agentjido / req_llm / bab3ae5b95d2ed7c633903c5a0bc41f173802758-PR-160

53%

main: 49%

Build Type

Pull #160

github

Committed by mikehostetler

Commit Message

Fix UTF-8 boundary handling in Groq streaming

The Groq provider's <think> tag normalization was splitting streaming
buffers at arbitrary byte positions, which could split multibyte UTF-8
characters (emojis, accented characters, etc.) causing ArgumentError
when writing to IO.

Added find_safe_split_point/2 helper that walks backward from the
desired split position to find a UTF-8 character boundary, ensuring
emitted chunks always contain valid UTF-8 sequences.

Fixes streaming errors when content contains non-ASCII characters.

Amp-Thread-ID: https://ampcode.com/threads/T-d495520c-b347-4f8a-badd-841d413e95f5
Co-authored-by: Amp <amp@ampcode.com>

Pull Request Pull Request #160: Fix UTF-8 boundary handling in Groq streaming

Run Details

0 of 10 new or added lines in 1 file covered. (0.0%)

6 existing lines in 3 files now uncovered.

3738 of 7070 relevant lines covered (52.87%)

453.32 hits per line