The community member noticed that the VLLM (Very Large Language Model) requires the model parameter in the payload, but it is not being sent from the llama-index-llms-vllm. Additionally, the payload to the /chat/completions is being translated to prompt, when it should be messages: [{role:'',content:''},{role:'',content:''}].
In the comments, another community member suggests that the structure of the VLLM server class was likely created that way when it was first added, and that it needs modification. Others agree that this needs to be addressed.
The final comment explains that the reason for this is that the /chat/completions endpoint cannot be used, and the /completions endpoint should be used instead, as the payload is not compatible with the OpenAI format.
