Field-Monitoring / pag /monitor.py
ammarnasr's picture
small fix
56d64f4
# monitor.py
import os
import utils
import streamlit as st
import geopandas as gpd
from authentication import greeting, check_password
from senHub import SenHub
from datetime import datetime, timedelta
from sentinelhub import SHConfig
import requests
import process
from zipfile import ZipFile
import plotly.express as px
import threading
import pandas as pd
import grpc
import pb.timesfm_pb2_grpc
import pb.timesfm_pb2
from openai import OpenAI
from dotenv import load_dotenv
import time
import dotenv
load_dotenv()
def check_authentication():
if not check_password():
st.stop()
config = SHConfig()
config.instance_id = '44e79764-8b9d-43b0-a4bf-15799db2899d'
config.sh_client_id = '4ae34b53-3f81-4ba0-9c7d-b6fb0606dac3'
config.sh_client_secret = '3IPSSqE75fqK38vP85hxttR9PJEs5OxX'
config.sh_timesfm_IP = "34.121.141.161"
try:
OpenAI_key = os.getenv('OPENAI_KEY')
client = OpenAI(api_key= OpenAI_key)
except:
OpenAI_key = "sk-"
client = OpenAI(api_key= OpenAI_key)
def select_field(gdf):
st.markdown("""
<style>
.stSelectbox > div > div {cursor: pointer;}
</style>
""", unsafe_allow_html=True)
names = gdf['name'].tolist()
names.append("Select Field")
field_name = st.selectbox("Select Field", options=names, key="field_name_monitor", help="Select the field to edit", index=len(names)-1)
return field_name
def calculate_bbox(df, field):
bbox = df.loc[df['name'] == field].bounds
r = bbox.iloc[0]
return [r.minx, r.miny, r.maxx, r.maxy]
def get_available_dates_for_field(df, field, year, start_date='', end_date=''):
bbox = calculate_bbox(df, field)
token = SenHub(config).token
headers = utils.get_bearer_token_headers(token)
if start_date == '' or end_date == '':
start_date = f'{year}-01-01'
end_date = f'{year}-12-31'
data = f'{{ "collections": [ "sentinel-2-l2a" ], "datetime": "{start_date}T00:00:00Z/{end_date}T23:59:59Z", "bbox": {bbox}, "limit": 100, "distinct": "date" }}'
response = requests.post('https://services.sentinel-hub.com/api/v1/catalog/search', headers=headers, data=data)
try:
features = response.json()['features']
except:
print(response.json())
features = []
return features
@st.cache_data
def get_and_cache_available_dates(_df, field, year, start_date, end_date):
dates = get_available_dates_for_field(_df, field, year, start_date, end_date)
print(f'Caching Dates for {field}')
return dates
# def get_cuarted_df_for_field(df, field, date, metric, clientName):
# curated_date_path = utils.get_curated_location_img_path(clientName, metric, date, field)
# if curated_date_path is not None:
# curated_df = gpd.read_file(curated_date_path)
# else:
# process.Download_image_in_given_date(clientName, metric, df, field, date)
# process.mask_downladed_image(clientName, metric, df, field, date)
# process.convert_maske_image_to_geodataframe(clientName, metric, df, field, date, df.crs)
# curated_date_path = utils.get_curated_location_img_path(clientName, metric, date, field)
# curated_df = gpd.read_file(curated_date_path)
# return curated_df
def get_cuarted_df_for_field(df, field, date, metric, clientName, dates=None):
curated_date_path = utils.get_curated_location_img_path(clientName, metric, date, field)
if curated_date_path is not None:
curated_df = gpd.read_file(curated_date_path)
else:
download_date_data(df, field, [date], metric, clientName,)
curated_date_path = utils.get_curated_location_img_path(clientName, metric, date, field)
print("curr selected date processed")
if dates:
old_dates = [prev_date for prev_date in dates if prev_date != date]
