Generated our changelog for last 3 weeks

Author: jahooma

scripts/changelog/2025-05-18-version-10283-release-enhanced-assistant-experience-system-stability.mdx

@@ -0,0 +1,69 @@
+---
+title: "Version 1.0.283 Release - Enhanced Assistant Experience & System Stability"
+description: "Week 20, 2025 — May 12 to May 18"
+---
+
+### 🚀 New Features
+
+- **Self-Review Capability**: Added self-review feature to improve response quality
+- **Enhanced Help Menu**: Implemented better help menu with improved user guidance
+- **Deterministic Sampling**: Introduced seedrandom package and deterministic sampling with `sampleSizeWithSeed` for more consistent behavior
+
+### 🔧 Improvements
+
+- **Assistant Message Decoupling**: Each assistant message is now decoupled from previous messages for better context management
+- **Enhanced User Guidance**:
+  - Added clearer current working directory (cwd) prompting and parameter support
+  - Improved instructions for tool results
+  - Enhanced prompting about using the end_turn tool
+- **Client Architecture**: Made client singleton for better resource management
+- **Terminal Command Security**: Terminal commands now only allowed from project root directory
+- **File Processing**: Added file path information in error messages to assistant for better debugging
+- **Training Data Optimization**:
+  - Request key files for training and merge with non-obvious files
+  - Only record file training data when files are requested
+  - Users no longer charged for filepicker training data
+
+### 🐛 Bug Fixes
+
+- **XML Processing**:
+  - Fixed saxy undefined defaultEncoding issue by always using UTF-8
+  - Added safeguards for agent-provided XML
+  - Improved warnings for invalid XML
+- **File Operations**:
+  - Made regex for parseReadFilesResult more lenient with whitespaces
+  - Fixed out-of-order chunks issue
+  - Added carriage returns to all newlines
+- **History Management**: Fixed bug with deleting history
+- **Balance System**:
+  - Initialize balance breakdown with all grant types (#144)
+  - Always itemize principals
+  - Removed invoice.created event (all invoices charged at zero dollars)
+- **Testing**:
+  - Fixed failing saxy tests
+  - Added referenced_by tag to simplify-tool-results tests
+  - Resolved type errors in web and backend related to GrantType
+
+### 🔄 System Changes
+
+- **Database Optimization**:
+  - Added index on message for finished_at
+  - Changed index to be on finished_at AND id for better query performance
+- **Tool Management**:
+  - Removed end_turn + await_tool_results tools
+  - Added back end_turn tool with simplified reprompting for pending tool results
+- **Agent Behavior**:
+  - Reduced maximum consecutive assistant messages
+  - Less agentic behavior in lite mode
+  - Decreased chunk limit for better performance
+  - Enhanced prompt history with non-unique prompts moved to end
+
+### 🛠️ Technical Improvements
+
+- **Error Handling**: Improved error message passing from processFileBlock and processStrReplace
+- **Logging**:
+  - Added user_input tracking (distinct from agent_step)
+  - Removed prompt_sent logging
+  - Turned off credit log spam
+  - Added agent_step event tracking in main prompt
+- **CI/CD**: Multiple GitHub Actions fixes and improvements for better build reliability

scripts/changelog/2025-05-25-version-10295-enhanced-ai-models-and-user-experience-improvements.mdx

@@ -0,0 +1,74 @@
+---
+title: "Version 1.0.295 - Enhanced AI Models and User Experience Improvements"
+description: "Week 21, 2025 — May 19 to May 25"
+---
+
+### 🚀 New Features
+
+- **Claude Opus Model Support**: Added support for the new Claude Opus model with improved labeling (Opus + Sonnet)
+- **Slash Menu System**: Introduced interactive slash command menu with enhanced formatting and descriptions
+- **Project Setup Command**: Added `/init` command to help users initialize their project configuration
+- **File Picker Grading**: Added groundwork for file-picker grading functionality with new relabel options in UI
+
+### 🔧 Model & AI Improvements
+
+- **Upgraded to Claude Sonnet 4**: Switched from Sonnet 3.6 to the more powerful Sonnet 4 model
+- **Updated Flash Model**: Updated to Flash 2.5 (May 20th preview version)
+- **O4-Mini Integration**: Now using O4-Mini for large file search and replace operations
+- **Enhanced Prompting**: Added Anthropic-recommended prompts and improved instruction following
+- **Reduced Verbosity**: Streamlined Gemini thinking output now that Claude 4 provides stronger performance
+
+### 🛡️ Security & Safety
+
+- **Stronger Script Protection**: Enhanced prompting to prevent unauthorized script execution
+- **Background Process Warnings**: Added prompting alerts for background processes
+- **Improved Safety Instructions**: Added better instructions for escaping and handling dangerous operations
+
+### 🐛 Bug Fixes
+
+- **Code Search Fix**: Resolved code search functionality and added proper spacing before periods
+- **Quote Preservation**: Fixed issue where quotes were incorrectly removed from code search patterns
+- **XML Entity Parsing**: Corrected XML entity parsing issues
+- **Command/Ctrl Click**: Fixed broken command/ctrl click functionality
+- **Windows Terminal Fix**: Resolved terminal command hanging issues on Windows
+- **Newline Handling**: Fixed bugs related to newline processing
+
+### ⚡ Performance & UX Enhancements
+
+- **Larger File Support**: Enabled relace functionality for larger files with configurable limits
+- **Improved Context Handling**: Enhanced Context 7 functionality with increased timeout
+- **Better Directory Management**:
+  - Added `isSubdir` function for better directory handling
+  - Improved startup behavior when launched from home or root directories
+  - Enhanced current working directory prompting
+- **Readline Improvements**: Made readline functionality public for better accessibility
+- **Enhanced Analytics**:
+  - Added user input completion tracking with latency metrics
+  - Improved analytics for help menu usage
+  - Added failure tracking for update processes
+
+### 🎨 Interface Improvements
+
+- **Restored Classic Coloring**: Brought back randomized coloring system (selects 3 colors first)
+- **Better Help System**:
+  - Improved help command table spacing and formatting
+  - Renamed "Diagnostics Check" to "Project Setup" for clarity
+  - Enhanced startup menu for home directory scenarios
+- **Slash Command Enhancements**:
+  - Simplified command processing with filter + sort chains
+  - Added dedicated methods for command cleaning and validation
+  - Improved error handling for unknown commands
+
+### 🔧 Technical Improvements
+
+- **Dependency Updates**: Switched from `readable-stream` to `node:stream` for better Node.js compatibility
+- **Parameter Handling**: Improved parsing to ignore new parameters while processing existing ones
+- **Event Tracking**: Enhanced event tracking system with better failure handling
+- **Content Management**:
+  - Improved pasted content consideration for slash menu
+  - Better handling of fresh prompts and content clearing
+  - Enhanced user message tagging system
+
+---
+
+_This release focuses on significantly improving AI model capabilities, user safety, and overall experience while maintaining backward compatibility._

