Spaces:
Runtime error
Runtime error
| from langchain.prompts.prompt import PromptTemplate | |
| from langchain.chains import ConversationalRetrievalChain | |
| from langchain.chat_models import ChatOpenAI | |
| # from langchain.llms import OpenAI | |
| # from langchain.chains.conversational_retrieval.prompts import CONDENSE_QUESTION_PROMPT,QA_PROMPT | |
| # _template = """Given the following conversation and a follow up question, rephrase the follow up question to be a | |
| # standalone question. | |
| # | |
| # | |
| # Chat History: | |
| # {chat_history} | |
| # Follow Up Input: {question} | |
| # Standalone question:""" | |
| # CONDENSE_QUESTION_PROMPT = PromptTemplate.from_template(_template) | |
| # | |
| # template = """You are given an context and a question find an answer from the given | |
| # context and provide an answer. If the answer is not in the context then | |
| # simply say "No information found". Don't try to make an answer. And also don't use "According to the information | |
| # provided in the given links," Question: {question} ========= {context} ========= Answer in Markdown: """ | |
| # QA_PROMPT = PromptTemplate(template=template, input_variables=["question", "context"]) | |
| _template = """Given the following conversation and a follow up question, rephrase the follow up question to be a | |
| standalone question. | |
| Chat History: | |
| {chat_history} | |
| Follow Up Input: {question} | |
| Standalone question:""" | |
| CONDENSE_QUESTION_PROMPT = PromptTemplate.from_template(_template) | |
| prompt_template = """Give an answer to the question based on the context below and try to explain in detail and do | |
| add source links if exist in context, and if you don't find the answer in the context, then simply say "That's a great | |
| question, and I want to make sure I provide you with the most accurate and helpful information possible. | |
| Unfortunately, our current pool of insights does not have an answer to this.", don't try to make up an answer And | |
| also don't use "According to the information provided in the given links, " while giving an answer. | |
| {context} | |
| Question: {question} | |
| Helpful Answer:""" | |
| QA_PROMPT = PromptTemplate( | |
| template=prompt_template, input_variables=["context", "question"] | |
| ) | |
| def get_chain(vectorstore): | |
| from langchain.chains.llm import LLMChain | |
| # from langchain.callbacks.streaming_stdout import StreamingStdOutCallbackHandler | |
| from langchain.chains.question_answering import load_qa_chain | |
| llm = ChatOpenAI(temperature=0) | |
| streaming_llm = ChatOpenAI(streaming=True, temperature=0, max_tokens=300) | |
| question_generator = LLMChain(llm=llm, prompt=CONDENSE_QUESTION_PROMPT) | |
| doc_chain = load_qa_chain(streaming_llm, chain_type="stuff", prompt=QA_PROMPT) | |
| # qa_chain = ConversationalRetrievalChain.from_llm( | |
| # llm, | |
| # vectorstore.as_retriever(search_kwargs={"k": 8, "include_metadata": True}), | |
| # condense_question_prompt=CONDENSE_QUESTION_PROMPT, | |
| # qa_prompt=QA_PROMPT | |
| # ) | |
| qa_chain = ConversationalRetrievalChain(return_source_documents=True, | |
| retriever=vectorstore.as_retriever( | |
| search_kwargs={"k": 8, "include_metadata": True}), | |
| combine_docs_chain=doc_chain, question_generator=question_generator) | |
| return qa_chain | |