download_thread = threading.Thread(target=download_date_data, name="Downloader", args=(df, field, old_dates, metric, clientName))
download_thread.start()
curated_date_path = utils.get_curated_location_img_path(clientName, metric, date, field)
curated_df = gpd.read_file(curated_date_path)
return curated_df
# def check_and_download_date_data(df, field, date, metric, clientName,):
# curated_date_path = utils.get_curated_location_img_path(clientName, metric, date, field)
# if curated_date_path is not None:
# curated_df = gpd.read_file(curated_date_path)
# else:
# process.Download_image_in_given_date(clientName, metric, df, field, date)
# process.mask_downladed_image(clientName, metric, df, field, date)
# process.convert_maske_image_to_geodataframe(clientName, metric, df, field, date, df.crs)
# curated_date_path = utils.get_curated_location_img_path(clientName, metric, date, field)
# curated_df = gpd.read_file(curated_date_path)
# return curated_df
def download_date_data(df, field, dates, metric, clientName,):
for date in dates:
process.Download_image_in_given_date(clientName, metric, df, field, date)
process.mask_downladed_image(clientName, metric, df, field, date)
process.convert_maske_image_to_geodataframe(clientName, metric, df, field, date, df.crs)
# print(f"finished downloading prev dates data")
return
def track(metric, field_name, src_df, client_name):
st.subheader(":green[Select Date and Start Monitoring]")
dates = []
date = -1
if 'dates' not in st.session_state:
st.session_state['dates'] = dates
else:
dates = st.session_state['dates']
if 'date' not in st.session_state:
st.session_state['date'] = date
else:
date = st.session_state['date']
if True:
start_date = '2024-01-01'
today = datetime.today()
end_date = today.strftime('%Y-%m-%d')
year = '2024'
dates = get_and_cache_available_dates(src_df, field_name, year, start_date, end_date)
# Add None to the end of the list to be used as a default value
#sort the dates from earliest to today
dates = sorted(dates)
#Add the dates to the session state
st.session_state['dates'] = dates
# Display the dropdown menu
if len(dates) > 0:
st.markdown("""
<style>
.stSelectbox > div > div {cursor: pointer;}
</style>
""", unsafe_allow_html=True)
dates.append(-1)
date = st.selectbox('Select Observation Date: ', dates, index=len(dates)-1, key=f'Select Date Dropdown Menu - {metric}')
if date != -1:
st.write(f'You selected: {date}')
#Add the date to the session state
st.session_state['date'] = date
else:
st.write('Please Select A Date')
else:
st.info('No dates available for the selected field and dates range, select a different range or click the button to fetch the dates again')
st.markdown('---')
st.subheader('Show Field Data')
# If a field and a date are selected, display the field data
if date != -1:
# Get the field data at the selected date
with st.spinner('Loading Field Data...'):
# Get the metric data and cloud cover data for the selected field and date, to enable background download set dates=dates
metric_data = get_cuarted_df_for_field(src_df, field_name, date, metric, client_name, dates=None)
cloud_cover_data = get_cuarted_df_for_field(src_df, field_name, date, 'CLP', client_name, dates=None)
#Merge the metric and cloud cover data on the geometry column
field_data = metric_data.merge(cloud_cover_data, on='geometry')
# Display the field data
avg_clp = field_data[f'CLP_{date}'].mean() *100
avg_metric = field_data[f'{metric}_{date}'].mean()
st.write(f'Field Data for (Field ID: {field_name}) on {date}')
col1,col3,col5,col2,col4 = st.columns(5)
