Architecture

Repository baseline: bf19250

System Topology

MT5 / MQL5
  └─ CentaurQuant.mq5
      ├─ Core / Data
      ├─ Execution
      ├─ Models
      └─ Network / SDP
             │
             │ TCP
             ▼
      Python Router
        ├─ AI / LLM analyzer
        ├─ Telemetry
        └─ Database

MQL5 Components

• MQL5/Experts/CentaurQuant.mq5 — composition root.
• MQL5/Include/Core/CSymbolNormalizer.mqh — runtime symbol normalization.
• MQL5/Include/Data/DataHarvester.mqh — data acquisition.
• MQL5/Include/Data/ContextPackager.mqh — context preparation.
• MQL5/Include/Execution/COrderBlockScanner.mqh — OB/FVG scanner.
• MQL5/Include/Execution/COrderExecutor.mqh — order execution.
• MQL5/Include/Execution/CHistoryTracker.mqh — history tracking.
• MQL5/Include/Models/CFuzzyFusion.mqh — fuzzy fusion.
• MQL5/Include/Models/CLLMClient.mqh — LLM client integration.
• MQL5/Include/Network/CSDPEncoder.mqh — SDP serialization.
• MQL5/Include/Network/CSocketClient.mqh — bounded socket transport.

Python Components

• Python/router/main.py — central gateway.
• Python/models/analyzer.py — analysis engine.
• Python/database/telemetry.py — telemetry persistence.
• Python/ingest_telemetry.py — telemetry ingestion.
• Python/assistant_bridge.py — assistant integration.
• Python/tests/test_sdp_schema.py — SDP QA.
• Python/publish_forge.py — Forge publication pipeline.

Research-Relevant Architecture

Current research is concentrated on the setup-detection pipeline, particularly COrderBlockScanner and its evidence funnel.

EXP-003A demonstrated that scanner outputs can be invalidated by array-orientation handling even when the underlying pattern population exists. Scanner correctness is therefore a prerequisite for interpreting setup counts.

Architecture vs Research Status

Implemented components and research validation are separate dimensions. A working component does not by itself validate the trading hypothesis built on its output.