vahid-ahmadi · Nov 3, 2026 · Nov 3, 2026 · Nov 3, 2026 · Nov 3, 2026 · Nov 3, 2026
diff --git a/CHANGELOG.md b/CHANGELOG.md
@@ -8,6 +8,26 @@ The version below tracks the `policyengine-macro` package in `integration/`.
 `CITATION.cff` and `integration/pyproject.toml` must agree with the topmost
 heading here; `tests/test_repo_hygiene.py` asserts that they do.
 
+## [Unreleased]
+
+### Site redesign stack (#200)
+
+Bullets by PR, in merge order. #226 and #229 wait on the decision in #218; #231 and #232 sit on #206. Drop a PR's bullet here if it has not merged into #200 when #200 merges.
+
+- **#200: site redesign.** Every page uses the Autumn Budget 2026 layout (header band, title band, sub-tabs, cards with expanders), with `DESIGN.md` as the spec. URLs, numbers and data are unchanged.
+- **#204: question router on `/models` and a new `/glossary`.** Model cards filter by question and country; `/connect` gives copy-paste setup and `/contact` lists self-serve routes. No numbers changed.
+- **#206: model pages.** One-column methodology, one overview order, and model names instead of ids.
+- **#206: FRB/US comparators.** The worked-shock reference moves from the 2014 FEDS Note (gap −0.4pp, core PCE −0.08pp) to Laforte (2018) for the current model (gap −0.486pp, core PCE −0.0345pp, unemployment +0.219pp at q8); multiplier ranges move to CBO's converted ranges (for example Coenen 0.7–1.0 → CBO 0.45–1.82).
+- **#205: forecast-evaluation scoreboard.** `/forecasts` leads with the record (band coverage, mean absolute error beside the no-change benchmark, forecasts scored), the archive is flat, and charts carry units and sources. `forecasts/scorecard.json` is unchanged.
+- **#219: `/models/compare`.** One table per question (six questions × five models) with the exact call, a sourced benchmark and a verdict. The FRB/US benchmark is Laforte (2018) (band −0.53…−0.40 → −0.486…−0.459).
+- **#216: integration-layer scenario sweep fixes.** `score_reform(model="obr")` places PolicyEngine-UK costings on fiscal-year quarters (default window `2026Q1..2030Q4` → `2026Q2..2031Q1`); `svar_forecast` returns simple y/y rates rather than log differences; `frbus_shock` reports rates as level deviations; `hank_shock` returns the `tax` series; invalid inputs (NaN sizes, unknown question types, out-of-window years) are refused before solving.
+- **#226: hosted OBR pin** `1ed742c` → `e4d7faa` (obr-macroeconomic-model main). After a redeploy, hosted Bank Rate +100bp ΔGDP q4/q8/q12 moves −446/−525/−490 → −431/−482/−416 £m, matching the site, and the default `score_reform` cumulative ΔGDP for a basic-rate rise moves +£5.6bn → −£18.3bn.
+- **#229: weekly drift check.** `integration/drift/canaries.json` compares one cheap hosted call per model with the number the site quotes for it and reports value, provenance and revision mismatches.
+- **#231: BoE SVAR replication figures** are labelled with their run vintage (boe-var-model `845e2a4`) and their `results/` links are pinned to it. No number changes.
+- **#232: BoE SVAR figures updated to the 2026Q2 data refresh** (boe-var-model #24, `d4ec3e7`): accepted draws 744 → 765, ESS 356.1 → 368.0, global FEVD share of UK GDP 37.4 → 38.7 and of UK CPI 42.3 → 41.5, rolling origins 49 → 50, smallest adjusted q 0.36 → 0.39 (published spec 0.40 → 0.42), fast-config GDP/CPI 42.0/37.6 → 40.1/40.1.
+
+<!-- Site redesign stack (#200) entries go above this line; data pipeline (#221, #228) entries go below it. Keeping them apart lets both stacks merge in either order. -->
+
 ## [0.2.0] — 2026-08-11
 
