What (One-liner)

The latest official sovereign-curve cross-section, carried exactly as published (percent + compounding convention travelling with the data) — one curve node feeding both an engine chain and an independent verification chain.

How to use

Pick the sovereign curve (UK gilts nominal by default) and run — it fetches the official cross-section (cached daily). Wire it upstream of ldi_funding_gap_v1 (the engine consumes curve_points) and/or a script node (bind curve_points.tenors / .spot_rates_pct / .annual_compounded for an independent recompute). The LDI cross-check template shows the full pattern.

Why it does NOT normalise

BoE GLC and MoF publish annually-compounded rates; GSW and ECB publish continuously-compounded ones. This carrier tags the convention (compounding + bind-safe annual_compounded) instead of converting, so every consumer applies ln(1+r) independently — which puts the normalisation step itself inside the cross-check's verified surface.

Input / Output contract

Input: curve (boe_nominal / boe_real / gsw_nominal / ecb / jp_mof).

Output (v2): curve_series (rectangular history panel: dates/dates_ord/tenors/rates_pct + convention tags) + series_pct.tN per-tenor flat series sharing ONE dates axis (script-bindable without triggering the lane's shortest-length alignment) + curve_points (= the panel's last row; one rule) plus flat display scalars (rate_2y_pct, rate_10y_pct, rate_30y_pct, slope_10y_2y_bp, as_of, n_tenors, staleness). Unfetchable curve → status: unavailable with the reason.

Known limitations

1. Official daily estimates with a publication lag (age_days / stale declared). 1a. Panel depth = what the source adapter holds (the BoE adapter currently merges only the '2025 to present' archive workbook, ~1.6y and growing; GSW 1961→, MoF 1974→) — date_min/n_dates declare it, never padded. Deeper BoE history = a curve_sources adapter extension, tracked separately.
2. Display spots are exact-tenor lookups — honest null when a source lacks that tenor.
3. jp_mof carries par yields flagged as a spot approximation.