Project Architecture & Principles
MIDAS ATLAS is an engineering-led project to design and operate a real-time automated trading runtime with explicit safety gates and measurable behavior.
01 The Architecture Narrative
At runtime, MIDAS ATLAS consumes market, account, depth, and order messages and routes them into period-specific ledgers. Those ledgers are enriched with technical indicators and cross-order metrics, then exposed to strategy/risk modules for decisioning.
A supervisor control plane exposes the mode lifecycle, health/readiness, and test orchestration through modular web APIs and the operator UI. The goal is to make every phase auditable: from ingest and ledger updates to indicator calculations and strategy/risk decisions.
Tech Stack
- Rust runtime (midas-engine)
- React/Vite operator frontend
- RabbitMQ transport bridge
- JForex integration layer
- Postgres/Timescale telemetry
Canonical System Topology Diagram
02 Project Principles
Deterministic by Default
Backtest and replay paths exercise the same logic routes as the live engine, ensuring behavioral consistency across environments.
Observable, Not Opaque
Strategy decisions are not "black boxes." Every reject or accept intent carries an audit reason visible in the operator UI.
Engineering-First
Measured results over marketing hype. The roadmap is prioritized by architectural robustness and data integrity.
Future Extensions
MIDAS ATLAS is designed as a modular execution research platform. JForex is the current bridge; future adapter work is planned so strategy and risk logic can be evaluated across multiple execution venues, including MT4/MT5.
Near-Term
- Richer ledger inspection views
- Risk orchestration via Web UI
- Expanded replay analytics
Planned
- Multi-broker adapter design
- MT4/MT5 bridge feasibility
- Large-scale parameter exploration