AI Dossier

Architecture Documentation

System architecture, design decisions, and technical deep-dives for the Dossier project.

Overview

This section contains comprehensive architectural documentation for developers and contributors who want to understand how Dossier is built.

Contents

System Components

Core Components

  • @ai-dossier/core - Verification and parsing library
  • @ai-dossier/cli - Command-line verification tool
  • @ai-dossier/mcp-server - Model Context Protocol integration
  • @ai-dossier/worktree-pool - Pre-warmed git worktree management

File Structure

  • Monorepo organization (npm workspaces)
  • Package interdependencies
  • Build and publish pipeline

Security Architecture

  • Cryptographic verification flow
  • Key management (Ed25519, AWS KMS)
  • Threat model and mitigations

See ../../security/ for detailed security architecture.

Architecture Decision Records

We use ADRs to document significant architectural decisions. Each ADR describes:

  • The context and problem
  • The decision made
  • The consequences and trade-offs

Browse adr/ for all architecture decisions.

Design Principles

  1. Security First: Cryptographic verification before execution
  2. Simplicity: Minimal dependencies, clear abstractions
  3. Extensibility: Plugin architecture for custom verifiers
  4. Standards-Based: Follow OSS best practices
  5. Developer Experience: Clear APIs, good documentation

For Contributors

External Resources


Rendered from docs/architecture/README.md in the repository. Edit it there.