We don't publish
your competitive advantage.

IFHuggingFacePipeline does not apply the model's chat template when using ChatPromptTemplate, resulting in a generic 'Human/AI' format instead of the model's expected tokens (e.g., <|user|>, <|assistant|>).

THENReplace HuggingFacePipeline with ChatHuggingFace, which internally converts the message list and leverages the tokenizer's apply_chat_template() to format prompts correctly. Additionally, ensure langchain-huggingface version >= 0.0.4 or manually apply the fix from commit 4796b7e to avoid a related bug (issue #22804).

IFWhen using LangChain ConversationChain with a HuggingFace model that has a specific chat template, and the prompt fails to match that template, the model may generate responses that repeat conversation history and loop.

THENEnsure that the input prompt is formatted using the model's own chat template. For HuggingFace models, use tokenizer.apply_chat_template() or leverage LangChain's ChatHuggingFace wrapper which handles formatting automatically. Avoid manually writing a template that doesn't match the model's expected tokens (e.g., system/user/assistant).

IFConversation with LangChain's ConversationChain using HuggingFace models results in the model repeating or looping after initial responses due to incorrect chat template formatting.

THENEnsure the prompt template matches the model's expected chat format. For HuggingFace models, use the tokenizer's apply_chat_template method to format the chat history and input. Alternatively, use LangChain's ChatPromptTemplate with MessagesPlaceholder and a system message that includes context, avoiding manual prompt templates that omit assistant labels or special tokens.