src/streamlit_app.py

import streamlit as st
import pandas as pd
import numpy as np
import plotly.express as px
from datetime import datetime, timedelta
from dateutil.parser import parse

# ---------------------------
# App Config and Theming
# ---------------------------
st.set_page_config(
    page_title="Procurement Agent – S/4HANA Embedded Analytics (Demo)",
    page_icon="🧭",
    layout="wide",
    initial_sidebar_state="expanded",
)

# Subtle CSS polish for a premium feel
st.markdown(
    """
    <style>
    .kpi-card {
        padding: 14px 16px; border-radius: 12px; background: #0a0a0a0d;
        border: 1px solid #e6e6e6; box-shadow: 0 1px 2px rgba(0,0,0,0.04);
    }
    .metric-label { font-size: 12px; color: #666; margin-bottom: 6px; }
    .metric-value { font-size: 26px; font-weight: 700; }
    .metric-sub { font-size: 12px; color: #999; }
    .stChatFloatingInputContainer { border-top: 1px solid #eee; }
    .st-emotion-cache-1avcm0n { padding-top: 0 !important; }
    .rounded-img { border-radius: 50%; }
    </style>
    """,
    unsafe_allow_html=True,
)

# ---------------------------
# Data Loading (Synthetic “CDS-like”)
# ---------------------------
@st.cache_data
def load_data():
    df = pd.read_csv("data/synthetic_procurement.csv", parse_dates=["PO_Date","DeliveryDate","GR_Date","IR_Date"])
    # Derived fields similar to embedded analytics
    df["DaysToDeliver"] = (df["DeliveryDate"] - df["PO_Date"]).dt.days
    df["IsOpen"] = df["Status"].eq("Open")
    return df

df = load_data()

# ---------------------------
# Sidebar Filters
# ---------------------------
with st.sidebar:
    st.image("https://huggingface.co/front/assets/huggingface_logo-noborder.svg", width=120)
    st.title("Procurement Agent")
    st.caption("S/4HANA Embedded Analytics – Learning Demo (Synthetic data)")

    # Time filter
    max_date = df["PO_Date"].max()
    default_start = max_date - timedelta(days=45)
    date_range = st.date_input("PO Date Range", (default_start, max_date))

    # Codelists
    company = st.multiselect("Company Code", sorted(df["CompanyCode"].unique().tolist()))
    plants = st.multiselect("Plant", sorted(df["Plant"].unique().tolist()))
    mat_groups = st.multiselect("Material Group", sorted(df["MaterialGroup"].unique().tolist()))
    suppliers = st.multiselect("Supplier", sorted(df["Supplier"].unique().tolist()))
    buyers = st.multiselect("Buyer", sorted(df["Buyer"].unique().tolist()))
    status_sel = st.multiselect("Status", sorted(df["Status"].unique().tolist()))

    st.markdown("---")
    st.subheader("Demo actions")
    if st.button("Reset Filters"):
        st.session_state.clear()
        st.rerun()

# Apply filters
def apply_filters(df):
    dff = df.copy()
    if isinstance(date_range, tuple) and len(date_range) == 2:
        start_date, end_date = pd.to_datetime(date_range[0]), pd.to_datetime(date_range[1])
        dff = dff[(dff["PO_Date"] >= start_date) & (dff["PO_Date"] <= end_date)]
    if company:
        dff = dff[dff["CompanyCode"].isin(company)]
    if plants:
        dff = dff[dff["Plant"].isin(plants)]
    if mat_groups:
        dff = dff[dff["MaterialGroup"].isin(mat_groups)]
    if suppliers:
        dff = dff[dff["Supplier"].isin(suppliers)]
    if buyers:
        dff = dff[dff["Buyer"].isin(buyers)]
    if status_sel:
        dff = dff[dff["Status"].isin(status_sel)]
    return dff

fdf = apply_filters(df)

# ---------------------------
# KPI Header
# ---------------------------
def kpi_card(label, value, sub=""):
    st.markdown(
        f"""
        <div class="kpi-card">
            <div class="metric-label">{label}</div>
            <div class="metric-value">{value}</div>
            <div class="metric-sub">{sub}</div>
        </div>
        """,
        unsafe_allow_html=True,
    )

col1, col2, col3, col4 = st.columns(4)
with col1:
    total_po = fdf["PO_ID"].nunique()
    kpi_card("Purchase Orders", f"{total_po:,}", "Unique POs in selection")
with col2:
    spend = fdf["NetValue"].sum()
    kpi_card("Net Spend", f"${spend:,.0f}", "Sum of PO item values")
with col3:
    avg_lt = fdf["LeadTimeDays"].mean() if len(fdf) else 0
    kpi_card("Avg Lead Time", f"{avg_lt:.1f}d", "Supplier cycle time")
with col4:
    otif = fdf["OTIF"].mean() * 100 if len(fdf) else 0
    kpi_card("OTIF", f"{otif:.0f}%", "On-time in-full rate")

st.markdown("")