col1.metric(f":orange[Average {metric}]", value=f"{avg_metric :.2f}")
col2.metric(":green[Cloud Cover]", value=f"{avg_clp :.2f}%")
#Get Avarage Cloud Cover
# If the avarage cloud cover is greater than 80%, display a warning message
if avg_clp > 80:
st.warning(f'⚠️ The Avarage Cloud Cover is {avg_clp}%')
st.info('Please Select A Different Date')
df = field_data.copy()
df['latitude'] = df['geometry'].y
df['longitude'] = df['geometry'].x
# Create a scatter plot
fig = px.scatter_mapbox(
df,
lat='latitude',
lon='longitude',
color=f'{metric}_{date}',
color_continuous_scale='RdYlGn',
range_color=(0, 1),
width= 800,
height=600,
size_max=15,
zoom=13,
)
# Add the base map
with st.expander("Show Map", expanded=False):
token = open("token.mapbox_token").read()
fig.update_layout(mapbox_style="satellite", mapbox_accesstoken=token)
st.plotly_chart(fig, use_container_width=True)
#Dwonload Links
# If the field data is not empty, display the download links
if len(field_data) > 0:
# Create two columns for the download links
download_as_shp_col, download_as_tiff_col = st.columns(2)
# Create a shapefile of the field data and add a download link
with download_as_shp_col:
#Set the shapefile name and path based on the field id, metric and date
extension = 'shp'
shapefilename = f"{field_name}_{metric}_{date}.{extension}"
path = f'./shapefiles/{field_name}/{metric}/{extension}'
# Create the target directory if it doesn't exist
os.makedirs(path, exist_ok=True)
# Save the field data as a shapefile
field_data.to_file(f'{path}/{shapefilename}')
# Create a zip file of the shapefile
files = []
for i in os.listdir(path):
if os.path.isfile(os.path.join(path,i)):
if i[0:len(shapefilename)] == shapefilename:
files.append(os.path.join(path,i))
zipFileName = f'{path}/{field_name}_{metric}_{date}.zip'
zipObj = ZipFile(zipFileName, 'w')
for file in files:
zipObj.write(file)
zipObj.close()
# Add a download link for the zip file
with open(zipFileName, 'rb') as f:
st.download_button('Download as ShapeFile', f,file_name=zipFileName)
# Get the tiff file path and create a download link
with download_as_tiff_col:
#get the tiff file path
tiff_path = utils.get_masked_location_img_path(client_name, metric, date, field_name)
# Add a download link for the tiff file
donwnload_filename = f'{metric}_{field_name}_{date}.tiff'
with open(tiff_path, 'rb') as f:
st.download_button('Download as Tiff File', f,file_name=donwnload_filename)
else:
st.info('Please Select A Field and A Date')
def monitor_fields():
st.title(":orange[Field Monitoring]")
row1,row2 = st.columns([1,2])
with row1:
current_user = greeting("Let's take a look how these fields are doing")
if os.path.exists(f"fields_{current_user}.parquet"):
gdf = gpd.read_parquet(f"fields_{current_user}.parquet")
field_name = select_field(gdf)
if field_name == "Select Field":
st.info("No Field Selected Yet!")
else:
metric = st.radio("Select Metric to Monitor", ["NDVI", "LAI", "CAB"], key="metric", index=0, help="Select the metric to monitor")
st.write(f"Monitoring {metric} for {field_name}")
with st.expander("Metrics Explanation", expanded=False):
st.write("NDVI: Normalized Difference Vegetation Index, Mainly used to monitor the health of vegetation")
st.write("LAI: Leaf Area Index, Mainly used to monitor the productivity of vegetation")
st.write("CAB: Chlorophyll Absorption in the Blue band, Mainly used to monitor the chlorophyll content in vegetation")
# st.write("NDMI: Normalized Difference Moisture Index, Mainly used to monitor the moisture content in vegetation")
else:
st.info("No Fields Added Yet!")
