app.py

import streamlit as st
import pandas as pd
import yaml
import json
import s3fs

import datetime

from src.data import label_gee
from src.report.plot_predicted_values import plot_predicted_values, get_last_prediction_path

st.title("Cyanobacteria Monitoring - Guaíba Lake")

fs = s3fs.S3FileSystem()

with open("params.yaml") as config_file:
    config = yaml.safe_load(config_file)

@st.cache_data(ttl=datetime.timedelta(hours=24))
def load_data(config, _fs):
    last_pred_path = get_last_prediction_path(config, fs)

    predicted_values = pd.read_parquet(_fs.open(last_pred_path))
    _, ciano = label_gee.load_data(config, _fs)

    return predicted_values, ciano

predicted_values, ciano = load_data(config, fs)
last_value = predicted_values.tail(1)

st.header("Last predicted value")
st.markdown(
    f"""
    **{round(last_value['y_pred'].values[0])} cells mL-1 on {last_value['date'].values[0]} ({pd.to_datetime(last_value['date']).dt.day_name().values[0]}).**
    """
)

fig = plot_predicted_values(predicted_values, ciano, config)
st.plotly_chart(fig, use_container_width=True)

#csv = convert_df(predicted_values)

#st.download_button(
#    label="Download predictions as CSV",
#    data=csv,
#    file_name="predictions.csv",
#    mime="text/csv",
#)

st.header("About the project")

mae_file = open(config["evaluate"]["metrics_file"])
mae = json.load(mae_file)

st.markdown(
    f"""
A machine learning model to predict the cyanobacteria concentration at Guaíba Lake
using Sentinel 2A data.

Mean absolute error of {round(mae["mae"])} cells/mL
"""
)