# ---------------------------
# Tabs: Overview | Supplier Insights | Explorer | Simulations
# ---------------------------
tab1, tab2, tab3, tab4 = st.tabs(["Overview", "Supplier Insights", "Explorer", "Simulations"])

with tab1:
    c1, c2 = st.columns([1.3, 1])
    with c1:
        st.subheader("Spend by Supplier")
        if len(fdf):
            fig = px.bar(
                fdf.groupby("Supplier", as_index=False)["NetValue"].sum().sort_values("NetValue", ascending=False),
                x="Supplier", y="NetValue", color="Supplier", height=380, template="plotly_white",
                hover_data={"NetValue":":,.0f"}
            )
            st.plotly_chart(fig, use_container_width=True)
        else:
            st.info("No data for selected filters.")

        st.subheader("Material Group Mix")
        if len(fdf):
            fig2 = px.pie(
                fdf, names="MaterialGroup", values="NetValue", hole=0.45, template="plotly_white",
                height=380
            )
            st.plotly_chart(fig2, use_container_width=True)

    with c2:
        st.subheader("Lead Time by Supplier")
        if len(fdf):
            g = fdf.groupby("Supplier", as_index=False)["LeadTimeDays"].mean().sort_values("LeadTimeDays")
            fig3 = px.bar(g, x="LeadTimeDays", y="Supplier", orientation="h", height=380, template="plotly_white")
            st.plotly_chart(fig3, use_container_width=True)

        st.subheader("OTIF by Supplier")
        if len(fdf):
            g2 = fdf.groupby("Supplier", as_index=False)["OTIF"].mean()
            g2["OTIF%"] = (g2["OTIF"] * 100).round(1)
            fig4 = px.scatter(g2, x="Supplier", y="OTIF%", size="OTIF%", color="Supplier", height=340, template="plotly_white")
            st.plotly_chart(fig4, use_container_width=True)

with tab2:
    st.subheader("Supplier Scorecard")
    sup = st.selectbox("Choose supplier", sorted(df["Supplier"].unique().tolist()))
    sdf = fdf[fdf["Supplier"] == sup]
    if len(sdf):
        c1, c2, c3 = st.columns(3)
        with c1:
            kpi_card("Spend", f"${sdf['NetValue'].sum():,.0f}")
        with c2:
            kpi_card("Avg Price", f"${sdf['NetPrice'].mean():.2f}/unit")
        with c3:
            kpi_card("OTIF", f"{(sdf['OTIF'].mean()*100):.0f}%")

        st.markdown("")

        c4, c5 = st.columns(2)
        with c4:
            st.caption("Lead time trend (by PO date)")
            trend = sdf.sort_values("PO_Date").groupby("PO_Date", as_index=False)["LeadTimeDays"].mean()
            fig5 = px.line(trend, x="PO_Date", y="LeadTimeDays", markers=True, template="plotly_white", height=340)
            st.plotly_chart(fig5, use_container_width=True)
        with c5:
            st.caption("Price distribution")
            fig6 = px.histogram(sdf, x="NetPrice", nbins=10, template="plotly_white", height=340)
            st.plotly_chart(fig6, use_container_width=True)

        st.subheader("Recent PO Lines")
        st.dataframe(
            sdf.sort_values("PO_Date", ascending=False)[
                ["PO_ID","PO_Item","PO_Date","Material","Quantity","OrderUnit","NetPrice","NetValue","DeliveryDate","Status","LeadTimeDays","OTIF"]
            ].head(10),
            use_container_width=True, height=300
        )
    else:
        st.info("No lines for selected supplier within current filters.")

with tab3:
    st.subheader("Interactive Explorer")
    dims = ["CompanyCode","Plant","MaterialGroup","Supplier","Buyer","Status"]
    sel_dim = st.selectbox("Dimension", dims, index=3)
    sel_mea = st.selectbox("Measure", ["NetValue","Quantity","NetPrice","LeadTimeDays","OTIF"], index=0)

    if len(fdf):
        g = fdf.groupby(sel_dim, as_index=False)[sel_mea].mean() if sel_mea in ["NetPrice","LeadTimeDays","OTIF"] else \
            fdf.groupby(sel_dim, as_index=False)[sel_mea].sum()
        fig7 = px.bar(g.sort_values(sel_mea, ascending=False).head(15), x=sel_dim, y=sel_mea, color=sel_dim, template="plotly_white", height=420)
        st.plotly_chart(fig7, use_container_width=True)
        st.dataframe(g.sort_values(sel_mea, ascending=False), use_container_width=True, height=260)
    else:
        st.info("Adjust filters to see data.")

