dashboard_phase7.py

"""
dashboard_phase7.py  (Phase 7e)
===============================
Interactive ESG-tilted optimiser dashboard.

Run:  streamlit run dashboard_phase7.py
(Pulls live SX5E data via the BL data layer on first load; needs network.)

Standalone by design -- keeps the Phases 1-6 dashboard.py untouched.
Controls:
  - Reporting year   -> moves the 1.5C pathway target (-7% p.a.)
  - Carbon target    -> the binding decarbonisation constraint
  - Per-name cap     -> UCITS-style concentration limit
"""
import numpy as np
import pandas as pd
import streamlit as st

from phase7.esg_data import build_universe
from phase7.esg_optimiser import (min_tracking_error, run_scenarios,
                                  cost_of_green_frontier, benchmark_waci)
from sfdr2.alignment import pathway_target
from sfdr2.exclusions import screen_portfolio

st.set_page_config(page_title="SFDR 2.0 ESG-Tilted Optimiser", layout="wide")
st.title("ESG-Tilted Minimum-Tracking-Error Optimiser")
st.caption("Phase 7 - closes the SFDR engine's decarbonisation residual by "
           "tilting the SX5E benchmark to the 1.5C pathway under a tracking-"
           "error budget. Returns and covariance from the Black-Litterman "
           "data layer (real, Ledoit-Wolf shrunk).")


@st.cache_data(show_spinner="Pulling SX5E data via the BL data layer...")
def load_bundle(years: int):
    b = build_universe(years=years)
    # cache-friendly: return plain objects
    return {
        "tickers": b["tickers"], "names": b["names"],
        "Sigma": b["Sigma"], "w_bench": b["w_bench"],
        "holdings": b["holdings"], "r_f": b["r_f"],
    }


bundle = load_bundle(3)
tickers, holdings = bundle["tickers"], bundle["holdings"]
parent_waci = benchmark_waci(holdings, tickers, bundle["w_bench"].values)

# --- sidebar ---------------------------------------------------------------
st.sidebar.header("Mandate parameters")
year = st.sidebar.slider("Reporting year", 2024, 2035, 2026, 1)
tgt_default = pathway_target(parent_waci, year)
cap = st.sidebar.slider("Per-name cap (UCITS-style)", 0.05, 1.0, 0.12, 0.01,
                        help="Concentration limit. Lower caps force the tilt "
                             "across more names and raise tracking error.")
carbon_target = st.sidebar.slider(
    "Carbon-intensity target (tCO2e/EURm)",
    min_value=float(round(0.5 * tgt_default, 1)),
    max_value=float(round(parent_waci, 1)),
    value=float(round(tgt_default, 1)), step=0.5,
    help=f"1.5C pathway target this year = {tgt_default:.1f}. "
         "Lower = deeper decarbonisation = higher tracking error.")

st.sidebar.markdown("---")
st.sidebar.caption("ESG attributes are sector-archetype proxies (see README), "
                   "not audited issuer figures. Market data is real.")

# --- headline optimisation -------------------------------------------------
r = min_tracking_error(bundle["Sigma"], bundle["w_bench"].values, holdings,
                       tickers, carbon_target=carbon_target, floor=None,
                       max_weight=cap, year=year)

c1, c2, c3, c4 = st.columns(4)
c1.metric("Tracking error", f"{r['te_bps']:.0f} bps",
          help="Annualised active risk vs the SX5E benchmark")
c2.metric("Achieved WACI", f"{r['waci']:.1f}",
          delta=f"{r['waci'] - parent_waci:+.1f} vs parent",
          delta_color="inverse")
c3.metric("1.5C pathway target", f"{tgt_default:.1f}",
          delta="aligned" if r["waci"] <= tgt_default + 1e-6 else "above",
          delta_color="off")
c4.metric("Active share", f"{r['active_share']*100:.0f}%")

if not r["success"]:
    st.error("Optimisation infeasible at this carbon target and name cap. "
             "Raise the target or the cap.")

st.divider()

# --- cost-of-green frontier ------------------------------------------------
left, right = st.columns([1.4, 1])
with left:
    st.subheader("The cost of green")
    st.caption("Active risk vs decarbonisation. Convexity means each extra "
               "tonne removed costs more than the last.")
    df = cost_of_green_frontier(bundle, year=year, n_points=25, floor=None,
                                max_weight=cap)
    feas = df[df.feasible].copy()
    chart_df = feas.set_index("intensity_cut_pct")[["tracking_error_bps"]]
    chart_df.columns = ["Tracking error (bps)"]
    st.line_chart(chart_df, height=360)
    pathway_cut = 100.0 * (parent_waci - tgt_default) / parent_waci
    st.caption(f"1.5C pathway = {pathway_cut:.0f}% intensity cut vs parent "
               f"benchmark (WACI {parent_waci:.1f} -> {tgt_default:.1f}).")

with right:
    st.subheader("Compliance-path cost")
    st.caption("Tracking-error cost of each SFDR 2.0 qualifying route, at the "
               "1.5C pathway target.")
    sc = run_scenarios(bundle, year=year, max_weight=cap)
    rows = [{"Path": v["label"], "TE (bps)": round(v["te_bps"], 0),
             "WACI": round(v["waci"], 1),
             "Qual wt %": round(v["qualifying_weight"] * 100, 0),
             "Feasible": "Yes" if v["success"] else "No"}
            for v in sc["results"].values()]
    st.dataframe(pd.DataFrame(rows), hide_index=True, use_container_width=True)
    scr = screen_portfolio(holdings)["summary"]
    st.caption(f"On this benchmark, {scr['n_excluded']} names are PAB-excluded "
               f"({scr['weight_excluded_pct']:.0f}% of weight); qualifying "
               "weight stays above 70%, so the three paths coincide here.")

st.divider()

# --- resulting weights -----------------------------------------------------
st.subheader("Resulting tilt vs benchmark")
wb = bundle["w_bench"].values
active = r["weights"] - wb
wt_df = pd.DataFrame({
    "Name": bundle["names"],
    "Benchmark %": (wb * 100).round(2),
    "Optimised %": (r["weights"] * 100).round(2),
    "Active %": (active * 100).round(2),
    "Carbon intensity": [h["scope12_intensity_tco2_per_eur_m"] or np.nan
                         for h in holdings],
}).sort_values("Active %")
st.dataframe(wt_df, hide_index=True, use_container_width=True)
st.caption("The optimiser cuts high-carbon names (negative active weight) and "
           "overweights low-carbon ones, minimising active risk subject to the "
           "carbon constraint.")