Skip to content
AI Help Wiki

Agentic-First Documentation

A traditional wiki is built for humans to write and humans to read. A modern knowledge base, however, must serve two audiences: human contributors and the AI agents / LLMs that ingest, analyze, and retrieve this data.

The AI Help Wiki is built as an Agentic-First Wiki to bridge the gap between human collaboration and automated accuracy.

Why Not a Traditional Wiki?

Section titled “Why Not a Traditional Wiki?”

Traditional wikis (like MediaWiki or Wikipedia) have significant drawbacks:

  • Vandalism and Spam: Anyone can make live edits, requiring manual rollback after the damage is done.
  • No Automated Quality Gates: Standard wikis cannot verify code blocks, check formatting rules, or lint links.
  • Heavy Stack Requirements: They require databases, authentication servers, and specialized hosting.

Why Not Edit GitHub Directly?

Section titled “Why Not Edit GitHub Directly?”

While Git-backed documentation (Docs-as-Code) solves quality control, it introduces a high contribution barrier:

  • Git Friction: Contributors must sign up, fork the repository, clone, manage branches, and open pull requests.
  • Non-Technical Barrier: Casual users who notice a typo or want to submit a simple definition are locked out of contributing.

The AI Help Wiki resolves this by providing a direct On-Site Markdown Editor that creates branches and pull requests under the hood, making Git completely invisible to casual contributors.

Our Agentic-First Value Proposition

Section titled “Our Agentic-First Value Proposition”

The AI Help Wiki introduces automated features that go beyond standard markdown hosting:

1. Continuous Factual Grounding (The AI Council)

Section titled “1. Continuous Factual Grounding (The AI Council)”

Every pull request is automatically analyzed by three specialized AI agents before a human moderator is notified:

  • The North Star Guardian: Ensures the content aligns with our educational guidelines.
  • The Ground Truth Validator: Uses Google Search Grounding to cross-reference API names, libraries, version numbers, and conceptual statements with the web, alerting if it finds incorrect info or hallucinations.
  • The Structural Gatekeeper: Enforces formatting, structure, and link verification.

2. Structured Ingestion for RAG Systems

Section titled “2. Structured Ingestion for RAG Systems”

Instead of raw, unformatted HTML, this wiki is compiled as optimized static assets accompanied by search indexes. Because it lives in structured markdown files, external AI agents, loaders, and Retrieval-Augmented Generation (RAG) pipelines can ingest our data cleaner and faster:

  • Clean Structure: Frontmatter metadata (titles, descriptions, tags) is easily parseable.
  • No DOM Clutter: Zero client-side JS by default keeps the HTML output clean from script pollution.

3. Step-by-Step Chain-of-Thought Feedback

Section titled “3. Step-by-Step Chain-of-Thought Feedback”

If a contributor submits a page that fails linting or factual checks, the AI Council does not reject it silently. It posts a detailed, polite, actionable markdown review directly back onto the PR. Contributors get instant, step-by-step guidance on how to fix their work, improving their contribution without moderator intervention.