Mentality Pipeline / v0.1

Thesis

Capture the trader,
not just the market.

Every quant system records market data tick by tick. The person pressing the button remains invisible. This project records heart rate, gaze, facial expression, and on-screen context — synchronized on a single monotonic clock with market ticks — so every decision point becomes a structured, queryable row in a multimodal dataset. Runs on a webcam alone; pair a BLE heart-rate strap for clinical-grade HR.

What each stream captures

One moment, one row

What the pipeline produces at a single decision point. Every column shares the same tsMs — the level of synchronization required to detect patterns that span market state and trader state simultaneously.

tsMs        42517.231
market      symbol=BTCUSDT price=63_420.10 change1m=-3.12%
wearable    bpm=108 rmssd=22ms // baseline 74 → +46%
webcam      rppg_bpm=105 signal_q=0.71 peak_hz=1.78
face_ctx    gaze=left browInnerUp=0.72 jawOpen=0.31 head_pitch=-8°
event       type=warning_breach payload={source:"wearable",dev:46.2}
→ inferable pattern  sharp drop + brow tension + gaze averted + HR spike = panic-response candidate

This row format is what gets persisted per session. Pattern discovery is a downstream task; the pipeline's job is to make that discovery possible by recording the right signals at the right resolution.

Scope