Google Gemini stop reason not handled with strictOpenAICompliance: true #1200
Description
What Happened?
With strictOpenAICompliance: true, we are still seeing Google's Gemini Flash 2.0 model return a finish reason of MAX_TOKENS, ideally this would be translated to match Open AI's specifications.
What Should Have Happened?
A stop reason of length is returned when Gemini Flash 2.0 runs out of tokens.
Relevant Code Snippet
Use Gemini Flash 2.0 with a particularly long block of text and ask it to add comments to the entire block. Gemini will then return a finish reason of MAX_TOKENS.
Your Twitter/LinkedIn
No response