app.py

# app.py — MyPlan Telco Recommender (V10 Final Hotfix)
import json
from pathlib import Path
import pandas as pd
import streamlit as st
from agents import Coordinator

# ---- Data path (robust) ----
_DEFAULT_NAME = "enhanced_multi_user_telecom_data.csv"
DEFAULT_DATA = (Path(__file__).parent / _DEFAULT_NAME).resolve()

def resolve_data_path() -> Path:
    p = DEFAULT_DATA
    if p.exists():
        return p
    alt = Path(__file__).parent / "data" / _DEFAULT_NAME
    if alt.exists():
        return alt.resolve()
    return p  # non-existent; caller handles error

# ---- CSS: force black text on light background ----
def inject_light_theme():
    css = [
        "<style>",
        ":root{--bg:#ffffff;--panel:#ffffff;--panel2:#fafafa;--panel3:#f4f6f8;--text:#111111;--muted:#444444;--accent:#1a73e8;}",
        "html,body,.stApp,[data-testid='stAppViewContainer'],[data-testid='stHeader']"
        "{background:var(--bg)!important;color:var(--text)!important;font-family:'Inter','Segoe UI',system-ui,-apple-system,sans-serif;}",
        "section[data-testid='stSidebar']{background:var(--panel2)!important;color:var(--text)!important;border-right:1px solid #e6e6e6;}",
        "*,.stMarkdown p,.stMarkdown span,.stMarkdown li,h1,h2,h3,h4,h5,h6,label{color:var(--text)!important;}",
        "a{color:var(--accent)!important;}",
        ".stSelectbox [role='combobox']{background:#ffffff!important;color:#111111!important;border:1px solid #d0d5dd!important;border-radius:8px;}",
        ".stSelectbox [role='combobox'] *{color:#111111!important;}",
        ".stSelectbox input{color:#111111!important;background:#ffffff!important;}",
        ".stSelectbox svg{color:#111111!important;fill:#111111!important;}",
        ".stSelectbox [data-baseweb='popover'],.stSelectbox [data-baseweb='menu']{background:#ffffff!important;color:#111111!important;border:1px solid #d0d5dd!important;}",
        ".stSelectbox [data-baseweb='option']{background:#ffffff!important;color:#111111!important;}",
        ".stSelectbox [data-baseweb='option']:hover{background:#f2f4f7!important;color:#111111!important;}",
        "div[data-baseweb='input'] input, textarea, .stNumberInput input{background:#ffffff!important;color:#111111!important;border:1px solid #d0d5dd!important;}",
        ".stButton>button, .stDownloadButton>button{background:#ffffff!important;color:#111111!important;border:1px solid #d0d5dd!important;border-radius:12px;padding:.6rem 1rem;font-weight:600;}",
        "[role='tablist']{border-bottom:1px solid #e6e6e6;}",
        "[role='tab']{background:#ffffff!important;color:#111111!important;border:1px solid #d0d5dd!important;border-bottom:none!important;margin-right:.25rem;border-top-left-radius:8px;border-top-right-radius:8px;}",
        "[role='tab'][aria-selected='true']{background:#f9fafb!important;color:#111111!important;}",
        ".card,.card-big{background:#ffffff!important;border:1px solid #e6e6e6;border-radius:14px;padding:1rem;color:#111111!important;}",
        ".card-title{color:#333333!important;font-weight:600;font-size:.9rem;text-transform:uppercase;letter-spacing:.02em;}",
        "code, pre{background:#f7f7f9!important;color:#111111!important;border:1px solid #e6e6e6!important;}",
        "hr{border-color:#e6e6e6;}",
        "</style>",
    ]
    st.markdown("\n".join(css), unsafe_allow_html=True)

def md_to_html(md: str) -> str:
    return md

st.set_page_config(page_title="MyPlan Telco Recommender — V10 Final", layout="wide")
inject_light_theme()

st.title("📶 MyPlan Telco Recommender — V10 Final")

# Sidebar controls
with st.sidebar:
    st.subheader("Controls")
    provider = st.selectbox("LLM Provider", ["Local Stub", "OpenAI", "Gemini", "Claude"], index=0, key="prov")
    model = st.selectbox("Model / Variant", ["stub-v1", "gpt-4o-mini", "gemini-1.5", "claude-3-haiku"], index=0, key="model")
    st.caption("Text is forced to BLACK to ensure visibility.")

# Inputs
c1, c2, c3, c4 = st.columns(4)
with c1:
    voice = st.number_input("Monthly Voice Minutes", min_value=0, max_value=10000, value=700, step=10)
with c2:
    data_gb = st.number_input("Monthly Data (GB)", min_value=0, max_value=1000, value=22, step=1)
with c3:
    sms = st.number_input("Monthly SMS", min_value=0, max_value=5000, value=40, step=5)
with c4:
    roam = st.number_input("Monthly Roaming Minutes", min_value=0, max_value=2000, value=0, step=5)

c5, c6, c7 = st.columns(3)
with c5:
    intl = st.number_input("Monthly International Minutes", min_value=0, max_value=2000, value=0, step=5)
with c6:
    bill_amt = st.number_input("Average Monthly Bill (₹)", min_value=0, max_value=100000, value=599, step=10)
with c7:
    disc = st.number_input("Avg. Discount/Waiver (₹)", min_value=0, max_value=100000, value=0, step=10)

late = st.number_input("Avg. Late Fee (₹)", min_value=0, max_value=100000, value=0, step=10)

st.markdown("---")
st.subheader("Get Recommendation")

user_inputs = {
    "voice_minutes": voice,
    "data_gb": data_gb,
    "sms_count": sms,
    "roaming_minutes": roam,
    "intl_minutes": intl,
    "bill_amount": bill_amt,
    "discount_amount": disc,
    "late_fee": late,
}

