Hello Community,
I am using LangChain based ConversationChain for my task. I have done prompt engineering using PromptTemplate where I have to input_varibale 1) input 2) histroy. I have also added format_instructions and output parser in prompt template to get structure output.
Now before adding format_instructions and output_parser the whole conversation flow was working fine and chatgpt was working as intruscted in prompt. But when I added format_instructions and output parser it is replying same answer again and again with small changes.
I have checked how history is append at the end of the prompt as I have format_instructions I think that is causing the issue.
Example with format_instructions:
Human: Hello
AI : '''json
{
"AI": "How can I help",
"flag": "Responsive",
}'''
Instead if I have correct history like
Human: Hello
AI : "How can I help",
Now How to fix this issue ?