 Access and verification release. Nothing about any model's behaviour or any

diff --git a/DESIGN.md b/DESIGN.md
diff --git a/README.md b/README.md
@@ -30,7 +30,7 @@ verbatim — that registry is authoritative and this table is a copy of it.
 | 02 | **`obr-macro`** — OBR macroeconometric emulator | validated for selected scenarios | [PolicyEngine/obr-macroeconomic-model](https://github.com/PolicyEngine/obr-macroeconomic-model) |
 | 03 | **`boe-svar`** — Bank of England structural VAR replication | validated replication for selected outputs | [PolicyEngine/boe-var-model](https://github.com/PolicyEngine/boe-var-model) |
 | 04 | **`frb-us`** — Federal Reserve FRB-US implementation | validated software replication with scope limits | [PolicyEngine/us-frb-model](https://github.com/PolicyEngine/us-frb-model) |
-| 05 | **`us-hank`** — US two-asset HANK (Auclert–Bardóczy–Rognlie–Straub 2021) | validated replication for hosted stylized-shock experiments; VAR-free sequence-space HANK; not a forecaster; distributional outputs are first-order approximations | [PolicyEngine/us-hank-model](https://github.com/PolicyEngine/us-hank-model) |
+| 05 | **`us-hank`** — US two-asset HANK (Auclert–Bardóczy–Rognlie–Straub 2021) | steady state replicates the paper (six of seven solved parameters); shock signs are tested, magnitudes are not validated; not a forecaster; distributional outputs are steady-state statistics and an MPC-proportional split of the aggregate | [PolicyEngine/us-hank-model](https://github.com/PolicyEngine/us-hank-model) |
 | 06 | **`psl-og`** — OG-UK overlapping generations model | research prototype; calibrated counterfactual | [PSLmodels/OG-UK](https://github.com/PSLmodels/OG-UK) |
 | 07 | **`define-uk`** — DEFINE-UK ecological stock-flow consistent model | experimental; partial replication — baseline macro block replicates manual Table 4; scenario deltas gated on the pinned oracle run, the published scenario definitions, and two paper anchors (no numeric v1.1 scenario results are published); unlicensed upstream is never hosted, so hosted calls return run instructions | [PolicyEngine/define-uk-model](https://github.com/PolicyEngine/define-uk-model) (upstream [DEFINE-model/DEFINE_UK_1.1](https://github.com/DEFINE-model/DEFINE_UK_1.1)) |
 | — | More model classes (incl. OG-USA) | planned | — |

diff --git a/assets/make_current_outlooks.py b/assets/make_current_outlooks.py
@@ -6,9 +6,12 @@
 import csv
 import json
 import re
+import sys
 from pathlib import Path
 
 ROOT = Path(__file__).resolve().parents[1]
+sys.path.insert(0, str(ROOT))
+import chartkit  # noqa: E402  (stdlib-only, repo root)
 
 
 def points(values, xs, ymap):
@@ -28,40 +31,42 @@ def obr_chart() -> str:
     def ymap(v):
         return bot - (v - 99) / 11 * (bot - top)
 
+    top_v = ymap(110)
     out = [
-        '<figure class="vfig current-outlook">',
-        '<svg class="vchart" viewBox="0 0 800 326" role="img" aria-labelledby="obr-current-t obr-current-d">',
+        f'<svg class="vchart" data-chart="obr-current-outlook" viewBox="{x0 - 6} {top_v - 6:.1f} '
+        f'{x1 - x0 + 12} {bot - top_v + 12:.1f}" role="img" aria-labelledby="obr-current-t obr-current-d">',
         '<title id="obr-current-t">March 2026 OBR outlook for real GDP and household consumption</title>',
         '<desc id="obr-current-d">Indexed line chart from 2026Q1 to 2031Q1. Real GDP rises from 100 to %.1f and household consumption from 100 to %.1f.</desc>'
         % (g[-1], c[-1]),
-        '<text class="vc-lab" x="62" y="25">March 2026 EFO outlook · 2026Q1 = 100</text>',
     ]
     for tick in (100, 102, 104, 106, 108, 110):
         y = ymap(tick)
         out += [
             f'<line class="vc-grid" x1="{x0}" y1="{y:.1f}" x2="{x1}" y2="{y:.1f}"/>',
-            f'<text class="vc-tick" x="52" y="{y + 4:.1f}" text-anchor="end">{tick}</text>',
+            chartkit.ytick(x0, y, str(tick)),
         ]
+    # Both series are the OBR's published outlook, nothing here is "ours":
+    # GDP in the official ink, drawn solid (no model to contrast it with),
+    # consumption as the neutral. A legend, not end labels: the two paths
+    # cross and end close together, both rising into the edge.
     out += [
-        f'<polyline class="vc-s1" points="{points(g, xs, ymap)}"/>',
-        f'<polyline class="vc-s2" points="{points(c, xs, ymap)}"/>',
-        f'<circle class="vc-s1-dot" cx="{xs[-1]:.1f}" cy="{ymap(g[-1]):.1f}" r="3.5"/>',
-        f'<circle class="vc-s2-dot" cx="{xs[-1]:.1f}" cy="{ymap(c[-1]):.1f}" r="3.5"/>',
-
+        f'<polyline class="vc-s2 vc-solid" points="{points(g, xs, ymap)}"/>',
+        f'<polyline class="vc-s3" points="{points(c, xs, ymap)}"/>',
+        f'<circle class="vc-s2-dot" cx="{xs[-1]:.1f}" cy="{ymap(g[-1]):.1f}" r="3.5"/>',
+        f'<circle class="vc-s3-dot" cx="{xs[-1]:.1f}" cy="{ymap(c[-1]):.1f}" r="3.5"/>',
     ]
-    for i in (0, 4, 8, 12, 16, 20):
-        out.append(
-            f'<text class="vc-tick" x="{xs[i]:.1f}" y="302" text-anchor="middle">{q[i]}</text>'
-        )
-    out += [
-        "</svg>",
-        '<div class="olg-legend">'
-        '<span class="li"><span class="ln ln-s1"></span>Real GDP</span>'
-        '<span class="li"><span class="ln ln-s2"></span>Household consumption</span>'
-        '</div>',
+    ticks_at = (0, 4, 8, 12, 16, 20)
+    for j, i in enumerate(ticks_at):
+        # Narrow: keep 2026Q1, 2028Q1 and the last tick (2030Q1 would touch it).
+        out.append(chartkit.xtick(xs[i], bot, q[i], j, len(ticks_at) - 1,
+                                  minor=j % 2 == 1 or j == len(ticks_at) - 2))
+    out.append("</svg>")
+    return "\n".join([
+        '<figure class="vfig current-outlook">',
+        chartkit.figure("\n".join(out), unit="March 2026 EFO outlook, index, 2026Q1 = 100",
+                        legend=[("ln-s2s", "Real GDP"), ("ln-s3", "Household consumption")]),
         "</figure>",
-    ]
-    return "\n".join(out)
+    ])
 
