Skip to content

Latest commit

 

History

History
76 lines (60 loc) · 2.65 KB

File metadata and controls

76 lines (60 loc) · 2.65 KB

Changelog

All notable changes to this project will be documented in this file.

The format is based on Keep a Changelog, and this project adheres to Semantic Versioning.

[1.0.0] - 2025-12-25

Added

  • Core Package (@agentledger/core)

    • Ledger class for managing audit sessions
    • Hash chain creation and verification for tamper-evident logging
    • Merkle tree construction and proof generation
    • Ed25519 digital signatures for cryptographic attestation
    • Multiple storage backends: InMemoryStorage, FileSystemStorage, S3Storage
    • Entry types: ModelCall, ToolInvocation, DecisionPoint, HumanApproval, StateSnapshot, ContentReference
    • Export formats: JSONL, Splunk CIM, Elastic ECS, FINRA 4511, EU AI Act
    • Zod schema validation for all entry types
  • OpenAI Package (@agentledger/openai)

    • AuditedOpenAI wrapper class for automatic audit logging
    • Support for both streaming and non-streaming completions
    • Automatic token counting and cost estimation
    • Content storage option for full prompt/completion capture
    • audited decorator for function-level auditing
    • Convenience functions: createAuditedOpenAI, auditedChatCompletion
  • CLI Package (@agentledger/cli)

    • verify command for validating audit log integrity
    • export command for converting to compliance formats
    • replay command for viewing session timelines
    • summary command for statistics and compliance scoring
    • Colorful terminal output with progress indicators
  • Testing

    • Comprehensive Jest test suite with 162+ tests
    • 90%+ coverage on core cryptographic functions
    • Integration tests for CLI commands
    • Mock OpenAI client for testing without API keys
  • Examples

    • Interactive demo showing 3-turn conversation
    • Mock OpenAI fallback when no API key present
    • Automatic export to multiple compliance formats
  • CI/CD

    • GitHub Actions workflow for Node 18, 20, 22
    • TypeScript type checking
    • Security audit
    • Demo verification

Security

  • Zero external cryptographic dependencies (uses Node.js crypto module only)
  • Ed25519 signatures for tamper-proof attestation
  • SHA-256 hashing for content integrity
  • Hash chain verification for sequential integrity

Compliance

  • FINRA 4511 Books and Records format
  • EU AI Act Article 12 logging requirements
  • 7-year default retention period
  • Human review workflow support
  • Decision point documentation

[Unreleased]

Planned

  • Anthropic SDK integration
  • LangChain callback handler
  • Real-time S3 streaming
  • Web dashboard for log visualization
  • Webhook notifications for compliance events