Activate langchain functionality. This adds a simple reply_chain, which takes one message from the user and generates a response. System prompt is included as normal but I have also added a "character" prompt as seen in this prompt hack to combat the generic Chat voice and protect against prompt injection.
Next steps will involve building whole message histories into the chain, and using more complicated langchain features. Should we merge this in first?
Activate langchain functionality. This adds a simple reply_chain, which takes one message from the user and generates a response. System prompt is included as normal but I have also added a "character" prompt as seen in this prompt hack to combat the generic Chat voice and protect against prompt injection.
Next steps will involve building whole message histories into the chain, and using more complicated langchain features. Should we merge this in first?