Components

Git2Docs consists of several interconnected components that work together to generate and serve documentation.

GitHub Integration

Git2Docs connects to your repositories via a GitHub App. This provides:

  • Secure access - Fine-grained permissions to only the repos you select
  • Webhook events - Automatic triggers on releases and pushes
  • Repository cloning - Access to clone repos for analysis

The GitHub App can be installed at the organization level, granting access to multiple repositories at once.

Documentation Engine (Worker)

The worker is the core processing engine that analyzes your codebase and coordinates with AI to generate documentation. It runs as a background service and handles:

  • Repository cloning - Fetches code at specific tags/branches
  • Code parsing - Uses tree-sitter to parse TypeScript, Python, and other languages
  • Artifact extraction - Identifies functions, classes, types, and their relationships
  • AI orchestration - Manages the three-stage pipeline (Comprehend → Plan → Synthesize)
  • Change detection - Tracks what changed between versions using dual-hash pattern
Dual-Hash Pattern: Each code artifact has two hashes - a signature hash (public API) and content hash (implementation). This enables smart change detection: signature changes require doc updates, implementation changes may only need review.

AI Language Model

Git2Docs uses Claude (Anthropic's AI) to understand code and generate documentation. The AI is used in multiple stages:

  • Comprehension - Understanding what the product does and who it's for
  • Planning - Designing the documentation structure and table of contents
  • Synthesis - Writing the actual documentation content

Each AI call is carefully crafted with context about the codebase, existing documentation, and the specific task at hand.

Data Store

PostgreSQL stores all documentation data with a multi-tenant architecture:

TablePurpose
organizationTeams and their settings
repositoryConnected GitHub repos
doc_briefProduct briefs from comprehension stage
doc_spaceDocumentation spaces (e.g., "User Guide")
doc_nodeIndividual documentation pages
doc_sectionContent sections within pages
doc_section_versionVersion history with provenance tracking
brand_configBranding settings (logo, colors)

Web Application

The Next.js web app provides two main interfaces:

Documentation Portal

A public-facing site where users browse documentation. Features include:

  • Product landing pages with cards for each documentation space
  • Hierarchical navigation with sidebar
  • Full-text search (coming soon)
  • Custom branding per organization

Team Dashboard

A private interface for teams to manage their documentation:

  • Connect and manage GitHub repositories
  • Trigger documentation generation
  • Review and edit AI-generated content
  • Configure branding and settings
  • View generation progress in real-time