Select Slider and plotly_events (custom-componet)

I am having a hard time understanding how to order my streamlit app. There is a select_slider bar that is used to filter a dataframe and a plotly_event which displays a chart from the filtered dataframe. I want to link the select_slider and plotly_event together but don’t know how to structure my app.

plotly_event is an interactive object that allows me to click a point on the line chart and record its position within the chart. I am storing information from the clicks using st.session_state.

When I move the slider, I would like to reload plotly_event such that there is no cached information from the previous click event on the previous chart. Right now I am stuck because I cannot make the chart until I have the filtered dataframe. I have looked at the select_slider documentation and on_change seems close to what I want.

import pandas as pd
import numpy as np
import streamlit as st
import as px
from streamlit.state.session_state import SessionState
from streamlit_plotly_events import plotly_events

def df_filter(df):
    sel_slider = st.select_slider("Choose a water year", 
                                  options = pd.date_range('1894-09-30','2020-09-30', freq = 'A-SEP'),
    start_date = sel_slider + pd.DateOffset(years=0)
    end_date = sel_slider + pd.DateOffset(years=1)
    filtered_df = df.loc[(> start_date) & ( <= end_date), :]
    filtered_df.flow = filtered_df.flow.astype(float)
    return filtered_df, end_date.year

if __name__ == '__main__':
    st.session_state.count = 0
    df = pd.DataFrame(data = {'date':pd.date_range('1983-09-30', '2020-08-01', freq='D'),
                              'flow':np.random.random(pd.date_range('1983-09-30', '2020-08-01', freq='D').shape)})    
    filtered_df, end_date = df_filter(df)

    fig = px.line(data_frame = filtered_df, x = 'date', y = 'flow')
    plot_name_holder = st.empty()
    clickedPoint = plotly_events(fig, key=f"line{st.session_state.count}")
    if len(clickedPoint) == 1:
        st.session_state[str(end_date)] = clickedPoint[0]['x']
        plot_name_holder.write(f"Clicked Point: {end_date}, {clickedPoint[0]['x']}")
        st.session_state.count +=1