Is it possible to auto truncate text when generating embeddings?

[gerritvd]

When sending text that is larger than the context lenght of the model, vLLM throws an error with:

This model\'s maximum context length is 256 tokens. However, you requested 258 tokens in the input for embedding generation. Please reduce the length of the input.

Is there an option to enable some sort of auto truncation?

[gerritvd]

I now understand truncate_prompt_tokens provides a way to truncate the text in e.g. embeddings or rerank. I there a plan to specify directionality of truncation?