v1.1.0 - OpenAI Support & Configurable Models

What's New

Major Features

• OpenAI Support: Added support for OpenAI GPT models as a translation provider
• Configurable Models: Choose specific models for both Gemini and OpenAI providers

New CLI Options

• --gemini-model <model>: Select Gemini model (default: gemini-2.0-flash-lite)
  • Supports: gemini-2.5-flash, gemini-2.0-flash-lite, gemini-pro, and more
• --openai-model <model>: Select OpenAI model (default: gpt-4o-mini)
  • Supports: gpt-4o, gpt-4o-mini, gpt-4-turbo, gpt-3.5-turbo, and more
• --provider openai: Use OpenAI as the translation provider

Configuration

• Set OPENAI_API_KEY environment variable to use OpenAI
• Automatic provider detection based on available API keys
• List available providers with --list-providers

Improvements

• Enhanced provider flexibility with three options: Gemini (cloud), OpenAI (cloud), or Ollama (local)
• Better model selection for different use cases and performance requirements
• Updated documentation with comprehensive examples

Technical Details

• Added OpenAI translator implementation with proper error handling
• Extended TranslatorFactory to support multiple cloud providers
• Comprehensive test coverage for all new features
• Maintained backward compatibility with existing configurations

Installation

```bash
npm install -g translator-ai@1.1.0
```

Example Usage

```bash

Use OpenAI with GPT-4o

translator-ai en.json -l es -o es.json --provider openai --openai-model gpt-4o

Use Gemini with 2.5-flash model

translator-ai en.json -l fr -o fr.json --gemini-model gemini-2.5-flash
```

v1.0.10 - Key Verification and Sorting

New Features

Key Verification with --check-keys

• Verify all source keys exist in translated output
• Exit with error if any keys are missing
• Detailed reporting of missing keys with full paths
• Helps catch translation API failures or formatting issues

Alphabetical Key Sorting with --sort-keys

• Sort all JSON keys alphabetically in output files
• Case-insensitive sorting for consistent ordering
• Recursive sorting through nested objects
• Makes diffs cleaner and more readable

Opt-in Metadata

• Metadata is now disabled by default for better compatibility
• Use --metadata flag to include translation metadata
• Prevents issues with strict i18n parsers

Improvements

• Preserve all source JSON keys including non-translatable values (version numbers, etc.)
• Better error reporting with detailed key paths
• Consistent output structure across all translation providers

Bug Fixes

• Fixed issue where non-translatable strings were being dropped from output
• Ensured version numbers, URLs, and other filtered content is preserved
• Improved type safety in TypeScript code

Usage Examples

# Verify all keys are present after translation
translator-ai en.json -l es -o es.json --check-keys

# Sort keys alphabetically for consistent output
translator-ai en.json -l fr -o fr.json --sort-keys

# Include metadata about the translation
translator-ai en.json -l de -o de.json --metadata

# Combine all features
translator-ai src/**/*.json -l es,fr,de -o "{dir}/{name}.{lang}.json" \
  --check-keys --sort-keys --preserve-formats --stats

Installation

npm install -g translator-ai@1.0.10

v1.0.9 - Advanced Translation Features

New Features

Language Detection

• Automatically detect source language with --detect-source flag
• 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)
• 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

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

Improvements

Ollama Reliability

• Increased retry attempts from 3 to 5 with exponential backoff
• Progressive prompt simplification on retries
• Enhanced JSON extraction with 5 different strategies
• Support for various response formats from different models

Documentation

• Updated README with examples of all new features
• Comprehensive .env.example with all configuration options

Testing & CI

• Added 21 new unit tests for new features
• All 79 unit tests passing
• All Ollama integration tests passing
• Fixed CI workflows to only run unit tests (no Ollama required)
• Fixed npm-publish workflow for reliable releases

Installation

npm install -g translator-ai@1.0.9

Example Usage

# Detect source language and translate to multiple languages
translator-ai content.json -l es,fr,de --detect-source -o translations/{lang}.json

# Dry run to preview what would be translated
translator-ai app.json -l es --dry-run -o app-es.json

# Preserve technical formats during translation
translator-ai config.json -l ja --preserve-formats -o config-ja.json

# Combine all features
translator-ai src/**/*.json -l es,fr,de -o "{dir}/{name}.{lang}.json" \
  --detect-source --preserve-formats --stats

v1.0.8 - License field update

What's Changed

Changed

• Updated license field in package.json to properly reflect CC-BY-NC-SA-4.0 with commercial attribution terms

This is a minor metadata update to ensure proper license identification in npm registry.

