from dotenv import load_dotenv import streamlit as st from langchain_experimental.agents import create_csv_agent from langchain.llms import HuggingFaceHub load_dotenv() def main(): st.set_page_config( page_title= "QnA with CSV", page_icon=":robot_face:", layout="wide" ) st.title("QnA with CSV 🤖") user_csv = st.file_uploader(label = "", type= "csv") if not user_csv: st.warning("Please Upload your CSV file 🚨") llm = HuggingFaceHub( repo_id = "mistralai/Mistral-7B-Instruct-v0.2", model_kwargs = { 'max_new_tokens': 249, 'temperature': 0.3, } ) user_input = st.text_input("Ask your Question about CSV files", disabled= not user_csv) with st.sidebar: st.subheader("Example Questions:") example_questions = [ "What is the total number of rows in the CSV?", "Can you show me the first 5 rows of the CSV?", "What are the column names in the CSV?", "How many columns does the CSV have?", "What is the data type of a specific column in the CSV?", "Can you provide a summary statistics for the numerical columns?", "Are there any missing values in the CSV?", "Can you filter the data based on a specific condition?", "What is the average value of a numerical column?", ] selected_example = st.selectbox("Select an example question:", example_questions, disabled= not user_csv) if not user_csv: st.warning("Please Upload your CSV file 🚨") if st.button("Use Selected Example"): user_input = selected_example if user_csv is not None: agent = create_csv_agent( llm = llm, path = user_csv, verbose = True, handle_parsing_errors=True ) if user_input is not None and user_input != "": st.write("User:", user_input) with st.spinner("Processing..."): response = agent.run(user_input) st.write("Bot:", response) if __name__ == "__main__": main()