[moremeds]

What

argon now computes the signal-lab gold factor model itself and shows it on the Gold tab. The model is a rolling 252-day fit of gold's daily move (XAUUSD) on the 10-year real yield (DFII10), a synthetic ICE dollar index and a safe-haven spread, with coefficients lagged two rows. Each day's move is split into real-yield, dollar, safe-haven, last-year's-pace and unexplained parts. Every research panel carries this label: Historical description, not a forecast · the direction test (Phase 3) failed.

• Engine macro/gold_factor.py → gold_factor_daily (migration 162, idempotent DDL, no DML).
• Job gold_factor_compute, 17:40 ET Mon–Fri on massive-0, after gold_fred_ingest. It fetches the last month of IB 1-day MIDPOINT closes (6 FX pairs + CMDTY XAUUSD) through xenon POST /historical/bars into macro_series_daily (source IB), then recomputes and upserts. It is healer-enrolled (gold_factor). Needs xenon ≥ v0.8.2, which is live on the mini (verified CASH 200 / CMDTY 200 / unknown 404 with argon's key).
• GET /api/gold/factor, plus the "What moved gold" band of six charts on the Gold tab.
• statsmodels is now declared. It was already installed through arch.

Evidence

• docs/research/2026-10-10-gold-factor-parity.json (scripts/research/gold_factor_parity.py --study <signal-lab worktree @ 2408775> --xenon <bars dir>):
  • from the study's IB closes: inputs, coefficients, intervals, parts and labels exact (4079 rows, max diff 0, 0 label mismatches);
  • from closes pulled through xenon feat(stock): per-stock Short-Vol card on Market Structure tab #174: the same, exact.
• The seed ran end-to-end against IB through a temporary xenon feat(stock): per-stock Short-Vol card on Market Structure tab #174 server into option_wizard_local: 28/28 calls OK, 4080 rows, identical to the CSV-seeded output on the 4079 shared dates.
• Local: CI lint gates; unit 3176 passed; integration 1887 passed, 9 skipped; vitest, typecheck and eslint green.

After merge + release

1. Run the seed on the mini (a prod write): scripts/backfill/gold_factor_seed.py. Without it the job refuses to run (fewer than 254 rows) and names the seed.
2. Confirm the first 17:40 ET run is green.

🤖 Generated with Claude Code