Full Changelog: v1.0.7...v1.0.8

v1.0.7 - Improved Ollama robustness and validation

What's Changed

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
• Strict 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

Installation

npm install -g translator-ai@1.0.7

Example Usage

# Basic translation
translator-ai --provider ollama -l es input.json -o output.json

# With verbose debugging
translator-ai --provider ollama --verbose -l pl input.json -o output.json

# Test with chatwoot example
translator-ai --provider ollama -l pl examples/test-more/chatwoot-en.json -o chatwoot-pl.json

Full Changelog: v1.0.6...v1.0.7

v1.0.6 - Ollama translation fixes and verbose logging

What's Changed

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

Installation

npm install -g translator-ai@1.0.6

Testing Ollama with Verbose Output

translator-ai --provider ollama -l es --verbose input.json -o output.json

Full Changelog: v1.0.5...v1.0.6

v1.0.5 - Renamed to translator-ai with Ollama support

What's Changed

Major Updates

• Renamed from translator-gemini to translator-ai to reflect multi-provider support
• Added Ollama Support for local translation without API costs
• Model Context Protocol (MCP) integration for AI assistants

Features

• 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
• MCP server implementation for direct integration with AI assistants

Improvements

• Abstracted translation logic into provider-based architecture
• Better handling of different LLM response formats
• Enhanced error handling and timeout management
• Comprehensive test suite with 67 tests covering all functionality

Installation

npm install -g translator-ai

Binary Names

• translator-ai - Main CLI tool
• translator-ai-mcp - MCP server for AI assistant integration

Full Changelog: v1.0.4...v1.0.5

v1.0.3 - Fix Keys with Dots

translator-gemini v1.0.3

🐛 Critical Bug Fix

Fixed a critical issue where JSON files with keys containing dots (like auth.login or legal.terms-of-service) would fail with "Cannot create property on string" error.

🔧 Technical Details

• Implemented proper key escaping using null byte separators
• Keys with dots are now preserved literally instead of being interpreted as nested paths
• This fix is essential for i18n files that use dot notation in their key names

📋 Example

The tool now correctly handles JSON structures like:

{
  "menu": {
    "legal": "Legal",
    "legal.terms-of-service": "Terms of Service"
  }
}

🧪 Testing

• Added comprehensive tests for keys with dots
• All 35 tests passing
• Tested with real-world i18n files from society-flow project

---

For installation and usage, see the README.

If you find this tool useful, consider supporting the development:

v1.0.2 - NPM Publishing Fix

translator-gemini v1.0.2

🔧 NPM Publishing Fixes

• Added --access public flag for npm publishing
• Fixed repository URL format
• Updated workflow for automation token support

📝 Important Note for npm Publishing

This release requires an npm automation token (not a regular publish token) to be set as npm_token in GitHub secrets for automated publishing to work.

---

For installation and usage, see the README.

If you find this tool useful, consider supporting the development:

v1.0.1 - Build Fix

translator-gemini v1.0.1

🐛 Bug Fixes

• Fixed build configuration for npm publishing
• Added proper file inclusion in package.json
• Updated postbuild script to handle missing files gracefully
• Added explicit include/exclude in tsconfig.json

📦 Package Improvements

• Ensures dist directory is properly included in npm package
• Fixed GitHub Actions workflow build process

🧪 Testing

• Updated version numbers in tests to match release

---

For installation and usage, see the README.

If you find this tool useful, consider supporting the development: