llm.py

# Server class for the project

# from langchain.llms import OpenLLM
import os
import openai
from langchain.embeddings.openai import OpenAIEmbeddings
import pinecone
from langchain.vectorstores import Pinecone
from langchain.llms import OpenAI
from langchain.text_splitter import RecursiveCharacterTextSplitter
from langchain.document_loaders import DirectoryLoader
from langchain.prompts import (
    ChatPromptTemplate,
    MessagesPlaceholder,
    SystemMessagePromptTemplate,
    HumanMessagePromptTemplate
)
from langchain.chains import ConversationChain
from langchain.chat_models import ChatOpenAI
from langchain.memory import ConversationSummaryMemory
from google.cloud import translate_v2 as translate



with open("openai_api_key.txt", "r") as f:
    api_key = f.read()

os.environ["OPENAI_API_KEY"] = api_key


class Server:
    "server class for the project"
    defalut_params = {
        "pine-cone_key": "da26b242-4825-43e9-b5dd-6a94047d2860",
        "pine-cone_env": "northamerica-northeast1-gcp",
        "index_name": "index-1"
    }

    def __init__(self):
        self.model_name: str = "gpt-3.5-turbo"
        self.model_id: str = "google/flan-t5-large"
        self.temperature: float = 0.94
        self.repetition_penalty: float = 1.2
        self.directory: str = "docs/"
        self.docs = None
        self.index = None
        self.chain = None
        self.chunk_size: int = 1000
        self.chunk_overlap: int = 100
        self.memory = ConversationSummaryMemory(memory_key='chat_history', return_messages=True,llm=OpenAI(temperature=0))

    # @staticmethod
    def load_docs(self):
        loader = DirectoryLoader(self.directory)
        self.docs = loader.load()

    # @staticmethod
    def split_docs(self):
        self.load_docs()
        text_splitter = RecursiveCharacterTextSplitter(chunk_size=self.chunk_size, chunk_overlap=self.chunk_overlap)
        self.docs = text_splitter.split_documents(self.docs)

    # @staticmethod
    def load_pinecone(self):
        pinecone.init(
            api_key=self.defalut_params["pine-cone_key"],  # find at app.pinecone.io
            environment=self.defalut_params["pine-cone_env"]  # next to api key in console
        )
        #   self.split_docs()
        embeddings = OpenAIEmbeddings()
        self.index = Pinecone.from_existing_index(self.defalut_params["index_name"], embeddings,
                                                  namespace="")
        print(self.index)

    # @staticmethod
    def get_similiar_docs(self, query, k=2, score=False):
        try:
            if len(query)!=0:
                if score:
                    similar_docs = self.index.similarity_search_with_score(query, k=k)
                else:
                    similar_docs = self.index.similarity_search(query, k=k)
                return similar_docs
        except  ValueError:
            return None
        
            

    def get_answer(self, query):
        similar_docs = self.get_similiar_docs(query)
        prompt = ChatPromptTemplate.from_messages([
            SystemMessagePromptTemplate.from_template(
               """"
             You are a question bot named Kutty.

                ###Instruction###

                You should provide information only about creating knowledge questions.

            Prompt: Generate Questions Using Bloom's Taxonomy

                Learning Objectives:
                    1. Explain the concept of photosynthesis.
                    2. Solve quadratic equations.
                    3. Analyze the causes of World War II.
                    4. Evaluate the impact of climate change.
                    5.Explain the process of mitosis and its importance in cell division.
                    6.Analyze the economic factors that contribute to inflation.
                    7.Evaluate the ethical implications of genetic engineering.
                    8.Describe the major functions of the human nervous system.
                    9.Apply the principles of supply and demand to explain market equilibrium.
                    10.Compare and contrast renewable and non-renewable sources of energy.
                    11.Analyze the impact of social media on interpersonal communication.
                    12.Explain the concept of cultural relativism in anthropology.
                    13.Evaluate the effectiveness of various methods for studying and retaining information.
                    14.Apply the scientific method to design and conduct an experiment about plant growth.

                    Using Bloom's Taxonomy, generate questions for each learning objective that span different cognitive levels. Provide questions for each of the following Bloom's levels:

                    1. Remember (Knowledge) :"list", "define", "recite"
                    2. Understand (Comprehension):"explain", "describe", "summarize"
                    3. Apply (Application):"solve", "use", "apply",
                    4. Analyze (Analysis):"analyze", "compare", "contrast",
                    5. Evaluate (Evaluation):"evaluate", "judge", "assess",
                    6. Create (Synthesis): "create", "design", "compose"


                Your task is to generate appropriate questions for each learning objective and Bloom's level following a similar pattern.

                describe only when the user asks.

                You should always be polite and respond briefly.

                If you don't know the answer,don't try to make one up.

               """
            ),
            MessagesPlaceholder(variable_name="chat_history"),
            HumanMessagePromptTemplate.from_template("{input}")
        ])
        llm = ChatOpenAI(temperature=0)
        conversation = ConversationChain(memory=self.memory, prompt=prompt, llm=llm)
        text = conversation.predict(input = str(similar_docs) +" "+ query)
        return text