Overview — System Vision V1
Source: Notion | Last edited: 2025-10-31 | ID: 2962d2dc-3ef...
Eon Quant OS (EQOS) is an AI-agent-centric, DSL-driven quantitative research and execution operating system.
It transforms the full lifecycle of quantitative strategies — from data to features, models, signals, and execution — into a declarative, composable, and self-evolving framework.
Core Idea
Section titled “Core Idea”“Turn quantitative research from ad-hoc scripts into a compilable system.”
- DSL as the single source of truth – every strategy, model, and execution logic is declared in a unified language.
- Compiler as the translator of ideas – converts DSL definitions into executable DAGs for full transparency.
- Runner as the execution engine – manages context, invokes plugins, and produces metrics and reports.
- Plugins as modular building blocks – all features, labels, models, backtests, and execution modules are plug-and-play.
- AI Agents as autonomous researchers – automatically generate, train, evaluate, and evolve strategies.
System Flow
Section titled “System Flow”DSL → Compiler → DAG → Runner → Plugin → Registry → Execution Gateway
Four-Layer Architecture
Section titled “Four-Layer Architecture”1️⃣ Core Layer – Research Core
Includes the DSL, Compiler, DAG, and Runner.
Defines what to compute and how data flows through the pipeline.
2️⃣ Plugin Layer – Functional Modules
Implements all operations — data access, feature engineering, labeling, training, backtesting — as modular plugins.
3️⃣ AI Agent Layer – Intelligent Research
Autonomous agents generate DSLs, run experiments, evaluate results, and continuously improve performance.
4️⃣ Execution Layer – Trading Gateway
A Rust-based gateway translates signal intents into real orders with risk and latency control.
Development Phases
Section titled “Development Phases”Design Principles
Section titled “Design Principles”- Declarative > Imperative – define what to do; the system decides how to execute.
- Modular > Monolithic – all components are independent, reusable, and testable.
- Autonomous > Manual – AI agents continuously discover and optimize strategies.
- Traceable > Ephemeral – every experiment and signal has a unique ID and full context.
Minimal Executable Pipeline (MVP)
Section titled “Minimal Executable Pipeline (MVP)”Process
- DSL (strategy.yaml)
- Compiler → generate dag.json
- Runner → execute DAG
- Plugins → train and backtest
- Outputs → metrics.json / bt_report.json / intent_log.jsonl Artifacts
- Feature and model outputs
- Backtest performance reports
- Signal intents for execution
- Reproducible experiment records
Vision
Section titled “Vision”EQOS aspires to become a
self-evolving quantitative research ecosystem powered by AI.
- For Human Researchers: a complete automation framework from idea to deployment.
- For AI Agents: an operating system they can read, modify, and execute.
- For Institutions: a foundation for standardized and reproducible quant research.