return
with row2:
if field_name != "Select Field":
track(metric, field_name, gdf, current_user)
if field_name != "Select Field":
st.title(":orange[Field Health Forecast]")
st.write(f"Press the button below to predict {metric} for the next 30 weeks")
# Reset session state if any of the required keys are missing
required_keys = ['api_token', 'api_token_confirmed', 'valid_until']
if any(key not in st.session_state for key in required_keys):
st.session_state['api_token'] = ''
st.session_state['api_token_confirmed'] = False
st.session_state['valid_until'] = ''
if not st.session_state['api_token_confirmed']:
st.warning("No Valid API Token Found")
with st.expander("Need a new API Token?", expanded=True):
st.markdown(utils.NEW_TOKEN_INSTRUCTIONS, unsafe_allow_html=True)
with st.expander("Token Usage history", expanded=False):
filename = f'{current_user}_tokens.csv'
if os.path.exists(filename):
token_usage = pd.read_csv(filename)
token_usage['is_expired'] = token_usage['valid_until'].apply(lambda x: 'Yes' if datetime.strptime(x, '%Y-%m-%d %H:%M:%S') < datetime.now() else 'No')
st.write(token_usage)
else:
st.write("No Token Usage History Found")
api_token = st.text_input("API Token", key="api_token_input", help="Enter the API Token From SNET")
if st.button("submit API Token", key="confirm_api_token"):
if utils.confirm_api_token(api_token)['valid']:
st.session_state['api_token'] = api_token
st.session_state['api_token_confirmed'] = True
st.session_state['valid_until'] = utils.load_token_expiration(api_token).strftime('%Y-%m-%d %H:%M:%S')
st.rerun()
else:
st.error(f"Invalid API Token; {utils.confirm_api_token(api_token)['message']}")
else:
now = datetime.now()
valid_until = datetime.strptime(st.session_state['valid_until'], '%Y-%m-%d %H:%M:%S')
time_remaining = valid_until - now
minutes_remaining = int(time_remaining.total_seconds() // 60)
seconds_remaining = int(time_remaining.total_seconds() % 60)
time_left_column, clear_token_column = st.columns([1,1])
with time_left_column:
st.success(f"API Token Confirmed. Token valid for {minutes_remaining} minutes and {seconds_remaining} seconds")
with clear_token_column:
if st.button("Clear API Token", key="clear_api_token"):
st.session_state['api_token'] = ''
st.session_state['api_token_confirmed'] = False
st.session_state['valid_until'] = ''
st.rerun()
with st.expander("Need a new API Token?", expanded=False):
st.markdown(utils.NEW_TOKEN_INSTRUCTIONS, unsafe_allow_html=True)
with st.expander("Token Usage history", expanded=False):
token_usage = utils.manage_user_tokens(current_user, st.session_state['api_token'], valid_until.strftime('%Y-%m-%d %H:%M:%S'))
token_usage['is_expired'] = token_usage['valid_until'].apply(lambda x: 'Yes' if datetime.strptime(x, '%Y-%m-%d %H:%M:%S') < datetime.now() else 'No')
st.write(token_usage)
lookback_days = st.slider("Select Lookback Days", 10, 360, 30, step=10,key="lookback_days", help="Large lookback days may take longer to load")
subcol1, subcol2, subcol3 = st.columns(3)
if subcol2.button(f'Predict & Recommend', key="predict_button", disabled=not st.session_state['api_token_confirmed']):
# start_date = '2024-01-01'
today = datetime.today()
end_date = today.strftime('%Y-%m-%d')
start_date = today - timedelta(days=lookback_days)
start_date = start_date.strftime('%Y-%m-%d')
year = '2024'
dates = get_and_cache_available_dates(gdf, field_name, year, start_date, end_date)
newest_date, oldest_date = dates[0], dates[-1]
number_of_months = (datetime.strptime(newest_date, '%Y-%m-%d') - datetime.strptime(oldest_date, '%Y-%m-%d')).days//30
my_bar = st.progress(0, text= f"Downloading Data for the last {number_of_months+1} months ...")
