features/research/unified_scoring.py

"""Unified Scoring (ADR-014 Phase 4).

Cross-model unified score = weighted average of demand/supply percentiles.
Displayed in "전체" mode when cross-model pair exists.
"""

import streamlit as st
import pandas as pd
import plotly.graph_objects as go

from core.supabase_client import (
    get_gap_scores, get_topic_clusters,
)


def render_unified_scoring(base_ctx: dict, pair: dict):
    """Render unified scoring view combining demand + supply signals.

    Args:
        base_ctx: Dashboard base context.
        pair: Cross-model pair dict from find_cross_model_pair().
    """
    campaign_chatgpt = pair["campaign_chatgpt"]
    campaign_gemini = pair["campaign_gemini"]

    matches = get_gap_scores(campaign_chatgpt, campaign_gemini)
    if not matches:
        st.info("Unified Scoring에 필요한 Cross-Model 데이터가 없습니다.")
        return

    # Fetch cluster counts for weight calculation
    chatgpt_clusters = get_topic_clusters(campaign_chatgpt, source="chatgpt")
    gemini_clusters = get_topic_clusters(campaign_gemini, source="gemini")

    chatgpt_volume = sum(c.get("fanout_count", 0) for c in chatgpt_clusters)
    gemini_volume = sum(c.get("fanout_count", 0) for c in gemini_clusters)
    total_volume = chatgpt_volume + gemini_volume

    # Weight by data volume (ADR-014 spec)
    w_chatgpt = chatgpt_volume / total_volume if total_volume > 0 else 0.5
    w_gemini = gemini_volume / total_volume if total_volume > 0 else 0.5

    st.caption(
        "Demand(ChatGPT)와 Supply(Gemini) 신호를 데이터 볼륨 비례로 통합한 점수입니다."
    )

    # Weight info
    c1, c2, c3 = st.columns(3)
    with c1:
        st.metric(
            "Demand 가중치",
            f"{w_chatgpt:.1%}",
            help=f"ChatGPT fanout 볼륨: {chatgpt_volume:,}",
        )
    with c2:
        st.metric(
            "Supply 가중치",
            f"{w_gemini:.1%}",
            help=f"Gemini citation 볼륨: {gemini_volume:,}",
        )
    with c3:
        st.metric("매칭 토픽", f"{len(matches)}개")

    st.markdown("---")

    # Compute unified scores
    scored = []
    for m in matches:
        demand_pct = float(m.get("demand_percentile") or 0)
        supply_pct = float(m.get("supply_percentile") or 0)
        unified = w_chatgpt * demand_pct + w_gemini * supply_pct
        scored.append({
            **m,
            "unified_score": unified,
        })

    # Sort by unified score DESC
    scored.sort(key=lambda x: x["unified_score"], reverse=True)

    # --- Unified Ranking Table ---
    st.markdown("#### Unified Score 랭킹")
    st.caption(
        "두 모델의 신호를 통합한 순위입니다. "
        "Unified Score가 높을수록 Demand와 Supply 모두에서 중요한 토픽입니다."
    )

    rows = []
    for i, s in enumerate(scored, 1):
        demand_pct = float(s.get("demand_percentile") or 0)
        supply_pct = float(s.get("supply_percentile") or 0)
        rows.append({
            "#": i,
            "토픽 (ChatGPT)": (s.get("chatgpt_label") or "")[:30],
            "토픽 (Gemini)": (s.get("gemini_label") or "")[:30],
            "Unified": f"{s['unified_score']:.4f}",
            "Demand": f"{demand_pct:.2%}",
            "Supply": f"{supply_pct:.2%}",
            "GapScore": f"{float(s.get('gap_score') or 0):.4f}",
            "Quadrant": s.get("quadrant", "NICHE").replace("_", " ").title(),
        })

    df = pd.DataFrame(rows)
    st.dataframe(df, use_container_width=True, hide_index=True)

    st.markdown("---")

    # --- Unified Score Distribution ---
    st.markdown("#### Unified Score vs GapScore")
    st.caption(
        "X축은 통합 중요도(높을수록 두 모델 모두 중요), "
        "Y축은 기회 크기(높을수록 콘텐츠 제작 ROI가 높음)."
    )

    _render_unified_scatter(scored)


def _render_unified_scatter(scored: list[dict]):
    """Scatter: Unified Score (x) vs GapScore (y)."""
    from .cross_model import QUADRANT_COLORS, QUADRANT_LABELS

    xs, ys, colors, hovers, sizes = [], [], [], [], []

    for s in scored:
        unified = s["unified_score"]
        gap = float(s.get("gap_score") or 0)
        quadrant = s.get("quadrant", "NICHE")
        chatgpt_label = s.get("chatgpt_label", "")

        xs.append(unified)
        ys.append(gap)
        colors.append(QUADRANT_COLORS.get(quadrant, "#9CA3AF"))
        sizes.append(max(8, min(25, unified * 30)))
        hovers.append(
            f"<b>{chatgpt_label}</b><br>"
            f"Unified: {unified:.4f}<br>"
            f"GapScore: {gap:.4f}<br>"
            f"Quadrant: {QUADRANT_LABELS.get(quadrant, quadrant)}"
        )

    fig = go.Figure()
    fig.add_trace(go.Scatter(
        x=xs,
        y=ys,
        mode="markers",
        marker=dict(
            size=sizes,
            color=colors,
            opacity=0.7,
            line=dict(width=0.5, color="#333"),
        ),
        text=hovers,
        hoverinfo="text",
        showlegend=False,
    ))

    fig.update_layout(
        title="Unified Score vs GapScore",
        xaxis_title="Unified Score (통합 중요도)",
        yaxis_title="GapScore (콘텐츠 기회)",
        height=450,
        template="plotly_white",
        hoverlabel=dict(bgcolor="white", font_size=12),
    )

    st.plotly_chart(fig, use_container_width=True, key="cross_model:unified_scatter", config={"displayModeBar": False})

    # Insight: Top-right quadrant = high importance + high opportunity
    high_unified = [s for s in scored if s["unified_score"] > 0.5]
    high_gap_and_unified = [
        s for s in high_unified
        if float(s.get("gap_score") or 0) > 0.05
    ]
    if high_gap_and_unified:
        st.info(
            f"Unified Score > 0.5 이면서 GapScore가 높은 토픽이 "
            f"**{len(high_gap_and_unified)}개** 있습니다. "
            f"이 토픽들은 두 모델 모두에서 중요하면서 콘텐츠 기회도 큰 최우선 영역입니다."
        )