from langchain.embeddings import SentenceTransformerEmbeddings import google.generativeai as genai import streamlit as st import os from langchain.vectorstores import Chroma import pickle from transformers import AutoTokenizer # Load tokenizer and ensure padding token is set tokenizer = AutoTokenizer.from_pretrained("bert-base-uncased") if tokenizer.pad_token is None: tokenizer.pad_token = tokenizer.eos_token or tokenizer.unk_token # Configure API key GOOGLE_API_KEY = os.getenv('Google_API_KEY') os.environ['Google_API_KEY'] = GOOGLE_API_KEY genai.configure(api_key=os.environ['Google_API_KEY']) # Load sentence transformer model model = SentenceTransformerEmbeddings(model_name="all-MiniLM-L6-v2") def setup(): embedding = pickle.load(open('embedding.pkl', 'rb')) document = pickle.load(open('document.pkl', 'rb')) db = Chroma.from_documents(document, embedding) return db def find_match(input_text): db = setup() matching_doc = db.similarity_search_with_score(input_text, k=2) return matching_doc[0][0].page_content + matching_doc[1][0].page_content def query_refiner(conversation, query): generation_config = { "temperature": 1, "top_p": 0.95, "top_k": 64, "max_output_tokens": 8192, } model = genai.GenerativeModel( model_name="gemini-1.5-flash", generation_config=generation_config, system_instruction=f"Your name is Sam, given the following user query and conversation log, formulate a question that would be the most relevant to provide the user with an answer from a knowledge base.\n\nCONVERSATION LOG: \n{conversation}\n\nQuery: {query}\n\nRefined Query:", ) chat_session = model.start_chat(history=[]) response = chat_session.send_message(f"Conversation Log:{conversation} Query:{query}") return response.text def get_conversation_string(): conversation_string = "" for i in range(len(st.session_state['responses']) - 1): conversation_string += "Human: " + st.session_state['requests'][i] + "\n" conversation_string += "Bot: " + st.session_state['responses'][i + 1] + "\n" return conversation_string