counter = 0
downloaded_prev_metrics = []
for index, date in enumerate(dates):
# time.sleep(0.1)
metric_data = get_cuarted_df_for_field(gdf, field_name, date, metric, current_user, dates = None)
# cloud_cover_data = get_cuarted_df_for_field(gdf, field_name, date, 'CLP', current_user, dates = None)
# field_data = metric_data.merge(cloud_cover_data, on='geometry')
avg_metric = metric_data[f'{metric}_{date}'].mean()
downloaded_prev_metrics.append((date, avg_metric))
counter = counter + 100/(len(dates))
my_bar.progress(round(counter), text=f"Downloading Data for the last {len(dates)//6} months: {round(counter)}%")
st.subheader('Predictions:')
# chart_data = pd.DataFrame(
# {
# "date": [metric[0] for metric in downloaded_prev_metrics],
# f"{metric}": [metric[1] for metric in downloaded_prev_metrics],
# }
# )
# st.area_chart(chart_data, x="date", y=f"{metric}")
channel = grpc.insecure_channel(f"{config.sh_timesfm_IP}:50051")
print("runing client request")
stub = pb.timesfm_pb2_grpc.PredictAgriStub(channel)
features = stub.predict_metric(iter([pb.timesfm_pb2.prev_values(value=metric[1], date=metric[0]) for metric in downloaded_prev_metrics]))
print("server streaming:")
predictions = []
for feature in features:
predictions.append(feature.value)
# do something with the returned output
# print(predictions)
future_dates = []
# print(dates[0])
curr_date = datetime.today()
for pred in predictions:
curr_date = curr_date + timedelta(days=7)
future_dates.append(curr_date.strftime('%Y-%m-%d'))
prev_dates = [metric[0] for metric in downloaded_prev_metrics]
history_metric_data = [metric[1] for metric in downloaded_prev_metrics]
future_metric_data = predictions
interval_dates = prev_dates
interval_dates.extend(future_dates)
history_metric_data.extend([0 for i in range(len(predictions))])
masked_future_metric_data = [0 for i in range(len([metric[1] for metric in downloaded_prev_metrics]))]
masked_future_metric_data.extend(future_metric_data)
# print(f"interval_dates:{len(interval_dates)}")
# print(f"history_metric_data:{len(history_metric_data)}")
# print(f"masked_future_metric_data:{len(masked_future_metric_data)}")
# print(predictions)
# print(interval_dates)
prediction_chart_data = pd.DataFrame(
{
f"history_{metric}_values": history_metric_data,
f"predicted_{metric}_values":masked_future_metric_data,
f"date": interval_dates,
}
)
# print(prediction_chart_data)
graph_col, recommendation_col = st.columns([1,1])
with graph_col:
st.area_chart(prediction_chart_data, x="date", y=[f"history_{metric}_values", f"predicted_{metric}_values"])
with recommendation_col:
st.subheader('Recommendation:')
with st.spinner("Generating Recommendation..."):
crop = gdf.loc[gdf['name'] == field_name].crop if 'crop' in gdf.columns else "Wheat"
try:
weeks = future_dates
gdf_loc = gdf.loc[gdf['name'] == field_name].reset_index(drop=True)
location = utils.get_region_from_coordinates(gdf_loc.geometry[0].centroid.y, gdf_loc.geometry[0].centroid.x)
prompt = f"""The Field Name is {field_name} and is located in {location}.
Analyze {crop} growth conditions for the next {len(weeks)} weeks starting from {weeks[0]} to {weeks[-1]} based on the Forecatsed {metric} values weekly.
{metric}: {predictions}
Provide a concise Short report:
1. Field Status (use format "Category: Status - One sentence comment", (e.g. Overall Health: Low - The NDVI values consistently below 0.2, indicating weak vegetative growth.)
- Overall Health:
- Growth Stage:
- Pest Risk:
- Disease Risk:
- Stress Level:
2. Yield Forecast:
[look online for the expected yield for the crop in the region based {metric} values]
3. Recommendation:
[one actionable advice reasoned based on the forecasted {metric} values, season, crop, and region]
"""
# prompt = f"given the {metric} values weekly for the next 30 weeks, comment if they are appropriate to grow {crop} (write one paragraph showing your conclusion): {metric} values:{predictions}"
response = client.chat.completions.create(
model="gpt-4o",
messages=[
{
"role": "user",
"content": prompt
}
],
temperature=1,
max_tokens=256,
top_p=1,
frequency_penalty=0,
presence_penalty=0
)
st.markdown(response.choices[0].message.content)
# save the recommendation
recommendation = response.choices[0].message.content
recommendation_filename = f'{current_user}_recommendations.md'
with open(recommendation_filename, 'a') as f:
f.write(f'\n\n## {field_name} - {datetime.now().strftime("%Y-%m-%d %H:%M:%S")}\n\n')
f.write(recommendation)
f.write('\n\n')
# Dwonload button for the recommendation
with open(recommendation_filename, 'rb') as f:
st.download_button('Download Recommendation', f,file_name=recommendation_filename)
except Exception as e:
st.code("Server Error: Could't generate recommendation!")
st.error(e)
if __name__ == '__main__':
check_authentication()
monitor_fields()