Home

Welcome to the Navier-Stocks Engineering Wiki

Navier-Stocks is a high-performance, event-driven, heterogeneous multi-agent algorithmic trading system explicitly engineered to capture short-term intraday momentum breakouts and liquidity flows on the Indian Stock Exchanges (NSE and BSE).

By modeling market order books, volumetric expansion, and price velocity as structural fluid vectors, the system treats market turbulence not as a random walk, but as a bounded computational problem governed by predictable mechanical thresholds.

---

💡 The Core Philosophy: Fluid Systems & Tight Constraints

Unlike typical monolithic or generic LLM-based trading scripts, Navier-Stocks is built from the ground up to operate within rigid operational boundaries:

• Resource-Constrained Optimization: Engineered to execute flawlessly on minimal hobby-grade cloud computing hardware. Every component uses highly optimized, lock-free memory structures to preserve performance.
• Deterministic Low Latency: The system splits the processing pipeline into a Fast Path (pure statistical/mathematical Go goroutines targeting sub-5ms internal routing) and a Slow Path (asynchronous AI analysis), ensuring the platform's trigger finger is never delayed by heavy runtime operations.
• Heterogeneous Intelligence: The system combines the mathematical predictability of classical quantitative agents with the advanced context-parsing capabilities of Large Language Models.

---

🤖 The Multi-Agent Ecosystem

Navier-Stocks relies on a coordinated 4-Agent Multi-Agent System (MAS) communicating asynchronously over a distributed message stream. Each agent owns a distinct logical boundary:

Agent Name	Type	Responsibility	Communication Channel
Quantitative Scout	Statistical / Go	Monitors live high-speed WebSocket tick data, maintains local in-memory ring buffers, and isolates volume breakouts.	signal.intent.TICKER
Sentiment Analyst	Cognitive / LLM	Asynchronously scrapes web news, interprets industry narratives, and logs contextual market bias with citations.	PostgreSQL Persistence
Risk Manager	Logical / Go	Validates active time buffers, checks lock-free sector concentration maps, applies the 2% account risk limit, and sizes orders.	signal.execute.TICKER
Trader Gateway	Interface	A swappable execution layer that handles transactional order placement (PaperTrader sandbox or live GrowwTrader).	External Broker API / DB

---

🛠️ Technological Architecture Blueprint

The ecosystem is fully containerized and orchestrated as a single microservice stack running via Docker Compose:

• Core Engine: Written entirely in Go, taking advantage of lightweight, native goroutines for high-concurrency event handling.
• Message Broker: NATS JetStream functions as the resilient event-driven backbone, securing at-least-once delivery semantics and persistent stream replays.
• State & Audit Database: PostgreSQL handles long-term historical transaction tracking, relational state auditing, and pre-computed sentiment matrices.
• Telemetry & Analytics: A decoupled Next.js application serving as a administrative dashboard to visualize live operational metrics, system logs, and portfolio health.

---

📖 Documentation Directory

Navigate through the system design specifications using the sidebar or via the quick links below:

• High-Level Architecture (HLD): The foundational technical overview detailing data topologies, system design goals, and inter-agent communication boundaries.
• Risk Guardrails & Timing Controls: Deep-dive into how the Risk Manager governs exposure using strict intraday time windows, 2% capital allocation logic, and concurrency limits.
• Indicator Reference: Detailed mathematical documentation for the Quantitative Scout's technical engines — VWAP, Bollinger Bands, and Volume Breakout.