Auto-generated. This article is rebuilt from FastAPI route introspection by scripts/build_api_kb.py. Edit the route docstring in app/ and re-run the script — do not edit this file directly.

GET /api/v1/signals/base-rates

Handler: app.signals.routes.get_base_rates
Methods: GET

Description

Historical Base Rate Engine — pattern-match current signals against daily_signals history and return forward-return distributions.

Query parameters:

• backfill_weight (default 1.0) — tunes how much weight the matcher gives historical rows tagged report_filename='historical-backfill' (T2-A inserts). ?backfill_weight=0.5 halves their contribution; =0 drops them entirely; live rows always weight at 1.0.
• mode (default "soft") — one of hard / soft / euclidean / tag / hybrid. soft (default) and euclidean are aliases — weighted-Euclidean distance over the full signal vector with top-K nearest neighbors. hard is the legacy bin-equality matcher. tag (C2, May 2026) filters history by coarse-tag overlap with today’s signals (_HYBRID_FILTER_TAGS set), no Euclidean ranking. hybrid (C2) does both — tag pre-filter then Euclidean rank within the filtered set; option C per DOCTRINE §6 Q1. Per DOCTRINE P8 the default stays soft — hybrid opts in. When the tag filter under-shoots _HYBRID_MIN_ANALOGUES the matcher falls back to Euclidean and surfaces hybrid_fallback=True per DOCTRINE P0.
• top_k (default 100) — Euclidean / tag-mode analogue-set cap.
• max_distance (optional) — Euclidean-mode only; filters analogues whose distance exceeds the cap (None = no cap).
• recency_half_life_days (default 730; T3-B) — exponential decay of analogue weight by trade-date age. 730 = 2-year half-life, so 2021 rows count ~half as much as 2025 ones (5y window straddles two rate regimes — the rate-cycle shift makes pre-2023 dispersion misleading without decay). Pass recency_half_life_days=0 (or any falsy value via =null) to disable.

Parameters

Curl

curl -s https://bigclawd.com/api/v1/signals/base-rates