| from langchain.llms import OpenAI | |
| def extract_prompt(text): | |
| start_tag = "<prompt>" | |
| end_tag = "</prompt>" | |
| start_index = text.find(start_tag) | |
| end_index = text.find(end_tag) | |
| if start_index != -1 and end_index != -1 and start_index < end_index: | |
| start_index += len(start_tag) | |
| return text[start_index:end_index].strip() | |
| return "" | |
| def generate_system_message(name, openai_key): | |
| llm = OpenAI(openai_api_key=openai_key) | |
| text = """Generate a prompt for a large language model that tell's it to behave like {0}. | |
| Make sure the instructions gguarantee that the LLM never breaks character and never switches roles and at all time behaves like if it would be {0}. | |
| Make sure the LLM doesn't repeat it's assumed role all the time. | |
| Start the prompt with <prompt> and end it with </prompt>""".format(name) | |
| prompt = llm(text) | |
| print(prompt) | |
| response = extract_prompt(prompt) | |
| print(response) | |
| return response | |
| def generate_welcome_message(name1, name2, openai_key): | |
| llm = OpenAI(openai_api_key=openai_key) | |
| response = llm("Imagine {0} and {1} are meeting by chance on the street. To start a conversation, {0} says: ".format(name2, name1)).replace('"','') | |
| print(response) | |
| return response | |
| def filtered_response(chat, stack): | |
| invalid_response = True | |
| while(invalid_response): | |
| invalid_response = False | |
| response = chat(stack) | |
| if ( "an AI language model" in response ) or ( "an AI assistant") in response: | |
| print( "FILTERED RESPONSE:" + response) | |
| invalid_response = True | |
| return response |