Metricore

Metricore is a decentralized risk controller built on Chainlink CRE that continuously 
monitors under-collateralized DeFi lending protocols and autonomously enforces protective 
actions when risk thresholds are breached.

I'm building Invalend Protocol an early-stage under-collateralized lending protocol on 
Base that enables traders to access 5x leverage with only 20% collateral. As I was 
designing the risk architecture, I realized the protocol's core mechanic creates a 
critical vulnerability: a market crash of just 20% can trigger liquidation cascades 
faster than any human operator can respond. I built Metricore as the autonomous 
monitoring and circuit-breaker layer that Invalend needs before going to production.

The problem it solves: existing solutions rely on centralized monitoring, manual 
intervention, or binary circuit breakers with no nuance. Metricore provides graduated, 
autonomous response proportional to actual risk conditions.

It works in three layers:

1. Monitoring
   A CRE Workflow runs every 5 minutes, reading live position health factors 
   and pool utilization directly from Base Sepolia, while fetching external market signals 
   (ETH price via CoinGecko, sentiment via Alternative.me Fear & Greed Index) through 
   NodeRuntime with consensus median aggregation across the DON.

2. Risk Classification
   A three-tier engine (ELEVATED / HIGH / CRITICAL) aggregates all 
   signals and determines the appropriate response: alert only, reduce leverage, or pause 
   new positions.

3. Trustless Execution
   The workflow submits a DON-signed report via evmClient.writeReport() 
   to a Chainlink KeystoneForwarder, which delivers it to MetricoreGateway.onReport(). The 
   gateway independently re-verifies all thresholds from on-chain state before executing, no direct authority is granted to the workflow itself.

For this hackathon submission, I'm using mock contracts that simulate Invalend's lending 
interface. Phase 2 replaces these with Invalend's production contracts as the protocol 
moves toward mainnet launch.

Stack:

• CRE Workflow: TypeScript, @chainlink/cre-sdk , Bun runtime
• Smart Contracts: Solidity ^0.8.20, Foundry, OpenZeppelin
• EVM Encoding: viem (encodeFunctionData, encodeAbiParameters, decodeFunctionResult)
• Blockchain: Base Sepolia (chainId 84532)
• External APIs: CoinGecko (ETH/USD price), Alternative.me (Fear & Greed Index)
• Alerting Backend: Express.js + TypeScript webhook receiver

Architecture:
The CRE Workflow uses CronCapability (every 5 minutes) to orchestrate three monitoring modules. All EVM reads use EVMClient.callContract() with viem-encoded calldata. External HTTP calls use NodeRuntime with consensusMedianAggregation — price and sentiment data reach DON-level consensus before influencing any risk decision.

For EVM writes, the workflow uses the correct CRE pattern: runtime.report() generates a DON-signed report containing the ABI-encoded action type, then evmClient.writeReport() submits it to the MockKeystoneForwarder (simulation) or KeystoneForwarder (production), which validates signatures and calls MetricoreGateway.onReport(). The gateway decodes the report, re-verifies thresholds independently from on-chain state, checks a 30-minute cooldown, and executes on the protocol only if all conditions are confirmed.

MetricoreGateway also stores a PriceSnapshot (price + timestamp) after each workflow run — solving CRE's stateless execution model by using the contract itself as persistent memory for the next run's hourly price delta calculation.