Starstruk Logo MarkStarstruk

lingo.dev@0.132.3

lingodotdev/lingo.devlingo.dev@0.132.3Feb 23, 2026by github-actions[bot]

AI Summary

Patch release improving JSON parsing reliability by stripping conversational text from LLM responses.

Key Highlights

  • Strip conversational preamble (e.g., 'OK') from LLM responses before JSON parsing
  • Use indexOf/lastIndexOf and jsonrepair as fallback for parsing

New Features

  • Improved LLM response parsing reliability

Full Release Notes

### Patch Changes

-   [#2006](https://github.com/lingodotdev/lingo.dev/pull/2006) [`4e4bea8`](https://github.com/lingodotdev/lingo.dev/commit/4e4bea8ee52ac97473b8edfc5489d15d551d68f6) Thanks [@cherkanovart](https://github.com/cherkanovart)! - fix: strip conversational preamble from LLM responses before JSON parsing

    Some models (e.g. Gemini 2.5 Flash Lite) prepend filler text like "OK" to their responses, causing JSON.parse to fail. Now structurally extracts the outermost JSON object using indexOf/lastIndexOf before parsing, with jsonrepair as fallback.