Al033/OTI

OTI is an open historical‑analogue research engine for macro markets. Users (macro analysts, portfolio managers, researchers) type a market event in plain English and receive a one‑page brief that picks three analogues from a curated 39‑event corpus, shows multi‑horizon asset moves (S&P, USTs, USD, gold, oil, HY OAS, VIX), synthesises the recurring failed trade, and identifies where consensus went wrong. Every brief also discloses all retrieved candidates with Jaccard/cosine/rerank/combined scores and a “where this might not fit” caveat. Key features: corpus‑of‑reasoning trajectories (point‑in‑time narratives), two‑phase retrieval/synthesis to prevent look‑ahead bias, hybrid retrieval (Jaccard + voyage‑4 embeddings + RRF + cross‑encoder rerank), numeric‑paraphrase guards and deterministic asset tables, negative‑analogue retrieval, calibrated 80% intervals (conformal leave‑one‑out), FRED/Stooq refresh, Self‑Verifier invariants, streaming UI, multi‑provider AI gateway, and a community PR pipeline for adding events.

AI-related keywords found in this repo's description, topics, or README summary — grouped by category. Each badge links to the corresponding ranking detail page.