The community member is using OpenAILike as their language model (LLM) and vLLM as an interference server. They are experiencing issues with the context and the model going back and forth. The comments suggest that setting is_chat_model=True in the LLM may help, which the community member seems to have fixed. Additionally, updating vLLM to handle special characters also helped. The community members discuss the prompt as a potential issue, and one member suggests that setting is_chat_model=True in the LLM should allow vLLM to handle the prompt formatting.
