svijayanand/Podcast_Oracle

1from helpers.utils import create_or_load_vectore_store2from helpers.import_envs import openai_api_key3from langchain_openai import ChatOpenAI4from langchain.schema import StrOutputParser5from langchain_core.runnables.passthrough import RunnablePassthrough6from langchain.prompts import ChatPromptTemplate7from helpers.model_utils import set_question_answer_llm8 9def answer_question(question, transcript_file_name, llm_choice=None):10    question_answer_llm = set_question_answer_llm(llm_choice)11 12    # Specify the path to the file you want to check13    vector_store = create_or_load_vectore_store(transcript_file_name=transcript_file_name)14 15    # create a prompt template to send to our LLM that will incorporate the documents from our retriever with the16    # question we ask the chat model17    prompt_template = ChatPromptTemplate.from_template(18        "Answer the {question} based on the following {context}."19    )20 21    # create a retriever for our documents22    retriever = vector_store.as_retriever()23 24    # create a parser to parse the output of our LLM25    parser = StrOutputParser()26 27    # 💻 Create the sequence (recipe)28    runnable_chain = (29        # TODO: How do we chain the output of our retriever, prompt, model and model output parser so that we can get a good answer to our query?30        {"context": retriever, "question": RunnablePassthrough()}31        | prompt_template32        | question_answer_llm33        | StrOutputParser()34    )35 36    answer = runnable_chain.invoke(question)37    print(answer)38    return answer39