app.py

import streamlit as st
import pandas as pd
import plotly.express as px
from datetime import datetime, timedelta
from simple_salesforce import Salesforce
from transformers import pipeline
from utils import fetch_salesforce_data, detect_anomalies, generate_pdf_report
import os

# Streamlit app configuration
st.set_page_config(page_title="LabOps Dashboard", layout="wide")

# Cache Salesforce connection
@st.cache_resource
def init_salesforce():
    try:
        return Salesforce(
            username=os.getenv("SF_USERNAME", st.secrets.get("sf_username")),
            password=os.getenv("SF_PASSWORD", st.secrets.get("sf_password")),
            security_token=os.getenv("SF_SECURITY_TOKEN", st.secrets.get("sf_security_token"))
        )
    except Exception as e:
        st.error(f"Failed to connect to Salesforce: {e}")
        return None

# Cache Hugging Face model
@st.cache_resource
def init_anomaly_detector():
    try:
        return pipeline(
            "text-classification",
            model="distilbert-base-uncased",
            tokenizer="distilbert-base-uncased",
            clean_up_tokenization_spaces=True
        )
    except Exception as e:
        st.error(f"Failed to initialize anomaly detector: {e}")
        return None

# Initialize connections
sf = init_salesforce()
anomaly_detector = init_anomaly_detector()

# Cache data fetching
@st.cache_data(ttl=10)  # Cache for 10 seconds to meet refresh requirement
def get_filtered_data(lab_site, equipment_type, date_start, date_end):
    query = f"""
        SELECT Equipment__c, Log_Timestamp__c, Status__c, Usage_Count__c, Lab__c, Equipment_Type__c
        FROM SmartLog__c
        WHERE Log_Timestamp__c >= {date_start.strftime('%Y-%m-%d')}
        AND Log_Timestamp__c <= {date_end.strftime('%Y-%m-%d')}
    """
    if lab_site != "All":
        query += f" AND Lab__c = '{lab_site}'"
    if equipment_type != "All":
        query += f" AND Equipment_Type__c = '{equipment_type}'"
    query += " LIMIT 100"  # Reduced for faster load in Hugging Face Spaces
    return fetch_salesforce_data(sf, query)

def main():
    if sf is None or anomaly_detector is None:
        st.error("Cannot proceed without Salesforce connection or anomaly detector.")
        return

    st.title("Multi-Device LabOps Dashboard")
    
    # Filters
    col1, col2 = st.columns(2)
    with col1:
        lab_site = st.selectbox("Select Lab Site", ["All", "Lab1", "Lab2", "Lab3"])
    with col2:
        equipment_type = st.selectbox("Equipment Type", ["All", "Cell Analyzer", "Weight Log", "UV Verification"])
    
    date_range = st.date_input("Date Range", [datetime.now() - timedelta(days=7), datetime.now()])
    
    # Validate date range
    if len(date_range) != 2:
        st.warning("Please select a valid date range.")
        return
    date_start, date_end = date_range

    # Fetch and process data
    with st.spinner("Fetching data..."):
        data = get_filtered_data(lab_site, equipment_type, date_start, date_end)
    if not data:
        st.warning("No data available for the selected filters.")
        return
    
    df = pd.DataFrame(data)
    df["Log_Timestamp__c"] = pd.to_datetime(df["Log_Timestamp__c"])
    df["Anomaly"] = df.apply(
        lambda row: detect_anomalies(f"{row['Status__c']} Usage:{row['Usage_Count__c']}", anomaly_detector),
        axis=1
    )

    # Pagination
    page_size = 10
    total_pages = max(1, len(df) // page_size + (1 if len(df) % page_size else 0))
    page = st.number_input("Page", min_value=1, max_value=total_pages, value=1, step=1)
    start_idx = (page - 1) * page_size
    end_idx = start_idx + page_size
    paginated_df = df[start_idx:end_idx]

    # Device Cards
    st.subheader("Device Status")
    for _, row in paginated_df.iterrows():
        anomaly = "⚠️ Anomaly" if row["Anomaly"] == "POSITIVE" else "✅ Normal"
        st.markdown(f"""
            **{row['Equipment__c']}** | Lab: {row['Lab__c']} | Health: {row['Status__c']} | 
            Usage: {row['Usage_Count__c']} | Last Log: {row['Log_Timestamp__c'].strftime('%Y-%m-%d %H:%M:%S')} | {anomaly}
        """)
    
    # Usage Chart
    st.subheader("Usage Trends")
    fig = px.line(
        df, 
        x="Log_Timestamp__c", 
        y="Usage_Count__c", 
        color="Equipment__c", 
        title="Daily Usage Trends",
        labels={"Log_Timestamp__c": "Timestamp", "Usage_Count__c": "Usage Count"}
    )
    fig.update_layout(xaxis_title="Timestamp", yaxis_title="Usage Count")
    st.plotly_chart(fig, use_container_width=True)

    # Downtime Chart
    st.subheader("Downtime Patterns")
    downtime_df = df[df["Status__c"] == "Down"]
    if not downtime_df.empty:
        fig_downtime = px.histogram(
            downtime_df, 
            x="Log_Timestamp__c", 
            color="Equipment__c", 
            title="Downtime Patterns",
            labels={"Log_Timestamp__c": "Timestamp"}
        )
        fig_downtime.update_layout(xaxis_title="Timestamp", yaxis_title="Downtime Count")
        st.plotly_chart(fig_downtime, use_container_width=True)
    else:
        st.info("No downtime events found for the selected filters.")

    # AMC Reminders
    st.subheader("AMC Reminders")
    amc_query = "SELECT Equipment__c, AMC_Expiry_Date__c FROM Equipment__c WHERE AMC_Expiry_Date__c <= NEXT_N_DAYS:14"
    amc_data = fetch_salesforce_data(sf, amc_query, retries=3)
    if amc_data:
        for record in amc_data:
            st.write(f"Equipment {record['Equipment__c']} - AMC Expiry: {record['AMC_Expiry_Date__c']}")
    else:
        st.info("No AMC expiries within the next 14 days.")

    # Export PDF
    if st.button("Export PDF Report"):
        with st.spinner("Generating PDF..."):
            try:
                pdf_file = generate_pdf_report(df, lab_site, equipment_type, [date_start, date_end])
                with open(pdf_file, "rb") as f:
                    st.download_button("Download PDF", f, file_name="LabOps_Report.pdf", mime="application/pdf")
            except Exception as e:
                st.error(f"Failed to generate PDF: {e}")

if __name__ == "__main__":
    main()