Smarter translation algorithm, glossary support, better cost and progress tracking, and general cleanup

[vedmakk]

Hi @strickvl,

I’m not sure if you're still actively maintaining this repository, but I wanted to thank you for building it. I used your tool to translate a 200-page book, and it worked great. Along the way, I made several improvements, added a new translation algorithm, and closed some loose ends. Since your work laid the foundation, I thought it would be good to contribute these changes back in case you (or others) find them useful.

This PR is a bit large—apologies for that—but here's a summary of the main changes:

• README: streamlined and expanded with all the information needed to set up and run the project
• Dependencies: updated package versions (LiteLLM, Pydantic, …)
• Cost estimation: now includes both input and output token calculations
• Progress information: added progress bar, current costs, updated cost estimates, and time estimations during translation
• New translation algorithm: context-aware algorithm for .txt files with smarter text splitting to avoid chunk-edge issues and provide more context to the model
• Checkpoint save/resume: Now fully wired up and working
• Glossary support: automatically builds a glossary of key terms to maintain consistent translations
• --max-cost option: Now safely exits the process if translation costs exceed the set limit
• Model Reasoning control: support for minimal/low/medium/high reasoning modes to adjust translation quality
• General code cleanup

There's still plenty of room for future improvements, but the tool now works quite well for .txt file translation. If you're still interested in using or developing it further, I'd be glad to sync up and discuss ideas.

Thanks again for your initial work on this project - it made translating a full book possible.

Kind regards,
vedmakk

[strickvl]

Oh lovely @vedmakk! I'll check this out over the weekend. I do still use this project myself, but haven't added new features recently. I'll take a look at your changes and test them out. Beyond the notes you made, anything you'd recommend I watch out for or try out when I'm testing?

[vedmakk]

Nice! The updated README should contain all the information needed to test the changes.

Also, take a look at the plans directory in the root. It includes detailed documentation for the new context-aware algorithm and the glossary feature. I usually create these documents after longer discussions with Claude Code to “plan” the implementation, and then let a new coding agent execute on it. This workflow both improves the quality of the implementation and leaves behind detailed documentation to use as a reference.

For testing, check out examples/elara_story.txt and examples/elara_story.pdf. The commands in the Quick Start section of the README point directly to these files, so they should work out of the box.