feat(Translate): use structured output to limit extraneous text#326
Merged
feat(Translate): use structured output to limit extraneous text#326
Conversation
kyteinsky
force-pushed
the
feat/noid/structured-translation
branch
4 times, most recently
from
fa7c77d to
74104af
Compare
Signed-off-by: Anupam Kumar <kyteinsky@gmail.com>
kyteinsky
force-pushed
the
feat/noid/structured-translation
branch
from
74104af to
8f86401
Compare
julien-nc
approved these changes
Jan 16, 2026
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Tested with:
The text used were like a normal English paragraph translated to German and Spanish, the Circles/Teams readme translated to German and a normal Hindi paragraph translated to English and Spanish. Also tested chunking.
It mostly works fine but some models have issues with long Hindi texts and many symbols in the Circles readme like CodeLlama, Nemo and Qwen3 0.6B, others work fine.
Structured outputs for OpenAI is only supported in "gpt-4o-mini, gpt-4o-2024-08-06, and later", maybe that isn't that big a problem?
https://platform.openai.com/docs/guides/structured-outputs?type-restrictions=string-restrictions&example=structured-data#structured-outputs-vs-json-mode