with tab4:
    st.subheader("What-if: Payment Terms and Delivery Delays")

    # Simple simulation: change payment terms and hypothetical delay impact on OTIF
    term_delta = st.slider("Payment term change (days)", -30, 30, 0, step=5)
    delay_rate = st.slider("Simulate delivery delay rate (%)", 0, 50, 10, step=5)

    def run_sim(df_in, term_delta, delay_rate):
        sim = df_in.copy()
        # Adjust payment days
        sim["PaymentDaysSim"] = sim["PaymentDays"] + term_delta
        # Apply simple OTIF penalty based on delay rate
        penalty = delay_rate / 100.0
        sim["OTIF_Sim"] = np.clip(sim["OTIF"] * (1 - penalty) + (1 - sim["OTIF"]) * (1 - penalty/2), 0, 1)
        # Assume carrying cost impact: +0.02% per extra payment day on spend
        delta_days = np.maximum(sim["PaymentDaysSim"] - sim["PaymentDays"], 0)
        sim["CarryingCostAdj"] = sim["NetValue"] * (0.0002 * delta_days)
        return sim

    if len(fdf):
        simdf = run_sim(fdf, term_delta, delay_rate)
        c1, c2, c3 = st.columns(3)
        with c1:
            kpi_card("OTIF (Simulated)", f"{(simdf['OTIF_Sim'].mean()*100):.0f}%")
        with c2:
            kpi_card("PaymentDays Δ", f"{term_delta:+d}d")
        with c3:
            kpi_card("Carrying Cost Adj", f"${simdf['CarryingCostAdj'].sum():,.0f}")

        st.caption("Supplier-level OTIF change")
        g = simdf.groupby("Supplier", as_index=False)[["OTIF","OTIF_Sim"]].mean()
        g["OTIF"] = (g["OTIF"]*100).round(1)
        g["OTIF_Sim"] = (g["OTIF_Sim"]*100).round(1)
        fig8 = px.bar(g.melt(id_vars="Supplier", value_vars=["OTIF","OTIF_Sim"], var_name="Metric", value_name="OTIF%"), x="Supplier", y="OTIF%", color="Metric", barmode="group", template="plotly_white", height=400)
        st.plotly_chart(fig8, use_container_width=True)
        st.dataframe(g.sort_values("OTIF_Sim", ascending=False), use_container_width=True, height=260)
    else:
        st.info("No data to simulate. Adjust filters.")

# ---------------------------
# Agentic Chat (Demo)
# ---------------------------
st.markdown("---")
st.subheader("Agent Assistant")
st.caption("Ask procurement questions, e.g., “Top suppliers by OTIF this month,” “Compare ACME vs GLOBAL_MFG on price and lead time,” “Show spend by RM group last 30 days.”")

if "messages" not in st.session_state:
    st.session_state.messages = [
        {"role": "assistant", "content": "Hello! I can analyze procurement data, compute KPIs, and run what‑if simulations. What would you like to see?"}
    ]

# Simple tool functions (CDS-like queries)
def tool_top_suppliers_by(metric="OTIF", topn=5):
    if not len(fdf): return "No data in current selection."
    g = fdf.groupby("Supplier", as_index=False)[metric].mean()
    if metric != "OTIF":
        # For value metrics that make sense as sum (e.g., NetValue)
        if metric in ["NetValue","Quantity"]:
            g = fdf.groupby("Supplier", as_index=False)[metric].sum()
    g = g.sort_values(metric, ascending=False).head(topn)
    return g

def tool_compare_suppliers(sup_a, sup_b):
    sub = fdf[fdf["Supplier"].isin([sup_a, sup_b])]
    if not len(sub): return "No data for those suppliers in current selection."
    stats = sub.groupby("Supplier").agg(
        Spend=("NetValue","sum"),
        AvgPrice=("NetPrice","mean"),
        AvgLead=("LeadTimeDays","mean"),
        OTIF=("OTIF","mean")
    ).reset_index()
    stats["OTIF%"] = (stats["OTIF"]*100).round(1)
    return stats

def tool_spend_by_dim(dim="MaterialGroup"):
    if not len(fdf): return None
    g = fdf.groupby(dim, as_index=False)["NetValue"].sum().sort_values("NetValue", ascending=False)
    return g

def tool_show_recent_po_lines(n=10):
    if not len(fdf): return None
    cols = ["PO_ID","PO_Item","PO_Date","Supplier","Material","Quantity","OrderUnit","NetPrice","NetValue","DeliveryDate","Status","LeadTimeDays","OTIF"]
    return fdf.sort_values("PO_Date", ascending=False)[cols].head(n)

