lingo.dev@0.132.3
·
1536 commits
to main
since this release
Patch Changes
-
#2006
4e4bea8Thanks @cherkanovart! - fix: strip conversational preamble from LLM responses before JSON parsingSome 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.