Dolibarr MCP Server

# Changelog All notable changes to the Dolibarr MCP Server are documented here. The project follows [Semantic Versioning](https://semver.org/spec/v2.0.0.html) and adopts the [Keep a Changelog](https://keepachangelog.com/en/1.0.0/) format. ## [2.1.0] - 2026-01-27 ### Added - **TOON Format Output**: Token-Oriented Object Notation as default output format - ~40% reduction in token usage compared to JSON - Self-documenting tabular format: `[N]{field1,field2}:` - Automatic fallback to JSON on encoding errors - Format selection via `format` parameter (toon, json, json_compact) - **DragonflyDB Cache**: High-performance async caching layer - Compatible with DragonflyDB and Redis (25x faster with DragonflyDB) - Entity-specific TTL strategies (30s-1h based on data volatility) - Automatic cache invalidation on mutations - Graceful degradation when cache unavailable - Cache statistics tracking (hits, misses, hit rate) - **New modules**: - `formats/`: TOON encoder and format selection - `cache/`: DragonflyDB client and TTL strategies - **New dispatch functions**: - `dispatch_tool_cached()`: Cache-aware tool execution - `dispatch_tool_formatted()`: Returns TOON/JSON string output ### Technical Improvements - Cache strategies per entity type (products=15min, invoices=30s, etc.) - Automatic cache invalidation map for mutation operations - Optional `redis` dependency for cache support ## [2.0.0] - 2026-01-27 ### Added - **Modular Architecture**: Complete refactoring into modular structure - `client/`: API client with `base.py` and `exceptions.py` - `server/`: MCP server with `main.py`, `tools.py`, `handlers.py`, `responses.py` - `transports/`: Separated STDIO and HTTP transports - `schemas/`: Centralized schema definitions with `base.py`, `fields.py`, `entities.py` - **Response Wrapper System**: Unified response format with `success_response()`, `error_response()`, `paginated_response()` - **Enhanced Error Handling**: Structured errors with codes, retriable flags, and correlation IDs - `DolibarrAPIError`, `DolibarrValidationError`, `DolibarrNotFoundError`, `DolibarrConnectionError` - **Dynamic Tool Dispatcher**: Replaced 59 if-statements with declarative `TOOL_REGISTRY` lookup table - **AI-Optimized Descriptions**: All 59 tools now have detailed descriptions following format: `[ACTION] [OBJECT]. [FIELDS RETURNED]. [CONSTRAINTS/NOTES].` - **Schema Documentation**: Full JSON Schema with descriptions, constraints, and examples ### Changed - **Token Efficiency**: Responses now include metadata for pagination and debugging - **Error Responses**: Now include `code`, `message`, `status`, `retriable`, `details` - **Compatibility**: Backward compatible with existing tests via `dispatch_tool_legacy()` ### Technical Improvements - Reduced code duplication by ~40% - Single lookup table for tool dispatch (was 59 if-statements) - Centralized field filtering definitions - Cleaner separation of concerns ### Added - Restored README.md and CHANGELOG.md after merge conflicts while preserving the streamlined structure shared with `prestashop-mcp`. - Documented platform-specific setup covering Linux/macOS shells, Windows Visual Studio `vsenv`, and the Docker workflow. ### Changed - Reconciled feature and tool descriptions so they capture both the detailed ERP coverage and the new documentation bundle layout. - Clarified configuration guidance around `pydantic-settings`, environment variables, and `.env` files. ### Removed - Obsolete references to legacy helper scripts and superseded documentation variants that were dropped during the repository cleanup. ## [1.1.0] ### Removed - Legacy helper scripts, installers, and manual test programs that duplicated the automated test-suite. - Alternative server implementations (`simple_client`, `standalone_server`, `ultra_simple_server`) in favour of the single `dolibarr_mcp_server`. - Redundant documentation fragments and variant requirements files that no longer reflected the current project layout. ### Changed - Rewrote the README to highlight the streamlined structure and provide concise installation/run instructions. - Clarified that `dolibarr_mcp_server.py` is the definitive MCP entry point. ## [1.1.0] ### Removed - Legacy helper scripts, installers and manual test programs that duplicated the automated test-suite - Alternative server implementations (`simple_client`, `standalone_server`, `ultra_simple_server`) in favour of the single `dolibarr_mcp_server` - Redundant documentation fragments and variant requirements files that no longer reflected the current project layout ### Changed - Rewrote the README to highlight the streamlined structure and provide concise installation/run instructions - Clarified that `dolibarr_mcp_server.py` is the definitive MCP entry point ## [1.0.0] ### 🎯 Major Restructuring This release represents a complete restructuring of the Dolibarr MCP Server to match the clean architecture of prestashop-mcp. ### Added - Professional README.md with comprehensive documentation. - Structured test suite in `tests/` directory. - Clean configuration management. - Docker support for easy deployment. - Comprehensive CRUD operations for all Dolibarr entities. ### Changed - Complete repository restructuring to match prestashop-mcp pattern. - Simplified dependencies in requirements.txt. - Cleaned up package structure in `src/dolibarr_mcp/`. - Updated pyproject.toml with proper metadata. - Streamlined .gitignore file. ### Removed - Outdated prototype assets from the first public release. ### Technical Improvements - Single, focused MCP server implementation. - Clean separation of concerns. - Better error handling. - Improved logging. - Async/await architecture throughout. ## [0.5.0] ### Added - Initial Dolibarr API integration. - Basic CRUD operations for customers, products, invoices. - MCP server implementation. - Docker configuration. ## [0.1.0] ### Added - Initial project setup. - Basic repository structure. - License and documentation. --- **Note**: Entries prior to v1.0.0 summarise the historical evolution of the project and are retained for context.