2.2 Detection Pipeline

Fetch → Normalize → Calculate → Filter → Display

This is the actual logic, 1:1 with the repo.

2.2.1 Fetch – Data Sources

Capture Engine continuously fetches live prices from:

• Jupiter Aggregator API – cross-DEX quotes.

• Raydium API – AMM pools.

• Orca API – concentrated liquidity AMM.

• Phoenix API – central limit order book.

The engine polls these APIs at a configurable interval (default: every 2 seconds).

2.2.2 Normalize – Unified Representation

Different DEXs expose:

• Different JSON shapes

• Different pair naming conventions

• Different decimal formats

The engine normalizes this into a unified internal format:

• Standard pair notation (e.g. SOL/USDC).

• Comparable price units across DEXs.

• Cleaned data structures that can be fed directly into the calculation step.

2.2.3 Calculate – Spreads & Net Profit

For every pair and DEX combination, Capture Engine computes:

• Spread % – raw price difference between buy venue and sell venue.

• Profit % – estimated net profit after fees & expected slippage.

• Basis points (bps) – profit translated into trading units (100 bps = 1%).

This is where “SOL is cheaper here, more expensive there” turns into a numeric edge you can sort and rank.

2.2.4 Filter – Noise Removal

To avoid spamming you with garbage:

• Opportunities with profit below min_profit_threshold (default 0.1%) are dropped.

• Only pairs and venue routes that clear the bar are passed forward.

You control how aggressive this filter is (details in 2.5 Configuration).

2.2.5 Display – Real-Time Terminal Output

Filtered opportunities are streamed to the terminal in real time, in a compact format like:

[16:15:02] SOL/USDC | jupiter@0.008822 → raydium@0.003801 | spread -56.91% | profit 1.20% (120bps)

Stats: 20 opportunities | Avg: 0.75% | Best: 2.30%

You get:

• Per-opportunity lines (what to buy where, what to sell where, estimated profit).

• A rolling stats line summarizing how the engine is performing (count, average profit, best profit).