Claude Conversation Logger

# Changelog All notable changes to the Claude Conversation Logger project will be documented in this file. ## [3.1.0] - 2025-08-26 ### 🤖 Agent System Migration to Claude Code - **BREAKING**: Migrated from 6 JavaScript agents to 4 Claude Code compatible agents - **Eliminated redundancy**: Removed SessionStateAnalyzer and RelationshipMapper agents - **New format**: All agents now use Claude Code standard markdown format with YAML metadata - **Optimized performance**: Token usage reduced by 15-40 tokens per operation - **Improved relationships**: Established clear parent-child agent hierarchy ### 📋 Agent Architecture Overhaul - **conversation-orchestrator-agent**: Main coordinator with intelligent decision making - **semantic-analyzer-agent**: Multi-layer analysis with technical entity extraction - **pattern-discovery-agent**: Advanced pattern detection with temporal trends - **auto-documentation-agent**: Multiple template types with token-optimized generation - **Standardized format**: All agents follow uniCommerce agent standard with proper YAML metadata ### 📚 Documentation Updates - **README.md**: Updated from v3.0.0 to v3.1.0 with 4 Claude Code agents - **Project structure**: Updated to show `.claude/agents/` directory with markdown files - **Metrics updated**: Reflected 4 agents instead of 6 in all statistics - **MCP tools table**: Updated agent names to match new system - **Mermaid diagrams**: Updated architecture diagrams for new agent system ### 🧹 Code Cleanup - **Removed obsolete files**: Deleted 6 JavaScript agent files and configurations - **Simplified MCP server**: Removed agent system imports and references - **Updated server health**: Now shows Claude Code agent status - **Cleaned references**: All obsolete JavaScript agent references removed ### ⚡ Performance Improvements - **Token optimization**: 70% more efficient than previous 6-agent system - **Faster startup**: Reduced initialization overhead with simpler agent architecture - **Better caching**: Optimized agent response caching for frequent operations - **Reduced complexity**: Simplified agent coordination and delegation patterns ## [2.1.4] - 2025-01-20 ### 🏗️ Architecture Simplification - **BREAKING**: Eliminated memory buffer system (over-engineered complexity) - Simplified to clean dual storage: MongoDB (persistent) + Redis (cache) - Removed MAX_MESSAGES and memory-related configurations - Streamlined data flow for better maintainability ### 📚 Documentation Overhaul - Corrected README.md architecture references and diagrams - Updated MCP usage examples to reflect current storage model - Fixed conversation examples with realistic current data - Modernized model references to claude-sonnet-4-20250514 - Eliminated outdated "Triple Storage System" references - Updated QUICK_START.md with correct storage description ### ⚡ Performance Optimization - Reduced memory footprint by 40% (eliminated unnecessary cache layer) - Dashboard reads directly from MongoDB with ~50ms response time - Simplified Redis configuration for optimal balance - Container startup time improved by 15% ## [2.1.3] - 2025-08-20 ### 📊 Added - Visual Dashboard - **NEW Endpoint**: `/dashboard` serves complete HTML dashboard with real-time statistics - **Interactive Charts**: Token distribution (donut chart) and project activity (bar chart) - **Key Metrics Cards**: Total messages, cost tracking, token consumption, active projects - **Model Usage Table**: Per-model statistics with detailed cost breakdown - **Auto-Refresh**: Dashboard updates automatically every 30 seconds - **Responsive Design**: TailwindCSS-based mobile-first responsive layout - **Real-Time Data**: Live statistics from current conversation logs - **Chart.js Integration**: Interactive, animated charts for better data visualization ### 🎨 Dashboard Features - **TailwindCSS**: Modern, responsive styling via CDN - **Chart.js**: Interactive donut and bar charts via CDN - **No Authentication**: Public endpoint for easy access - **Error Handling**: Graceful fallback if data unavailable - **Mobile Optimized**: Works perfectly on all device sizes ### 📚 Documentation Updates - **README**: Added comprehensive dashboard documentation with examples - **Endpoint Table**: Updated to include new `/dashboard` endpoint - **Visual Layout**: Added ASCII art representation of dashboard structure ## [2.1.2] - 2025-08-20 ### 🔧 Fixed - Critical Token Counting Bug - **MAJOR FIX**: Resolved severe token underreporting (99%+ error rate) in assistant responses - **Enhanced Hook**: Improved `parse_transcript_for_last_assistant_message()` to properly accumulate streaming response chunks - **Token Accuracy**: Fixed usage extraction to capture complete response statistics instead of partial chunks ### 🎯 Added - OpenTelemetry Integration - **New Endpoint**: `/api/token-usage` for OpenTelemetry-compliant token metrics - **Token Separation**: Individual records for `input`, `output`, `cacheRead`, `cacheCreation` token types - **Cost Estimation**: Automatic cost calculation by model with USD precision to 6 decimals - **Duration Tracking**: Response time measurement from first to last timestamp - **Enhanced Metadata**: Extended message records with `cost_usd` and `duration_ms` ### 🚀 Added - Token Analytics Dashboard - **New Endpoint**: `/api/token-stats` with detailed token usage analytics - **Enhanced Stats**: Updated `/api/stats` endpoint to include token metrics summary - **Filtering Support**: Token statistics by project, time period, and model - **Cost Breakdown**: Per-token-type cost analysis for accurate billing insights ### 🛠️ Technical Improvements - **Storage Functions**: Refactored `saveMessage()` and added `saveTokenUsage()` - **Unified Storage**: Created `saveToStorage()` for consistent data persistence - **Better Logging**: Enhanced console output to show token record types - **Validation**: Added input validation for token usage endpoints ### 📚 Documentation Updates - **README**: Added comprehensive OpenTelemetry section with examples - **API Documentation**: Updated endpoints table with new token metrics APIs - **Troubleshooting**: Added section explaining the token counting fix - **Usage Examples**: Provided curl examples for token statistics endpoints ### 🧪 Testing - **Validation Script**: Created comprehensive test suite for token parsing functions - **OpenTelemetry Compliance**: Verified all token types match specification requirements - **Integration Tests**: Validated end-to-end token tracking workflow ## [2.1.1] - 2025-08-20 ### 🎯 MCP Configuration Improvements - **Added .mcp.json support** - Project-specific MCP configuration with automatic detection - **Documented both configuration methods** - .mcp.json vs global settings.json - **Updated troubleshooting guide** with .mcp.json solutions ### 📚 Documentation - Enhanced README with comparison between .mcp.json and settings.json - Updated QUICK_START with both configuration options - Improved troubleshooting section with MCP-specific solutions ## [2.1.0] - 2025-08-20 ### 🎯 Major Fixes - **Fixed hook configuration structure** - Corrected nested structure required by Claude Code - **Fixed user message capture** - UserPromptSubmit hook now works correctly - **Improved token parsing** - Accumulates streaming chunks for accurate token counts ### ✨ Features - Added real conversation example in documentation - Enhanced troubleshooting section with common issues - Improved token usage explanation ### 📚 Documentation - Updated all configuration examples with correct hook structure - Translated QUICK_START.md to English - Added detailed explanation of token types and usage - Created comprehensive troubleshooting guide ### 🐛 Bug Fixes - Fixed duplicate hook configurations across settings files - Corrected parsing logic for streaming assistant responses - Resolved issue with missing user messages in database ### 🔧 Technical Details #### Correct Hook Structure ```json { "hooks": { "UserPromptSubmit": [ { "hooks": [ {"type": "command", "command": "python3 ~/.claude/hooks/api-logger.py"} ] } ] } } ``` #### Token Parsing Improvements - Accumulates content from all streaming chunks - Captures usage statistics from final chunk only - Preserves complete assistant responses ## [2.0.0] - 2025-08-19 ### Features - Initial monolithic container architecture - Triple Storage System (MongoDB + Redis + Memory) - MCP server integration - Automatic failover and redundancy ## [1.0.0] - 2025-08-18 ### Features - Basic conversation logging - Simple file-based storage - Initial hook implementation