Building

Real-Time Market Scanner

A scanner that ingests live market data and surfaces setups as they form — the tool I wished existed off the shelf, built correctness-first.

GitHubUpdated Aug 8, 2026

Stack

Next.jsPythonWebSocketsPostgresRedis

Problem

Off-the-shelf scanners only match a fixed menu of setups, run on a delay, and quietly fire on bad ticks. I couldn't define my own rules or trust the data.

Idea

A streaming pipeline where I define setups as composable rules, evaluated in real time on a feed I've made trustworthy first.

Why I'm building it

I wrote up the thinking behind this in Building My Own Market Scanner. Short version: a great scanner is a data-engineering project with a trading skin, and I wanted one I could actually trust and extend.

Current state

  • ✅ WebSocket ingest with reconnect + gap detection
  • ✅ Bar alignment and split/halt handling
  • 🔨 Incremental indicator engine (VWAP, RSI, relative volume)
  • ⏳ Rule matcher + alert delivery
  • ⏳ Web UI for defining setups

Development log

2026-08-08 — Reconnects were silently dropping the first few seconds of ticks. Fixed by rehydrating per-symbol state before resuming evaluation.

2026-07-30 — Moved indicators from batch recompute to incremental updates. ~40x less CPU intraday.

Lessons so far

The data layer is 80% of the work and 100% of the trust. Every "bug" so far has been a data-correctness problem wearing a strategy costume.

Follow the builds

I email when a project ships a meaningful update, and when I start something new. Roughly monthly, never spam.

✶ No spam, promise! Unsubscribe anytime ✶