide functionality for project testing solutions
  • Add project wiki generation
  • Add development modes: Vibe and Strict
  • Improve type definitions and global settings
  • Add AI-assisted review suggestion
  • Add commit model setting
  • Optimization CoStrict provider api performance
  • Fix known issues

[1.6.19]

  • Optimized gemini-cli provider
  • Optimized shell prompts
  • Sync roocode last commit
  • Fix known issues

[1.6.18]

  • Increase context window size for ZGSM provider
  • Add user ID display and improve cloud account handling
  • Enhance error handling with errorCode field in global settings
  • Improve stream cancellation and request handling
  • Enhance commit message generation with i18n support
  • Optimize rate limit retry message handling in UI
  • Sync roocode last commit
  • Fix known issues

[1.6.17]

  • Enhance rate limit detection and retry logic in API requests
  • Add getAllInstance() method to ClineProvider for better instance management
  • Update UI to properly handle rate limit retry messages
  • Fix known issues

[1.6.16]

  • Add auto-generation of commit messages based on local i18n language
  • Optimize request cancellation edge case handling
  • Enhance provider API completePrompt request, support system prompts and language metadata
  • Update internationalization
  • Change default request interval from 0s to 1s to reduce concurrency
  • Optimized CodeReview
  • Sync roocode last commit
  • Fix known issues

[1.6.15]

  • Increase max read character limit to 30,000
  • Add user quota information
  • Sync roocode last commit
  • Fix known issues

[1.6.14]

  • Optimize the 'line jump' for batch file reading
  • Optimize file read character limit
  • Sync roocode last commit
  • Fix known issues

[1.6.13]

  • New Chat UI (more readable and user-friendly)
  • Updates base URL retrieval in ErrorCodeManager
  • Refines localization strings for clarity
  • Adds os-locale package for improved language detection
  • Add file read character limit
  • Sync roocode
  • Fix known issues

[1.6.12]

  • Improve default values for zgsm configuration
  • Sync roocode
  • Fix known issues

[1.6.11]

  • Adds models refresh for CoStrict provider
  • Add line navigation for file jumps in chat interface
  • Implement chat cancellation and improve search functionality
  • Enhance metrics loading and pushing logic in PrometheusTelemetryClient
  • Update conditional class for provider renderer visibility
  • Enhance chat UI
  • Sync roocode
  • Update ci
  • Fix known issues

[1.6.10]

  • Improves chat search
  • Improves encoding handling for Windows terminal commands
  • Add zgsmAiCustomModelInfo config
  • Enhance model handling and configuration
  • Improve rate limiting
  • Sync roocode
  • Update ci

[1.6.9]

  • Fix binary file check

[1.6.8]

  • Add chat search
  • Enhance binary file detection and encoding handling
  • Enhanced models cache
  • Sync roocode
  • Fixed test cases
  • Fix known issues

[1.6.7]

  • Implement MetricsSerializer for metric persistence and locking
  • Optimized shell integration
  • Optimized models cache
  • Optimized message edit
  • Update docs
  • Sync roocode
  • Fix known issues

[1.6.6]

  • Fixed non-UTF8 encoding file issues
  • Improved shell integration command logic
  • Enhanced hidden directory detection as workspace root
  • Disabled codebase in non-workspace
  • Added history conversation editing mode
  • Performance optimization: cached device features
  • Updated i18n translations
  • Fixed test cases
  • Improved dialog "@" and "/" hide logic with ESC exit
  • Sync roocode

[1.6.5]

  • Optimize file watching and ignore patterns
  • Updates project references and contact information
  • Optimize error handling and localization
  • Sync roocode

[1.6.4]

  • Fix wsl2 your 131072x1 screen size is bogus. expect trouble

[1.6.3]

  • Optimize history cache warning
  • Default enable auto approve settings display in chat
  • Optimize workspace event handling and request management
  • Enhance process detection and health check
  • Enhance codebase index handling and confirmation dialog
  • Improve file monitoring performance
  • Update internationalization entries
  • Sync roocode
  • Fix known issues

[1.6.2]

  • Optimize login tips
  • Update mode icons
  • Update error message
  • Update model picker
  • Fix known issues

[1.6.1]

  • Add message rendering speed mode
  • Fix: Implements dropdown close behavior on outside click
  • Sync roocode
  • Fix known issues

[1.6.0]

  • Auto mode for Agent
  • Auto codebase search
  • Denied commands support
  • New codebase
  • New tools support (todolist, simpleReadFileTool etc.)
  • New telemetry report
  • Optimized tasks disk usage
  • Optimized CodeReview
  • Optimized UI
  • Sync roocode
  • Fix known issues

[1.5.12]

  • Update default modelID
  • Fix CodeReview path causing service exception
  • Add new commands for context
  • Fix known issues

[1.5.11]

  • Update docs
  • Update error message

[1.5.10]

  • Remove deprecated api
  • Fix known issues

[1.5.9]

  • Add Telemetry
  • Add AutoCommit
  • Optimized shell integration
  • Update models config (Grok and o4)
  • Fix known issues

[1.5.8]

  • Optimization context window
  • Update docs
  • Fix known issues

[1.5.7]

  • Project renamed to costrict
  • Optimized CodeReview
  • Code optimization

[1.5.6]

  • Update Error Code

[1.5.5]

  • Optimized Prompts
  • Add Model Permission Control
  • Fix known issues

[1.5.4]

  • Optimized CodeReview
  • Fix PowerShell command formatting
  • Update Error Code

[1.5.3]

  • Add Gemini CLI ProjectId Config
  • Fix known issues

[1.5.2]

  • Optimized CodeReview
  • Update Error Code

[1.5.1]

  • Support "Add To Context" ExplorerSubmenu
  • Update Error Code

[1.5.0]

  • Code optimization

[1.4.9]

  • Add CodeReview
  • Add Codebase
  • Enhanced Code Completion.
  • New Authentication Flow
  • Add Quota Management
  • Optimized UI

[1.4.8]

  • Add Gemini CLI provider
  • Fix known issues
  • Optimize Code

[1.4.7]

  • Fix model outputs Unix system commands causing PowerShell execution failure
  • Add timeout for model request
  • Optimize logs

[1.4.6]

  • Fix known issues

[1.4.5]

  • Prompts optimization
  • Optimize logs

[1.4.4]

  • Add qwen25-vl-32b model info to support imege input
  • Code optimization
  • Fix known issues

[1.4.3]

  • Add clearHistory method for task history cleanup when reset extension
  • Fix known issues

[1.4.2]

  • Update CoStrict provider and error handling improvements
  • Update vscode engine requirement to ^1.86.2

[1.4.1]

  • Update svg and add re-login button
  • Adds onboarding messages and new settings descriptions
  • Fix On-Premise model info
  • Add window reload check for command availability

[1.4.0]

  • Sync upstream changes

[1.3.3]

  • Fixed the issue of exceeding file read limits under specific scenarios
  • Added support for configuring custom OAuth2 endpoints

[1.3.2]

  • Remove unnecessary files to reduce the plugin size
  • Disable request resending in automatic retry mode upon 400 responses from the backend

[1.0.0]

  • Initial release with code auto-completion support