AI Translation Workflow

JekyllNet is no longer limited to manually maintained zh and en content. If _config.yml includes an ai section, the build can help create and maintain more locale variants.

Provider support

The current provider model supports:

• OpenAI
• DeepSeek
• Ollama
• OpenAI-compatible third-party providers through a custom base_url

How the multilingual flow works

The intended pattern is:

• define locales in _config.yml
• keep source-language pages in a predictable mirrored path structure
• let the builder generate translation_links
• use AI translation when you want to scale past hand-maintained copies

Cache, incremental reuse, and glossary

The translation workflow now includes three operational pieces that matter in practice:

• cache under .jekyllnet/translation-cache.json
• incremental translation so unchanged content is not re-requested
• glossary support for product names, repository names, and domain terms