PandasAI with Gemini API

[pgvetrivel]

Hi Team,

following is the code I have written to connect pandasai with gemini

import os
from dotenv import load_dotenv
import pandas as pd
from pandasai import Agent
from pandasai import SmartDatalake
import streamlit as st
from pandasai.responses.streamlit_response import StreamlitResponse
import google.generativeai as genai

initialize chat history in streamlit session state

if "chat_history" not in st.session_state:
st.session_state.chat_history = []

display chat history

for message in st.session_state.chat_history:
with st.chat_message(message["role"]):
st.markdown(message["content"])

input field for user's message

user_prompt = st.chat_input("Ask MIS Agent...")

genai.configure(api_key=os.getenv("API_KEY"))
google_llm = genai.GenerativeModel("gemini-1.5-flash")

if user_prompt:
# add user's message to chat history and display it
st.chat_message("user").markdown(user_prompt)
st.session_state.chat_history.append({"role":"user","content": user_prompt})

agent = Agent(sales_by_partner, config={"llm": google_llm})
assistant_response = agent.chat(user_prompt)

st.session_state.chat_history.append({"role":"assistant", "content": assistant_response})

# display pandasai response
with st.chat_message("assistant"):
    assistant_response

When I ask a question, it throws up a following error

ValidationError: 1 validation error for Config llm instance of LLM expected (type=type_error.arbitrary_type; expected_arbitrary_type=LLM)

Please help me.

[dosubot[bot]]

To resolve the "ValidationError: 1 validation error for Config llm instance of LLM expected" error, you need to ensure that the llm attribute in your Config object is an instance of a class that inherits from LLM or LangchainLLM. Since you are using the Gemini API, you should wrap it in a custom class that implements the required methods for the LLM interface.

Here's an example of how you might define a custom LLM class for the Gemini API:

from pandasai.llm import LLM

class GeminiLLM(LLM):
    def __init__(self, api_key):
        self.api_key = api_key
        # Initialize any other necessary attributes

    def call(self, instruction, context=None):
        # Implement the logic to call the Gemini API and return the response
        pass

# Usage
google_llm = GeminiLLM(api_key=os.getenv("API_KEY"))
agent = Agent(sales_by_partner, config={"llm": google_llm})

Make sure to implement the call method in your GeminiLLM class, as it is required for any class inheriting from LLM ^[1][2].

To continue talking to Dosu, mention @dosu.