[Python] Plot loads data as a standalone but not through streamlit

I am running the velocityanalysis.py file as a standalone file. when running it I get a nice plot for the forecast. I have everything set in a function. I am sending only the part of the function, not the whole code => ''' def generate_forecast(df_offerred, show_data_from, months_back, forecast_months): # Convert show_data_from to Timestamp if needed if isinstance(show_data_from, str): show_data_from = pd.to_datetime(show_data_from) # Make show_data_from timezone-aware (UTC) if show_data_from.tzinfo is None: show_data_from = show_data_from.tz_localize('UTC') else: show_data_from = show_data_from.tz_convert('UTC') now = datetime.now(timezone.utc) first_day_current_month = datetime(now.year, now.month, 1) end_of_prev_month = first_day_current_month - pd.Timedelta(days=1) start_of_range = first_day_current_month - pd.DateOffset(months=months_back)

# Parse and assign 'date' column
df_offerred['date'] = pd.to_datetime(
df_offerred['ClosedDate'].fillna(df_offerred['ChangedDate']),
utc=True, errors='coerce', infer_datetime_format=True
)

# Drop rows with NaT in 'date'
df_offerred = df_offerred.dropna(subset=['date'])

# Create 'month_start'
df_offerred['month_start'] = df_offerred['date'].dt.to_period('M').dt.to_timestamp()

# Drop NaT in 'month_start'
df_offerred = df_offerred.dropna(subset=['month_start'])

# Deduplicate
df_offerred = df_offerred.sort_values(by=["date"])
df_offerred = df_offerred.drop_duplicates(subset=['month_start', 'Id'], keep="last")

# Filter rows with date >= show_data_from
df_offerred = df_offerred.loc[df_offerred['date'] >= show_data_from]

# Offered points
points_offerred = df_offerred.groupby('month_start')['Points'].sum().reset_index()
points_offerred = points_offerred[points_offerred['month_start'] < pd.Timestamp(first_day_current_month)]
points_offerred['Points'] = points_offerred['Points'].fillna(0).astype(int)

# Handled points
df_handled = df_offerred[df_offerred["State"] == "Closed"] \
.sort_values(by="ChangedDate", ascending=False) \
.drop_duplicates(subset=["Id"], keep="first")
df_handled = df_handled.loc[df_handled['date'] >= show_data_from]

points_handled = df_handled.groupby('month_start')['Points'].sum().reset_index()
points_handled = points_handled[points_handled['month_start'] < pd.Timestamp(first_day_current_month)]
points_handled['Points'] = points_handled['Points'].fillna(0).astype(int)

# Forecast helper function
def forecast_points(points_df, forecast_months, label):
df = points_df.rename(columns={'month_start': 'ds', 'Points': 'y'})
df['ds'] = pd.to_datetime(df['ds']).dt.tz_localize(None)

if len(df) < 3:
start_date = df['ds'].min()
if pd.isna(start_date):
# fallback start date if min is NaT, e.g., first day of current month UTC
now = pd.Timestamp.now(tz='UTC').normalize()
start_date = now.replace(day=1).tz_localize(None)
future_dates = pd.date_range(start=start_date, periods=forecast_months, freq='MS')
empty_forecast = pd.DataFrame({
'month_start': future_dates,
f'forecasted_points_{label}': [0] * len(future_dates)
})
return empty_forecast

model = Prophet()
model.add_seasonality(name='monthly', period=30.5, fourier_order=5)
model.fit(df)
future = model.make_future_dataframe(periods=forecast_months, freq='MS')
forecast = model.predict(future)

forecast_result = forecast[['ds', 'yhat', 'yhat_lower', 'yhat_upper']]
forecast_result = forecast_result.rename(columns={
'ds': 'month_start',
'yhat': f'forecasted_points_{label}'
})

# Round forecast to nearest integer and convert to int
forecast_result[f'forecasted_points_{label}'] = forecast_result[f'forecasted_points_{label}'].round().astype(
int)

return forecast_result

# Get forecasts
forecast_offerred = forecast_points(points_offerred, forecast_months, "offered")
forecast_handled = forecast_points(points_handled, forecast_months, "handled")

# Merge actual and forecast
df_offerred_plot = pd.merge(points_offerred, forecast_offerred, on='month_start', how='outer').sort_values(
'month_start')
df_handled_plot = pd.merge(points_handled, forecast_handled, on='month_start', how='outer').sort_values(
'month_start')
'''


when I run the code through streamlit I get an empty plot ''' with col7: with st.container(border=True): st.subheader(" Forecast") c1, c2 = st.columns([1, 1])

# Add a date input for month/year selection (set day to 1 always)
selected_date = c1.date_input(
"Monitor Data From:",
value=datetime.date.today().replace(day=1),
min_value=datetime.date(2000, 1, 1),
max_value=datetime.date.today(),
key="forecast_date"
)

# Convert selected_date to the first day of that month as string
show_data_from = selected_date.strftime("%Y-%m-%d")

# Validate df_offerred before forecast
df_offerred = st.session_state.get("df_offerred", None)
if df_offerred is not None and df_offerred.dropna().shape[0] >= 2:
if st.session_state.loading_forecast or "chart_forecast" not in st.session_state:
with c2:
with st.spinner("Loading Forecast Chart..."):
try:
st.session_state.chart_forecast = generate_forecast(
df_offerred,
show_data_from,
24,
12
)
except ValueError as e:
st.warning(f"Forecast generation failed: {e}")
st.session_state.chart_forecast = None
st.session_state.loading_forecast = False

fig = st.session_state.chart_forecast
if fig:
for trace in fig.data:
if trace.name in ['Items Closed (Actual)', 'Story Points (Actual)']:
trace.x = pd.to_datetime(trace.x) # Ensure datetime x-axis
trace.mode = 'lines+markers'

fig.update_xaxes(type='date')
st.plotly_chart(fig, use_container_width=True, height=400)
else:
st.info("Forecast chart could not be generated due to lack of data.")
else:
st.warning("Not enough data available to generate a forecast.")
'''


i am using the same inputs in both cases and even when running it from the streamlit, the print for the forecasted values returns data

Continue reading...