Argus Margin Intelligence & Data Analysis System

AMIDAS is a high-performance quantitative trading and backtesting engine engineered entirely in Rust. Designed to operate 24/7 on edge hardware (Raspberry Pi), it bridges the gap between fundamental SEC financial data and automated, margin-aware portfolio management.

The system serves as a private "Quant Lab," built to identify market inefficiencies through dynamic scoring, while strictly enforcing risk management and realistic tax accounting. With full Interactive Brokers (IBKR) integration, AMIDAS provides a seamless pipeline from historical research to live market execution.

• Dual-Broker Architecture: Engineered for both speed and traceability. Utilizes a blazingly fast in-memory broker for combinatorial mass-backtesting, paired with a persistent simulated broker that logs every atomic transaction to SQLite for granular auditability.
• Live Market Integration: Fully integrated with the Interactive Brokers (IBKR) API, allowing strategies to transition from simulated environments to live/paper trading with minimal friction.
• Institutional-Grade Accounting: Implements a custom accounting engine that tracks latent liquidation taxes and gross dividend reinvestments. It accurately simulates local tax laws (e.g., prohibition of loss carryforwards) to calculate true net performance.
• Point-in-Time Accuracy: The architecture strictly prevents look-ahead bias and simulates real-world execution slippage, ensuring backtest results are highly realistic.
• Dynamic Combinatorial Scoring: The strategy engine dynamically evaluates equities across multiple fundamental dimensions (e.g., P/E, ROIC). It utilizes configurable buy/sell thresholds to ruthlessly cull underperforming assets.

The codebase is structured into specialized crates, ensuring a clean separation of concerns:

• amidas_core: The foundational domain models (Company, Fundamentals, Price) and shared business logic.
• amidas_errors: Centralized custom error definitions guaranteeing robust error propagation and predictable failure handling.
• amidas_storage: The persistence layer. Powered by SQLite and SQLx for ACID-compliant storage of market history, featuring RAM-based caching mechanisms for high-speed iterations.
• amidas_ingest: The asynchronous data harvester, specialized in fetching and parsing raw financial data from external APIs.
• amidas_scoring: The evaluation engine calculating dynamic scores based on customizable fundamental weights.
• amidas_invest: The allocation mastermind. Orchestrates the InvestManager, enforcing sector constraints, calculating precise position sizing, and executing portfolio rebalancing logic.
• amidas_broker: The execution layer. Features a multi-tiered design encompassing the high-speed InMemoryBroker, the auditable SimulatedBroker, and the IBKR Adapter for real-world order routing.
• amidas_backtest: The time-machine. Simulates thousands of strategy permutations against historical timelines, generating detailed performance (CAGR, Net ROI) and drawdown metrics.
• amidas_cli: The lightweight command-line interface to monitor, control, and evaluate the engine's asynchronous processes.

• Language: Rust (Stable)
• Concurrency: Tokio
• Database Interface: SQLx (Compile-time checked queries)
• Storage Engine: SQLite (Optimized for edge-device I/O)
• Brokerage API: Interactive Brokers (TWS/IB Gateway)

• Workspace & Modular Crate Structure
• Database Schema & Async Migration Logic
• Data Ingestion & RAM-based Valuation Caching
• Detailed Single-Strategy Backtest Engine (with DB-persisted audit trails)
• High-Speed In-Memory Mass Backtesting
• Interactive Brokers (IBKR) API Integration
• Dynamic Thresholding & Combinatorial Strategy Generator
• Resolve Mass-Backtest Concurrency State Leakage
• Margin Monitoring Dashboard (CLI Integration)