How Barfinex Works: The Five-Service Architecture

The Core Idea

Each stage of trading gets its own service. Services talk through typed events. Every decision is traceable.

This means you can update a strategy without restarting the data feed. Adjust risk rules without touching the AI. Trace any trade back to the signal, AI reasoning, and risk check that produced it.

The Five Services

Provider — Your Data Foundation

Provider is the single source of truth for all market data. No other service talks to exchanges directly.

What it does for you:

Connects to exchanges (Binance, Alpaca, more via plugins)
Normalizes data into clean OHLCV candles
Detects and repairs data gaps automatically
Exposes one unified API — the only endpoint you need

Detector — Your Strategy Engine

This is where your trading logic lives. You define rules, weights, and thresholds as a TypeScript config.

Evaluates rules on every new candle
Scores conditions and emits signals only when the threshold is met
Shows exactly which rules fired and why (full attribution)
Multiple strategies run in parallel, isolated from each other

Advisor — Your AI Decision Maker

Takes each signal through an 8-stage pipeline:

Fetches market context
Checks data quality
Scores the signal independently
Calculates conviction (0-1)
Calibrates against current market regime
Synthesizes with LLM (with hallucination detection)
Validates spread and risk/reward
Emits execution intent

Every stage is logged. 28+ event types for full traceability.

Inspector — Your Risk Manager

No order reaches the exchange without Inspector's approval. It checks:

Position size limits
Portfolio exposure caps
Drawdown thresholds
Loss streak limits
Cooldown periods

Also manages stop-loss and take-profit orders automatically.

Studio — Your Dashboard

Real-time view of the entire pipeline:

Live charts with signal overlays
AI decision logs with reasoning traces
Risk dashboard with active positions
Capital efficiency metrics

Studio is read-only — it observes but doesn't control.

How Data Flows

Exchange → Provider → Detector → Advisor → Inspector → Exchange
                                                    ↓
                          Studio (observes everything)

Every arrow is a typed event with a timestamp and audit record.

Deployment

docker-compose up starts the full stack locally. Each service can also run on separate infrastructure as you scale.