Master AI Agent Architecture

The figure above presents the architecture for NODO AI. The system design follows the agent/service-based, which is highly customizable, flexible, and scalable.

• Vault UI/UX: The UI/UX layer serves as the user-facing portal, allowing individuals to manage their investments, view analytics, and interact with the AI Assistant. Built using modern frontend frameworks, it integrates real-time data visualizations and provides seamless access to all user features. The conversational interface allows users to naturally express investment goals, which makes the experience more accessible and engaging.

Data Intelligence Layer:

• Data Management Service: This service is the central repository for all structured and unstructured data within the system. It supports both reads and writes from various agents and services. It uses a combination of storage technologies—such as PostgreSQL for structured metadata, InfluxDB or TimescaleDB for time-series data, and object storage for raw files.

• Data Transformation Service: The role of this service is to preprocess raw data received from the Data Collector Agent into clean, normalized, and structured formats. It applies schema mapping, timestamp alignment, and formatting logic to convert raw data into analyzable datasets. This ensures downstream agents receive data in a consistent format, reducing coupling and increasing modularity.

• Data Analyst Agent: This agent is responsible for converting structured data into actionable insights. It runs a collection of analyzer plugins that perform operations such as feature extraction, indicator generation (e.g., moving averages, volatility bands), synthetic data generation, and outlier detection. These processed datasets are then stored back in the Data Management Service and used by strategies to optimize decision making.

Market Making Layer:

• AI Agent for Market Making Execution Agent: The AI Agent is an intelligent interface backed by a natural language processing engine. It uses large language models (LLMs) fine-tuned for financial and investment domains. This backend interprets conversational inputs, classifies user intents, extracts relevant parameters, and generates structured output. The AI Agent bridges the gap between human language and technical system commands, improving usability and reducing friction.

• Backtest Environment: The Backtest Engine provides a sandboxed environment to evaluate the effectiveness of strategies before live deployment. It retrieves historical datasets from the Data Management Service and simulates strategy behavior over them. Performance is measured in terms of PnL, Sharpe ratio, drawdown, and other KPIs. This allows developers and operators to iteratively improve strategy logic.

• Strategy Execution Service: This component takes execution plans generated by the Master Agent and carries them out on-chain. It interacts with blockchain networks using Web3 libraries and signs transactions using secure keys. The service is built with retry logic, transaction state tracking, and logging to ensure execution reliability and traceability. It plays a crucial role in translating analytical decisions into actual yield-generating operations.

Smart Contract Layer:

• NODO Vault Contract: NODO Contracts are smart contracts deployed on supported blockchain networks. They manage the custody of user and system funds, facilitate deposits and withdrawals, and mint vault tokens as interest-bearing representations of user stakes. These contracts also perform trade execution under instruction from the Strategy Execution Service, ensuring transparency, auditability, and decentralization.

Module Name	Function	Inputs	Processing Logic
1. Real-time Data Aggregation Module	Stream and normalize market data	WebSocket feeds (price, depth, trades), Sui RPC, REST APIs	- Normalize multi-DEX data - Cache in Redis - Trigger anomaly alerts
2. Predictive Analytics Module	Forecast price & volatility	Aggregated time-series data, Historical OHLCV	- LSTM or Transformer model for time-series forecasting - Calculate asset correlations
3. Cross-Protocol Liquidity Analytics Module	Track liquidity across Sui protocols	Move object metadata, LP token state, wallet balances	- Decode Move LP objects - Calculate deployed capital, pool share, yield to date
4. Portfolio Management Module	Allocate capital across assets	Price forecasts, vol forecasts, correlation matrix	- Apply constrained optimization (mean-variance, Black-Litterman) - Enforce allocation rules (e.g., 80/20)
5. Liquidity Rebalancing Module	Enforce capital distribution	Target allocation map, current LP positions	- Compute delta between current and target - Construct Sui batch txns to shift LP
6. Volatility-Aware Position Sizing Module	Dynamically size positions	15-min rolling price data, baseline volatility	- Compute vol_score = stddev(window) - Adjust position sizes: size ∝ 1/vol_score
7. Circuit Breaker Module	Stop trading during anomalies	5-min price movement %, live trades	- Detect if abs(p_t - p_t-5m) / p_t-5m > 15% - Trigger halt status flag
8. Reserve Management Module	Maintain idle buffer capital	Total capital deployed, LP vault state	- Enforce 30% capital lockup - Monitor idle wallet liquidity
9. Hourly Backtesting Module	Benchmark live strategies	Historical pool data, live trade logs	- Simulate alternative strategies (spread curves, slippage models) - Compare 1-hour ROI and risk
10. A/B Algorithmic Testing Module	Evaluate competing MM models	AMM model variants, live pair data	- Deploy multiple sub-agents with config variants - Log and compare yield, slippage, gas costs
11. LP ROI Reporting Module	Track per-pair profitability	Trade logs, LP token valuation, fee receipts	- Compute Gross ROI, Net ROI (IL-adjusted) - Group by pair/agent/day