Handle multiple outputs?

[llermaly]

I'm using pandasAI with Streamlit, and trying to handle multiple outpus. I had to override the _format_results method to get the charts working.

Now I would like to display the chart + the DF but I'm having no luck.

My custom format method:

    def _format_results(self, result: dict):
        if result["type"] == "plot":
            import matplotlib.pyplot as plt
            import matplotlib.image as mpimg

            # Load the image file
            image = mpimg.imread(result["value"])

            # Display the image
            plt.imshow(image)
            plt.axis("off")
            st.pyplot(plt.gcf(), use_container_width=True)
            plt.close("all")
        else:
            return super()._format_results(result)

The output :

    return { "type": "dataframe", "value": top_3_people_df.to_html() }, { "type": "plot", "value": 'exports/charts/temp_chart.png' }

The error:

Traceback (most recent call last):
  File "/opt/homebrew/anaconda3/lib/python3.11/site-packages/streamlit/runtime/scriptrunner/script_runner.py", line 552, in _run_script
    exec(code, module.__dict__)
  File "/Users/gllermaly/Documents/projects/llamaindex-streamlit/pages/11_Pandas_AI.py", line 117, in <module>
    run_app()
  File "/Users/gllermaly/Documents/projects/llamaindex-streamlit/pages/11_Pandas_AI.py", line 108, in run_app
    res = st.session_state["pandas_ai"].chat(question)
          ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/opt/homebrew/anaconda3/lib/python3.11/site-packages/pandasai/smart_datalake/__init__.py", line 361, in chat
    return self._format_results(result)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/Users/gllermaly/Documents/projects/llamaindex-streamlit/pages/11_Pandas_AI.py", line 17, in _format_results
    if result["type"] == "plot":
       ~~~~~~^^^^^^^^
TypeError: 'NoneType' object is not subscriptable

Thanks!

[lwdnxu]

I have the same problem, is there a solution?

[gdcsinaptik]

Hey @llermaly and @lwdnxu currently PandasAI does not handle multiple concurrent outputs. Depending on the user query, it identifies the most relevant output type, generate the corresponding code, and then outputs it.