@st.cache_resource(show_spinner=False)
def load_coordinator(prov: str, m: str):
    data_path = resolve_data_path()
    if not data_path.exists():
        st.error(f"Data file not found: '{_DEFAULT_NAME}'. Place it beside app.py or in ./data/.")
        st.stop()
    df = pd.read_csv(str(data_path))
    coord = Coordinator(provider=prov, model=m)
    coord.fit(df)
    return coord

if st.button("Recommend Plan", type="primary"):
    try:
        coord = load_coordinator(provider, model)
    except Exception as e:
        st.exception(e)
        st.stop()

    try:
        rec = coord.recommend(user_inputs)
    except Exception as e:
        st.exception(e)
        st.stop()

    left, right = st.columns([1,1])
    with left:
        st.markdown("<div class='card'><div class='card-title'>Recommended Plan</div>", unsafe_allow_html=True)
        label = getattr(rec, "plan_label", None) or getattr(rec, "label", None) or (rec.get("plan_label") if isinstance(rec, dict) else rec.get("label", ""))
        st.metric("Plan", label if label is not None else "")
        conf = getattr(rec, "confidence", None) or (rec.get("confidence") if isinstance(rec, dict) else None)
        if conf is not None:
            st.metric("Recommendation Strength", f"{float(conf)*100:.1f}%")
        est = getattr(rec, "est_monthly_cost", None) or getattr(rec, "est_cost", None) or (rec.get("est_monthly_cost") if isinstance(rec, dict) else rec.get("est_cost", None))
        fair = getattr(rec, "fair_cost_range", None) or getattr(rec, "fair_range", None) or (rec.get("fair_cost_range") if isinstance(rec, dict) else rec.get("fair_range", None))
        if est is not None:
            if isinstance(fair, (list, tuple)) and len(fair)==2:
                st.write(f"Estimated Monthly Cost: **₹{est:.0f}** (fair range: ₹{fair[0]:.0f}–₹{fair[1]:.0f})")
            else:
                st.write(f"Estimated Monthly Cost: **₹{est:.0f}**")
        st.markdown("</div>", unsafe_allow_html=True)

    with right:
        narrative = getattr(rec, "narrative", None) or (rec.get("narrative") if isinstance(rec, dict) else "")
        st.markdown("<div class='card'><div class='card-title'>Persona & Why this plan</div>"
                    f"<div style='padding-top:.4rem'>{md_to_html(narrative)}</div></div>", unsafe_allow_html=True)

    # Similar subscribers with monthly averages
    try:
        # Safe retrieval without boolean-evaluating a DataFrame
        peers = None
        if hasattr(rec, 'peers'):
            peers = rec.peers
        elif isinstance(rec, dict):
            peers = rec.get('peers')
        if isinstance(peers, pd.DataFrame) and not peers.empty:
            st.markdown("#### Similar Subscribers (usage view · top 10)")
            cols = ["user_id","current_plan","bill_amount","voice_minutes","data_gb","sms_count","roaming_minutes","intl_minutes"]
            show_cols = [c for c in cols if c in peers.columns]
            view = peers[show_cols].head(10).copy()
            # enrich with dataset averages
            try:
                df = pd.read_csv(str(resolve_data_path()), parse_dates=["date"])
                df["month"] = df["date"].dt.to_period("M")
                monthly = df.groupby(["user_id","month"]).agg(
                    voice_minutes_used=("voice_minutes_used","sum"),
                    data_mb_used=("data_mb_used","sum"),
                    sms_count_used=("sms_count_used","sum"),
                    roaming_voice_minutes=("roaming_voice_minutes","sum"),
                    international_voice_minutes=("international_voice_minutes","sum"),
                ).reset_index()
                avg = monthly.groupby("user_id").mean(numeric_only=True).reset_index()
                if "data_mb_used" in avg.columns:
                    avg["data_mb_used"] = avg["data_mb_used"]/1024.0
                avg = avg.rename(columns={
                    "voice_minutes_used":"avg_voice_min",
                    "data_mb_used":"avg_data_gb",
                    "sms_count_used":"avg_sms",
                    "roaming_voice_minutes":"avg_roaming_min",
                    "international_voice_minutes":"avg_intl_min",
                })
                view = view.merge(avg, on="user_id", how="left")
            except Exception as e:
                st.info(f"Averaged usage enrichment skipped: {e}")
            st.dataframe(view, use_container_width=True)
            csv_bytes = view.to_csv(index=False).encode("utf-8")
            st.download_button("Download similar subscribers (CSV)", csv_bytes, file_name="similar_subscribers.csv", mime="text/csv")
    except Exception as e:
        st.info(f"Peers section skipped: {e}")

    with st.expander("📊 Stats & Rationale", expanded=False):
        numeric = getattr(rec, "numeric_rationale", None) or (rec.get("numeric_rationale") if isinstance(rec, dict) else {})
        features = {}
        if isinstance(numeric, dict):
            features = numeric.get("features", {})
        payload = {"features": features, "confidence": conf}
        st.code(json.dumps(payload, indent=2), language="json")
else:
    st.info("Adjust inputs and click **Recommend Plan** to see suggestions.")