 
 def boe_chart() -> str:
@@ -91,36 +96,38 @@ def band(lo, hi):
 
     ticks = range(int(vlo), int(vhi) + 1)
     out = [
-        '<figure class="vfig current-outlook">',
-        '<svg class="vchart" viewBox="0 0 800 326" role="img" aria-labelledby="boe-current-t boe-current-d">',
+        f'<svg class="vchart" data-chart="boe-current-outlook" viewBox="{x0 - 6} {top - 6} '
+        f'{x1 - x0 + 12} {bot - top + 12}" role="img" aria-labelledby="boe-current-t boe-current-d">',
         '<title id="boe-current-t">Current boe-svar UK real GDP growth forecast</title>',
         f'<desc id="boe-current-d">Fan chart from the latest complete quarterly data edge, {data["data_edge"]}, forecasting {q[0]} to {q[-1]}. Median with 68 and 90 percent predictive bands.</desc>',
-        f'<text class="vc-lab" x="62" y="25">Current forecast from {data["data_edge"]} · year-on-year %</text>',
     ]
     for tick in ticks:
         y = ymap(tick)
-        cls = "vc-axis" if tick == 0 else "vc-grid"
+        cls = "vc-zero" if tick == 0 else "vc-grid"
+        label = f"{tick:+d}%".replace("-", "\u2212") if tick else "0%"
         out += [
             f'<line class="{cls}" x1="{x0}" y1="{y:.1f}" x2="{x1}" y2="{y:.1f}"/>',
-            f'<text class="vc-tick" x="52" y="{y + 4:.1f}" text-anchor="end">{tick:+d}%</text>',
+            chartkit.ytick(x0, y, label),
         ]
     out += [
         f'<path class="vc-band90" d="{band(lo90, hi90)}"/>',
         f'<path class="vc-band68" d="{band(lo68, hi68)}"/>',
         f'<polyline class="vc-s1" points="{points(med, xs, ymap)}"/>',
         f'<circle class="vc-s1-dot" cx="{xs[-1]:.1f}" cy="{ymap(med[-1]):.1f}" r="3.5"/>',
-        '<text class="vc-note" x="738" y="25" text-anchor="end">90% / 68% bands · median</text>',
     ]
-    for i in (0, 4, 8, 12):
-        out.append(
-            f'<text class="vc-tick" x="{xs[i]:.1f}" y="302" text-anchor="middle">{q[i]}</text>'
-        )
-    out += [
-        "</svg>",
+    ticks_at = (0, 4, 8, 12)
+    for j, i in enumerate(ticks_at):
+        out.append(chartkit.xtick(xs[i], bot, q[i], j, len(ticks_at) - 1))
+    out.append("</svg>")
+    return "\n".join([
+        '<figure class="vfig current-outlook">',
+        chartkit.figure("\n".join(out), legend=[
+            ("ln-s1", "median"), ("sw-band68", "68% band"), ("sw-band90", "90% band"),
+        ], unit="UK real GDP, % change on a year earlier",
+            notes=[f"Current forecast from {data['data_edge']}."]),
         f"<figcaption>Bands are the model&rsquo;s raw posterior quantiles, not rescaled to measured coverage &mdash; the calibrated version is on the <a href='/forecasts'>forecast record</a>. Live-facing forecast, distinct from the frozen 2024Q2 validation experiment. Data through {data['data_edge']}; forecast begins {data['forecast_start']}. Estimated on {data['estimation_sample'].replace('-', '&ndash;')}; {data['draws']:,} posterior draws, {data['accepted']} accepted; five stochastic paths per accepted draw. Source and provenance: <code>papers/boe-svar/figures/current_forecast.json</code>.</figcaption>",
         "</figure>",
-    ]
-    return "\n".join(out)
+    ])
 
 
 def render(page: Path, name: str, content: str) -> str:

diff --git a/assets/make_hero.py b/assets/make_hero.py
@@ -23,6 +23,8 @@
 from pathlib import Path
 
