DOCCAD — GitHub-Native AI Documentation System
Welcome to DOCCAD, an open-architecture, GitHub-native documentation ecosystem designed for high-assurance engineering teams. DOCCAD transforms version-controlled canonical project knowledge into governed, evidence-backed derived views.
The Core Philosophy
Traditional technical documentation suffers from two failure modes:
- Manual Drift: Architecture documents, runbooks, and API guides slowly fall out of sync with actual code implementations.
- AI Hallucination & Pollution: Ungoverned AI tools inject unverified assertions, imaginary metrics, and deprecated practices directly into source repositories.
DOCCAD eliminates both problems by establishing Two Structurally Separated Content Planes:
- The Canonical Plane (
/docs): Human-authored, reviewed, and immutable to automated generation. This is the single source of truth. - The Derived Plane (
/views): AI-generated views (such as recruiter summaries, interview guides, and dedicated question answers) produced strictly from canonical evidence through PR-gated workflows.
Key System Tenets
- Docs-as-Code: Everything is a version-controlled file in Git. No runtime database, no vector database, and no proprietary content silos.
- AI-Free Static Serving: The published documentation is a static website (Docusaurus 3.x). Readers and search engines access documentation with zero runtime AI latency, cost, or outage vulnerability.
- Deterministic Level-1 Retrieval: Context assembly uses exact file paths, frontmatter dependency closures, and keyword scanning. Retrieval is deterministic, reproducible, and verifiable in pull requests.
- Hash-Based Mechanical Drift: Every generated view records the
sha256content hash of each canonical file used as evidence. When a source file changes, CI flags only affected derived pages as stale. - Anti-Overengineering: The platform is designed to be operated and maintained by a single capable engineer.
Architecture at a Glance
Navigation Quicklinks
- Architecture: System Overview | Content Planes | AI Generation Plane
- Decisions: ADR-001: Source of Truth | ADR-002: Docusaurus 3.x | ADR-003: Content Planes
- Workflow: Validation Gates | Drift Detection
- Security: Trust Boundaries | Prompt Injection Defenses