EFTS · client project
Trading signals, cross-checked before anyone acts
Traders get signals from many places at once. Acting on a single alert that nobody has checked is how money gets lost.
Context
EFTS captures TradingView alerts and other trading signals, checks each one against independent sources, and adds AI research with a short summary. It runs inside a subscription product with a user dashboard. I worked on it in a team of three for ten months.
How it works, step by step
Step 1 of 6
ShippedA signal is captured
When a TradingView strategy or another source the trader follows raises a signal, it's caught the moment it fires.
TradingView posts a webhook with the strategy's payload to the EFTS webhook API; other sources arrive through their own connectors.Step 2 of 6
ShippedEvery signal is kept, exactly once
Each signal gets a unique key and waits in a queue, so a burst of alerts is never lost or counted twice.
Signals are written to a job queue with an idempotency key. At-least-once delivery plus dedupe makes retries safe.Step 3 of 6
ShippedChecked against other sources
The signal is compared with market data and other independent sources. One that only a single source backs is flagged, not hidden.
A verification job pulls price data and other feeds for the same symbol and window, and records which sources agree and which don't.Step 4 of 6
ShippedAI research, summarised
AI reads the news around the symbol and writes a plain summary: what supports the signal, what argues against it, and the source of each point.
An LLM research step gathers context per symbol and writes a summary with sources attached; claims without a source are dropped.Step 5 of 6
ShippedThe plan is checked
Before a job runs, the system checks that the user's subscription includes it.
Stripe multi-tier subscriptions map to entitlements, which are checked before a job executes.Step 6 of 6
ShippedThe trader decides
The signal, its checks and the summary appear on the trader's dashboard. The decision stays with the trader.
A signal log in MongoDB feeds a React dashboard with live updates.
Decisions & tradeoffs
- Chose
- a queue with idempotency keys
- over
- processing signals directly inside the webhook handler
- because
- a lost or doubled signal gives the trader a wrong picture. At-least-once delivery with dedupe made retries safe.
What's real today
- Shipped6 steps
Client project. Signals and summaries inform a decision. They're not financial advice and don't promise any result.
What it achieved
- sources checked for each signal
- Several
- plain summary per signal, with sources
- 1
- of development
- 10 months
- engineers on the team
- 3
Stack & role
- React
- Node.js
- Express
- MongoDB
- Stripe