 ROOT = Path(__file__).resolve().parents[1]
+sys.path.insert(0, str(ROOT))
+import chartkit  # noqa: E402  (stdlib-only, repo root)
 PAGE = ROOT / "index.html"
 SRC = ROOT / "papers" / "boe-svar" / "figures" / "current_forecast.json"
 
@@ -61,7 +63,9 @@ def build() -> str:
     quarters, med, lo68, hi68, lo90, hi90 = load_fan()
 
     W, H = 640, 400
-    x0, x1 = 30.0, 604.0
+    # The value ticks sit in a gutter right of the plot, clear of the bands and
+    # the median (at 1440 the "+1%" label used to sit on the median path).
+    x0, x1 = 30.0, 580.0
     top, bot = 40.0, 344.0  # plot area for the value range below
     v_lo, v_hi = -1.7, 3.4
     step = (x1 - x0) / (len(quarters) - 1)
@@ -98,8 +102,8 @@ def band(lo, hi):
     )
 
     out = [
-        f"{MARK_OPEN}",
-        f'<svg class="hero-fan" viewBox="0 0 {W} {H}" preserveAspectRatio="xMidYMid meet" '
+        f'<svg class="hero-fan" data-chart="hero-fan" viewBox="{x0 - 6:g} {top - 6:g} '
+        f'{x1 - x0 + 12:g} {bot - top + 12:g}" '
         f'role="img" aria-labelledby="hero-fan-t hero-fan-d">',
         '<title id="hero-fan-t">boe-svar: UK real GDP growth forecast, median with 68% and 90% credible bands</title>',
         f'<desc id="hero-fan-d">{esc(desc)}</desc>',
@@ -111,8 +115,9 @@ def band(lo, hi):
         out.append(
             f'<line class="{cls}" x1="{n(x0)}" y1="{n(y)}" x2="{n(x1)}" y2="{n(y)}"/>'
         )
+        label = f"{tick:+d}%".replace("-", "\u2212") if tick else "0%"
         out.append(
-            f'<text class="hf-tick" x="{n(x0)}" y="{n(y - 7)}">{tick:+d}%</text>'
+            f'<text class="vc-tick hf-tick vc-gutter" x="{n(x1 + 8)}" y="{n(y)}">{label}</text>'
         )
 
     out.append(f'<path class="hf-b90" d="{band(lo90, hi90)}"/>')
@@ -128,16 +133,15 @@ def band(lo, hi):
     for i, q in enumerate(quarters):
         year, qtr = q.split("Q")
         if i == 0 or qtr == "1":
+            anchor = "start" if i == 0 else "middle"
             out.append(
-                f'<text class="hf-tick hf-q" x="{n(xs[i])}" y="366">{esc(year)}</text>'
+                f'<text class="vc-tick hf-q vc-x" x="{n(xs[i])}" y="{n(bot)}" '
+                f'text-anchor="{anchor}">{esc(year)}</text>'
             )
-    # No in-SVG caption: the <figcaption> next to this figure says the same thing
-    # at real body size. Below 1000px the viewBox scales to roughly half, which
-    # would render these 11px labels at ~6px, so style.css hides .hf-tick there
-    # and the figure reads as a shape with the prose caption carrying the detail.
+    # The labels are lifted into HTML by chartkit, so they stay 12 px at every
+    # width; the <figcaption> next to this figure carries the rest.
     out.append("</svg>")
-    out.append(MARK_CLOSE)
-    return "\n".join(out)
+    return "\n".join([MARK_OPEN, chartkit.figure("\n".join(out), wrap_class="vc-hero"), MARK_CLOSE])
 
 
 BLOCK_RE = re.compile(re.escape(MARK_OPEN) + r".*?" + re.escape(MARK_CLOSE), re.DOTALL)
@@ -146,7 +150,7 @@ def band(lo, hi):
 def render_page(html: str) -> str:
     if not BLOCK_RE.search(html):
         sys.exit(f"no {MARK_OPEN} … {MARK_CLOSE} block in {PAGE}")
-    return BLOCK_RE.sub(lambda _m: build().replace("\\", "\\\\"), html)
+    return BLOCK_RE.sub(lambda _m: build(), html)
 
 
 def main():