Artifacts
Git2Docs works with several key artifacts that flow through the system, from code analysis to published documentation.
Product Brief
The product brief is a structured representation of your product extracted from analyzing the codebase. It serves as the foundation for all documentation generation.
{
"productName": "Git2Docs",
"oneLiner": "AI-powered documentation from code",
"targetAudience": "Software development teams",
"audienceLevel": "intermediate",
"deploymentModel": "saas",
"primaryWorkflows": [
"Connect GitHub repository",
"Generate documentation",
"Review and edit content",
"Publish to custom domain"
],
"keyConcepts": [
"Documentation spaces",
"Doc nodes and sections",
"Ownership states",
"Confidence tiers"
],
"components": [
"Worker (documentation engine)",
"Web app (portal and dashboard)",
"Database (content storage)"
]
}The brief is stored in the database. In practice you rarely hand-edit it: the Tell AI → Submit Reviewflow on the audit/edit page (and the runtime-validation findings ingest) both write brief patches through the AI consolidator, so the brief converges on what the validator and the maintainer collectively flag as wrong without anyone editing fields by hand.
Documentation Space
A space is a collection of related documentation pages. Each repository can have multiple spaces, such as:
- Getting Started - Onboarding and quickstart guides
- User Guide - In-depth usage documentation
- API Reference - Technical API documentation
- Architecture - System design and concepts
Spaces appear as cards on the documentation portal home page, allowing users to navigate to the documentation they need.
| Field | Description |
|---|---|
name | Display name of the space |
slug | URL-safe identifier |
description | Brief description shown on cards |
icon | Emoji or icon identifier |
ordering | Display order on the portal |
Doc Node
A doc node represents an individual documentation page. Nodes can be organized hierarchically with parent-child relationships.
| Field | Description |
|---|---|
title | Page title |
slug | URL path segment |
description | Page summary |
doc_type | Type of documentation (see below) |
source_type | ai_synthesized or api_reference |
parent_id | Parent node for hierarchy |
source_hints | File globs and references for synthesis |
Doc Types
Each page has a doc type that determines its structure and sections:
- guide - Step-by-step instructions for tasks
- tutorial - Learning-oriented walkthroughs
- concept - Explanations of key ideas
- architecture - System design documentation
- cli_reference - Command-line interface reference
- api_reference - API endpoint or function reference
- runbook - Operational procedures
- changelog - Version history and release notes
Doc Section
Each doc node contains multiple sections. Sections are the granular units of content that can be individually edited and tracked.
| Field | Description |
|---|---|
section_type | Type of section (overview, steps, etc.) |
ownership_state | ai_owned or human_owned |
ordering | Display order within the page |
Ownership States
- ai_owned - Content generated by AI. Will be updated on regeneration.
- human_owned - Content edited by a human. Protected from AI overwrites.
- contested - AI wants to update human-owned content. Requires resolution.
Section Version
Each section maintains a version history. This enables tracking changes over time and understanding where content came from.
| Field | Description |
|---|---|
content | The actual markdown content |
confidence_tier | high, medium, or low |
author_type | ai or human |
provenance | JSON tracking source artifacts |
is_current | Whether this is the active version |
Comment
A piece of maintainer or validator feedback pinned to a specific section in a specific version. The Tell AI button on the audit page creates one; the runtime validator's findings ingest creates one per failed artifact (tagged with source_finding_id). Comments are the structured signal the AI consolidator reads when the maintainer hits Submit Review.
| Field | Description |
|---|---|
body | The feedback text |
status | open · resolved · dismissed |
source_finding_id | UUID of the validation finding (null for hand-authored) |
last_regen_at | Receipt — last review run that included this comment |
author_user_id | Who wrote it (validator authorship coming in a follow-up) |
Validation Finding
A behavioural check produced by a validator: an external coding agent running Tier 1, or the worker's built-in Tier 0 static checker. One row per (section_version_id, tier, validator, artifact_kind, artifact_text); re-running a validator replaces its prior findings rather than appending.
| Field | Description |
|---|---|
tier | 0 static / 1 runtime / 2+ custom |
validator | Identifier of the validator (e.g. tier1:claude-code-session@<product>) |
artifact_kind | cli_invocation · http_request · others |
artifact_text | The exact command / call the docs claim |
status | passed · failed · skipped · unverifiable · no_schema |
evidence | JSON with the discrepancy text + substituted form + response excerpt |
Code Artifact
For API reference documentation, Git2Docs extracts code artifacts from your source files using tree-sitter parsing.
- Functions and methods
- Classes and interfaces
- Type definitions
- Constants and exports
Each artifact is tracked with dual hashes for smart change detection between versions.