scripts/changelog/2025-06-01-version-10305-release-enhanced-ai-models-new-commands-and-performance-improvements.mdx

@@ -0,0 +1,61 @@
+---
+title: "Version 1.0.305 Release - Enhanced AI Models, New Commands, and Performance Improvements"
+description: "Week 22, 2025 — May 26 to June 1"
+---
+
+### 🚀 New Features
+
+- **New AI Models**: Added support for multiple new fine-tuned models including `top-2-tuned` and `ft-file-picker 007` for improved performance
+- **Export Command**: Introduced `/export` command to summarize conversations and save them to files for easy sharing and documentation
+- **Ask Mode**: Implemented new `/ask` mode for streamlined interactions (mode switching now supports optional "/" prefix)
+- **Reset Command**: Added `/reset` command to clear conversation history and reset project file context cache
+- **Experimental Commands**: New experimental command framework for testing upcoming features
+
+### 🔧 Improvements
+
+- **AI SDK Migration**: Major refactoring to use unified AI SDK approach, consolidating provider-specific implementations and reducing code duplication
+- **Prompt Caching**: Reorganized agent messages and instructions for better prompt caching performance
+- **Message Management**:
+  - Added time-to-live (TTL) functionality for messages with automatic expiration
+  - Improved message schema migration to CoreMessage format
+  - Enhanced tool message handling with better error handling
+- **Performance Optimizations**:
+  - Increased output token budget for main agent
+  - Switched to Promise.allSettled for relabel operations
+  - More modular tools.ts refactoring
+  - Improved monthly usage calculations to use current month data
+
+### 🐛 Bug Fixes
+
+- **UI Fixes**:
+  - Fixed scrolling issues in file picker table
+  - Prevented unwanted navigation in admin file picker
+- **Message Handling**:
+  - Fixed duplicate "previous assistant" instructions appearing multiple times
+  - Resolved cache control issues with previous messages
+  - Fixed tool call ID extraction for different message formats
+- **System Stability**:
+  - Improved WebSocket connection handling with proper cleanup on exit
+  - Better error handling for PostHog analytics
+  - Fixed npm-app build issues
+
+### 🔄 Technical Changes
+
+- **Code Organization**:
+  - Removed legacy API files (deepseek-api.ts, openai-api.ts, gemini-api.ts)
+  - Consolidated LLM streaming implementations
+  - Improved eval system with better file organization
+- **Search Improvements**:
+  - Added '@' symbol for matching file paths and tokens
+  - Limited search matches to 20 results for better performance
+  - Added `code_search` as stop sequence
+- **Analytics**: Enhanced tracking with git repository statistics and improved user data collection
+
+### 📚 Documentation
+
+- Updated README with new fine-tuned model information
+- Improved npm package documentation
+
+---
+
+_This release focuses on consolidating our AI infrastructure while adding powerful new features for better user experience and performance._

scripts/changelog/docs.json

@@ -0,0 +1,19 @@
+{
+  "navigation": {
+    "tabs": [
+      {
+        "tab": "Changelog",
+        "groups": [
+          {
+            "group": "Q2 2025",
+            "pages": [
+              "changelog/2025-06-01-version-10305-release-enhanced-ai-models-new-commands-and-performance-improvements",
+              "changelog/2025-05-25-version-10295-enhanced-ai-models-and-user-experience-improvements",
+              "changelog/2025-05-18-version-10283-release-enhanced-assistant-experience-system-stability"
+            ]
+          }
+        ]
+      }
+    ]
+  }
+}