derevo

Results 10 comments of derevo

> Thanks for providing these details. Currently Langfuse does tokenization for openai and anthropic models and relies on ingesting token counts for other models. This does not seem to work...

I found best solution we move all to -> qdrant

We make it work by using this tool - [https://github.com/instructor-ai/instructor ](https://github.com/instructor-ai/instructor ) Complex structured output just does not works correctly with gemini models and langchain.