Charts visualization on Streamlit web app

[marcello-calabrese]

Hi, is there a better way to visualize charts through pandasai on Streamlit? I will explain, so to show the charts on streamlit when I ask to plot, I used matplotlib.use('TkAgg'), it works that basically show the chart in a new window but the request keeps running.

It might be a better way to render matplotlib through pandasai on a streamlit. Hope so.

Thanks a lot,

Marcello

[gventuri]

Hi @marcello-calabrese! Might be necessary to instruct the model to use block=False when plotting the chart. This should fix the issue hopefully!
I'll look into it further to figure it out, thanks a lot for reporting.

P.s. if you have any alternative approach in mind, feel free to share!

[marcello-calabrese]

Ciao @gventuri (immagino sei italiano 😊), thanks for your quick response. I will try to use block=false otherwise not sure if I create an input_text query extra for charts, and use st.py plot, will try

[gventuri]

Ciao @marcello-calabrese (yes, I'm from Italy 😊)! You should actually instruct the model to use block=false as it plots the chart for the time being, so try to mention it in the prompt. We will soon work on a patch to have it done automatically!

Let me know if it works!

[marcello-calabrese]

CIao @gventuri , I tried with middleware following the guide you mentioned above, of course after updating pandasai in the environment, but it gives me the error saying streamlitmiddleware does not exist. Let me know if you would like me to add the error screenshot. Thanks,

[gventuri]

@marcello-calabrese don't forget to import it:

from pandasai import PandasAI
from pandasai.middlewares.streamlit import StreamlitMiddleware

pandas_ai = PandasAI(llm, middlewares=[StreamlitMiddleware()])

Let me know if it works for you!

[marcello-calabrese]

Hi @gventuri , thanks i imported the Streanlit Middleware as you suggested above, but it doesnt create any chart, it only respond as below (see the screenshot);

[gventuri]

@marcello-calabrese thanks a lot for sharing. Would you mind sharing the code you are using so that I can try to reproduce it?

[gventuri]

@marcello-calabrese just released v0.5.5 that should fix this error!

[marcello-calabrese]

@marcello-calabrese thanks a lot for sharing. Would you mind sharing the code you are using so that I can try to reproduce it?

Hi, see the code below thanks. the code below is not with the version updated to 0.5.5, will try soon, thanks

instantiate LLM with PandasAI

llm = OpenAI(api_token='sk-XXXXXXXXXXXXXXXXXXXXXXXXXXXXXX', engine='gpt-3.5-turbo')

pandas_ai = PandasAI(llm, middlewares=[StreamlitMiddleware()], verbose=False)

if st.button('Generate Response'):
if question:
with st.spinner('Generating response, please wait...'):
# get the response from the chatbot
#pandas_ai.run(df,question, is_conversational_answer=True)
# show the response
st.write(pandas_ai.run(df,question, is_conversational_answer=True))

[gventuri]

I expect the code to work with 0.5.5, as I'm successfully running something very similar! Feel free to give it a try and let me know :)

[laodev1]

Just wanted to confirm that this does work in > 0.5.5

[marcello-calabrese]

Hi @gventuri , sorry for my late reply, I job took all of my time these days :-( and dint let me do what I love: data analytics and data science.

I can confirm that with the 0.5.5 version it works perfectly Thanks a lot GRAZIE :-)

[gventuri]

@laodev1 @marcello-calabrese thanks a lot for letting me know 🚀