# Cadence Mechanism

### What It Is

**Cadence** is a high-performance, Rust-based monitoring and alerting system designed for real-time DeFi protocol analysis. Its core function is to detect market anomalies, volatility spikes, and other critical events, and to proactively notify the Neko DeFAI Multiagent System.

It is **not** a passive data pipeline. Instead of providing raw market data, Cadence performs its own sophisticated analysis and sends structured, high-signal alerts only when specific, actionable conditions are met.

#### Your Benefit

Cadence acts as the nervous system for the Neko agent swarm. It handles the intensive, millisecond-grade task of monitoring market conditions, allowing the DeFAI Multiagent System to focus on high-level strategy, planning, and execution. This separation of concerns ensures the agent remains efficient and can react instantly to critical events without the overhead of constant market scanning.

{% hint style="info" %}
This monitoring/alerting system will be helpful and able to detect  and alert the multi-agent system for mass liquidation & cascading liquidation events (which can trigger ADL) both on HyperEVM and HyperCore. 
{% endhint %}

#### System Architecture

The integration between Cadence and the Neko agent is event-driven and uses a **push-based webhook model**, not polling or WebSockets. This ensures minimal latency from detection to action.

**Alert Workflow:**

1. **Continuous Monitoring**: The standalone **Cadence** system continuously ingests and analyzes real-time data from HyperEVM protocols (AMMs, lending venues) and the Hyperliquid order book. It uses specialized services like `Volatility Monitor`, `Anomaly Detector`, and `Health Score Monitor` to score market conditions.
2. **Anomaly Detection**: When a predefined condition is met (e.g., volatility spikes beyond a threshold, a pool's health score drops), Cadence generates a structured **`Alert` object**. This object contains the severity, alert type, and a snapshot of relevant `pool_metrics` at the moment of detection.
3. **Webhook Notification**: Cadence immediately sends the `Alert` object via an HTTP `POST` request to a dedicated endpoint on the agent's API server: `POST /webhook/cadence/alert`
4. **Alert Reception & Triage**: The agent's `handle_cadence_alert` function receives the alert and checks its `severity`.
   * **Non-Critical Alerts** (`Warning`, `Info`): The alert is logged for auditing and an acknowledgment is sent back to Cadence. No automated action is taken.
   * **Critical Alerts** (`Critical`): The agent immediately escalates the alert for processing by the DeFAI Multiagent System.
5. **Prompt Formatting**: For a critical alert, the agent transforms the structured JSON data into a natural language prompt, prefixed with **"URGENT ALERT from Cadence monitoring system"**. This prompt includes all the `pool_metrics` provided by Cadence, which the agent trusts as the real-time state of the market.

   *Example Formatted Query:*

   ```
   URGENT ALERT from Cadence monitoring system:
   Alert Type: volatility_spike
   Severity: critical
   Pool ID: pool_0xffaa4a3d97fe9107cef8a3f48c069f577ff76cc1
   Message: Extreme volatility detected in WHYPE/USDC pool - 85% spike

   Pool Metrics:
   - Price: $1.8505
   - Volume 24h: $1,500,000.00
   - Liquidity: $5,000,000.00
   - Volatility: 85.00%
   - Health Score: 3.20

   Please analyze this alert and determine if any immediate action is needed...
   ```
6. **Defensive Workflow Execution**: This urgent prompt is passed to the `run_agent_workflow`. The system's prompts (`PLANNER_SYSTEM_PROMPT`, `REPLANNER_PROMPT`, `ACTION_PLANNER_PROMPT`) are specifically enhanced to:

   * **Prioritize** these alerts over standard tasks.
   * Shift from yield optimization to **capital preservation**.
   * Generate a streamlined, time-sensitive plan that may involve immediate liquidity removal or range adjustments.
   * Execute actions with a **higher slippage tolerance** (e.g., 1-2%) to ensure transaction success in volatile conditions.

   <figure><img src="/files/KLMP4cP9hYQFcIGAIYlF" alt=""><figcaption></figcaption></figure>


---

# Agent Instructions: Querying This Documentation

If you need additional information that is not directly available in this page, you can query the documentation dynamically by asking a question.

Perform an HTTP GET request on the current page URL with the `ask` query parameter:

```
GET https://docs.neko.fun/architecture/ai-ml-architecture/cadence-mechanism.md?ask=<question>
```

The question should be specific, self-contained, and written in natural language.
The response will contain a direct answer to the question and relevant excerpts and sources from the documentation.

Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.