# Heuristic “planner” to route user intents to tools
def agent_router(prompt: str):
    p = prompt.lower().strip()
    # pattern routes
    if "top" in p and "supplier" in p and "otif" in p:
        n = 5
        for tok in p.split():
            if tok.isdigit():
                n = int(tok)
                break
        return ("top_suppliers_otif", {"topn": n})
    if "compare" in p and "supplier" in p:
        # naive extract A vs B
        tokens = p.replace("compare","").replace("supplier","").replace("suppliers","").replace(" vs "," ").split()
        # heuristic: choose two known supplier names intersected
        known = set(df["Supplier"].unique().tolist())
        picks = [t for t in tokens if t.upper() in known]
        if len(picks) >= 2:
            return ("compare_suppliers", {"a": picks[0].upper(), "b": picks[1].upper()})
        return ("compare_suppliers", {"a": "ACME_SUPPLY", "b": "GLOBAL_MFG"})
    if "spend" in p and ("material group" in p or "group" in p):
        return ("spend_by_dim", {"dim": "MaterialGroup"})
    if "recent" in p or ("last" in p and "po" in p):
        return ("recent_pos", {"n": 10})
    if "lead time" in p and "supplier" in p:
        return ("lead_by_supplier", {})
    if "price" in p and "supplier" in p:
        return ("price_by_supplier", {})
    # default: summary
    return ("summary", {})

def agent_execute(route, args):
    if route == "top_suppliers_otif":
        g = tool_top_suppliers_by("OTIF", topn=args.get("topn",5))
        if isinstance(g, str):
            return g, None
        g2 = g.copy()
        g2["OTIF%"] = (g2["OTIF"]*100).round(1)
        fig = px.bar(g2, x="Supplier", y="OTIF%", color="Supplier", template="plotly_white", height=360)
        return "Top suppliers by OTIF:", fig

    if route == "compare_suppliers":
        stats = tool_compare_suppliers(args.get("a"), args.get("b"))
        if isinstance(stats, str):
            return stats, None
        fig = px.bar(stats, x="Supplier", y=["Spend","AvgPrice","AvgLead","OTIF"], barmode="group", template="plotly_white", height=420)
        return "Comparison across Spend, AvgPrice, AvgLead, and OTIF:", fig

    if route == "spend_by_dim":
        g = tool_spend_by_dim(args.get("dim","MaterialGroup"))
        if g is None:
            return "No data for spend by dimension.", None
        fig = px.treemap(g, path=[args.get("dim","MaterialGroup")], values="NetValue", height=420)
        return f"Spend by {args.get('dim','MaterialGroup')}:", fig

    if route == "recent_pos":
        lines = tool_show_recent_po_lines(args.get("n",10))
        if lines is None:
            return "No recent PO lines found.", None
        st.dataframe(lines, use_container_width=True, height=260)
        return f"Showing {len(lines)} most recent PO lines.", None

    if route == "lead_by_supplier":
        if not len(fdf): return "No data.", None
        g = fdf.groupby("Supplier", as_index=False)["LeadTimeDays"].mean()
        fig = px.bar(g.sort_values("LeadTimeDays"), x="LeadTimeDays", y="Supplier", orientation="h", template="plotly_white", height=420)
        return "Average lead time by supplier:", fig

    if route == "price_by_supplier":
        if not len(fdf): return "No data.", None
        g = fdf.groupby("Supplier", as_index=False)["NetPrice"].mean()
        fig = px.bar(g.sort_values("NetPrice", ascending=False), x="Supplier", y="NetPrice", template="plotly_white", height=420)
        return "Average price by supplier:", fig

    # summary
    msg = f"In current selection: {fdf['PO_ID'].nunique()} POs, spend ${fdf['NetValue'].sum():,.0f}, avg lead time {fdf['LeadTimeDays'].mean():.1f}d, OTIF {(fdf['OTIF'].mean()*100):.0f}%."
    return msg, None

# Render chat history
for m in st.session_state.messages:
    with st.chat_message(m["role"], avatar="🧭" if m["role"]=="assistant" else "🧑🏻"):
        st.write(m["content"])

prompt = st.chat_input("Ask about procurement performance, suppliers, KPIs, or simulations…")
if prompt:
    st.session_state.messages.append({"role": "user", "content": prompt})
    with st.chat_message("user", avatar="🧑🏻"):
        st.write(prompt)

    with st.chat_message("assistant", avatar="🧭"):
        with st.status("Thinking…", expanded=False):
            route, args = agent_router(prompt)
            text, fig = agent_execute(route, args)

        if text:
            st.write(text)
        if fig is not None:
            st.plotly_chart(fig, use_container_width=True)
        st.session_state.messages.append({"role": "assistant", "content": text or "(shown as chart/table)"})