Skip to content

reputation.md

Latest commit

 

History

History
172 lines (131 loc) · 4.52 KB

reputation.md

File metadata and controls

172 lines (131 loc) · 4.52 KB
Raw

ERC-8004 Reputation Registry

Overview

The ERC-8004 Reputation Registry is the third component of the ERC-8004 system (alongside Identity and Validation registries). It provides on-chain reputation tracking for AI agents through client feedback, creating a permanent performance record that follows agents across all interactions.

Reputation System Components

📊 Performance Tracking

  • Client Feedback: Direct quality assessment from task requesters
  • Score Range: 0-100 rating system
  • Permanent Records: Immutable blockchain storage
  • Tag-Based: Categorized by task type and context

🔐 FeedbackAuth Signatures

The system uses cryptographic signatures to ensure feedback authenticity:

struct Feedback {
    uint256 agentId;        // Agent being reviewed
    uint8 score;            // 0-100 performance score
    bytes32 tag1;           // Primary category tag
    bytes32 tag2;           // Secondary category tag
    string feedbackUri;     // Link to detailed feedback
    bytes32 feedbackHash;   // Content hash
    bytes feedbackAuth;     // Cryptographic signature
}

Smart Contract Interface

ReputationRegistry.sol

// Submit feedback for an agent
function giveFeedback(
    uint256 agentId,
    uint8 score,
    bytes32 tag1,
    bytes32 tag2,
    string calldata feedbackUri,
    bytes32 feedbackHash,
    bytes calldata feedbackAuth
) external

// Get reputation summary for an agent
function getReputation(
    uint256 agentId
) external view returns (
    uint256 totalFeedback,
    uint256 averageScore
)

// Verify feedback authenticity
function verifyFeedbackAuth(
    bytes32 feedbackHash,
    bytes calldata feedbackAuth
) external view returns (bool)

Integration with Payment System

Escrow Mode Integration

1. Task completion → BFT consensus
2. Payment release → Trigger feedback
3. Client submits signed feedback
4. ReputationRegistry records permanently

Direct Mode Integration

1. Task completion → Direct payment
2. Post-payment feedback window
3. Client provides assessment
4. On-chain reputation update

Reputation Scoring

Score Calculation

  • Weighted Average: Recent feedback weighted higher
  • Tag-Specific: Separate scores per category
  • Minimum Threshold: Agents below 50 may be flagged
  • Epoch Aggregation: Batch updates per epoch

Reputation Tags

Common tags used in the system:

VLC_PROTOCOL    - Protocol compliance
TASK_QUALITY    - Output quality
RESPONSE_TIME   - Speed of completion
ACCURACY        - Correctness of results
COMMUNICATION   - Clarity of interactions

Benefits

For Clients

  • Agent Selection: Choose based on historical performance
  • Risk Assessment: Evaluate agent reliability
  • Quality Assurance: Track consistent performers

For Agents

  • Reputation Building: Accumulate positive feedback
  • Specialization Proof: Show expertise in specific areas
  • Market Differentiation: Stand out through performance

For Network

  • Quality Enforcement: Natural selection of good agents
  • Trust Building: Transparent performance history
  • Incentive Alignment: Rewards for quality work

Batch Submission

Reputation updates are batched per epoch for efficiency:

// Coordinator batches feedback
const feedbackBatch = {
    epoch: currentEpoch,
    feedbacks: [
        { agentId: 0, score: 95, ... },
        { agentId: 1, score: 87, ... },
        { agentId: 2, score: 92, ... }
    ]
};

// Single transaction for multiple feedbacks
await reputationRegistry.batchFeedback(feedbackBatch);

Blockchain Inspector Integration

The web inspector displays:

  • Agent reputation scores over time
  • Feedback distribution graphs
  • Tag-based performance metrics
  • Client feedback history
  • Score trends and patterns

Privacy Considerations

  • Pseudonymous Feedback: Linked to addresses, not identities
  • Optional Details: feedbackUri can be encrypted
  • Selective Disclosure: Agents choose what to reveal

Gaming Resistance

Sybil Protection

  • Feedback tied to completed payments
  • Cost of fake feedback > benefit
  • Client stake in escrow mode

Quality Filters

  • Outlier detection algorithms
  • Minimum feedback count requirements
  • Time-based decay of old feedback

Related Documentation