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/ai2/history

Handler: app.signals.ai_brief_routes.ai2_history
Methods: GET

Description

PUBLIC (the SECOND deliberate non-gated surface in this module, after /ai2/verdict) — /engine renders the AI-2 Tracker and /tape’s AI Read card reads the direction call + trade idea, both on a 60s refresh. Every served string is passed through the vendor-anonymity scrub. Never 500s, never 404s: an empty store returns 200 + count: 0 + honest empties.

Serves the full recorded verdict history (the persisted ai2_verdicts rows, newest-first, grouped per AI-1-read cycle, scrubbed) plus the DISTILLED public grade subset. An llm cycle carrying a trade idea also carries its gradeable trade_idea_meta sidecar (underlying ticker / profit direction / trading-day horizon — research/planning/prereg_ai2_trade_idea_2026-07-21.md) beside the prose, and grades.trade_idea carries the lane’s distilled scorecard (verdict + matured count; the research internals stay behind the internal-gated /ai2/trade-idea-grade). History IS the persisted store — there is no live-compute tier (unlike /ai2/verdict). The grade subset NEVER exposes the research internals (eras[], h1/h2/h3, bootstrap CIs, verdict_config_version hashes) that keep /ai2/risk-grade + /ai2/magnitude-grade internal.

Cache (mandatory): the two graders each full-scan daily_signals, so a public 60s-refresh page must not fan them out per request — a module-level probe cache + double-checked single-flight lock collapse a cold-start burst to ONE compute. limit is folded into the cache key so a smaller-limit request never serves a wider cached slice.

Response shape::

{schema_version, limit, count, cycles_recorded, tracking_started,
 cycles: [...], grades: {magnitude, risk_band}, reason}

Parameters

Curl

curl -s https://bigclawd.com/api/v1/ai2/history