translator-ai

# Changelog All notable changes to this project will be documented in this file. The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.0.0/), and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0.html). ## [1.0.9] - 2025-01-20 ### Added - **Language Detection**: Automatically detect source language with `--detect-source` flag - Uses AI to identify the language of input strings - No longer assumes English as the source language - **Multiple Target Languages**: Translate to multiple languages in a single command - Support for comma-separated language codes (e.g., `-l es,fr,de,ja`) - Optimizes API usage by reusing source analysis - Output patterns support `{lang}` variable for multi-language workflows - **Dry Run Mode**: Preview translations without making API calls with `--dry-run` - Shows detailed statistics about what would be translated - Displays cache hit rates and estimated API calls - Useful for cost estimation and debugging - **Format Preservation**: Maintain special formats unchanged with `--preserve-formats` - Preserves URLs, email addresses, and file paths - Keeps template variables intact ({{var}}, {0}, %s, ${var}, :param) - Maintains numbers with units, currency, dates, and version numbers - Protects hex colors and other technical formats ### Improved - README documentation with examples of new features - CLI help text with clearer descriptions ## [1.0.8] - 2025-01-20 ### Changed - Updated license field in package.json to properly reflect CC-BY-NC-SA-4.0 with commercial attribution terms ## [1.0.7] - 2025-01-20 ### Fixed - Ollama no longer falls back to source language when translation fails - Empty or missing translations now properly throw errors - Better handling of malformed JSON responses from LLMs ### Added - Retry logic with exponential backoff (3 attempts) for Ollama translations - Enhanced JSON extraction supporting multiple response formats - Validation to ensure all fields are properly translated ### Improved - More descriptive error messages for debugging - Robust parsing for responses with extra text before/after JSON - Test suite updated to handle retry behavior ## [1.0.6] - 2025-01-20 ### Fixed - Ollama translation response parsing for DeepSeek-R1 model - Handle JSON responses with 'translations' property wrapper - Improved error handling for various LLM response formats ### Added - `--verbose` flag for debugging Ollama API requests and responses - Detailed logging of Ollama request/response for troubleshooting ### Improved - CI workflows with better Ollama service health checks - Unit test separation for faster CI feedback ## [1.0.5] - 2025-01-20 ### Added - **Ollama Support**: Added support for local translation using Ollama with DeepSeek-R1 - New `--provider` flag to choose between 'gemini' or 'ollama' - `--ollama-url` to specify custom Ollama API endpoint - `--ollama-model` to use different Ollama models - `--list-providers` command to check available providers - Optimized prompts and parameters for DeepSeek-R1 format - Handles both JSON array and object responses from Ollama - No API costs - runs entirely on local hardware - **Model Context Protocol (MCP)**: Added MCP server for Claude Desktop integration - New `translator-ai-mcp` binary for MCP server - Supports both Gemini and Ollama providers via environment variable - Enables direct file translation within Claude conversations ### Changed - Abstracted translation logic into provider-based architecture - Updated CLI to support multiple translation backends - Enhanced error handling for different response formats ### Fixed - Improved JSON extraction from various LLM response formats - Better handling of thinking tags in DeepSeek responses ## [1.0.4] - 2025-01-20 ### Added - **Multi-file Processing**: Process multiple files with automatic deduplication - Accepts multiple input files and glob patterns - Identifies duplicate strings across files - Translates each unique string only once - Shows deduplication statistics and API call savings - Example: 23% savings when processing 4 files with shared strings ### Changed - CLI now accepts variadic arguments for multiple files - Output pattern supports `{dir}`, `{name}`, and `{lang}` variables ## [1.0.3] - 2025-01-20 ### Fixed - Critical bug with JSON keys containing dots (e.g., "auth.login") - Implemented proper key escaping using null bytes as separators - Fixed issue with society-flow test case ## [1.0.2] - 2025-01-20 ### Fixed - npm repository URL warnings - Package metadata improvements ## [1.0.1] - 2025-01-20 ### Added - GitHub Actions workflow for automated npm publishing - Cross-platform cache directory support - Windows: `%APPDATA%\translator-ai\` - macOS: `~/Library/Caches/translator-ai/` - Linux: `~/.cache/translator-ai/` ### Fixed - Build issues with postbuild script - TypeScript configuration ## [1.0.0] - 2025-01-20 ### Initial Release - Fast translation using Google's Gemini 2.0 Flash Lite model - Incremental caching system with SHA-256 hashing - Batch processing with dynamic sizing - Path preservation for complex JSON structures - Global npm installation support - Comprehensive CLI with multiple options - Performance statistics and progress indicators