docmd is uniquely positioned as an “AI-Ready” documentation engine. By following these best practices, you ensure that AI models (like ChatGPT, Claude, and GitHub Copilot) can understand and support your project with high accuracy.

1. Enable the LLM Plugin

The first step is enabling the native LLM plugin. This generates structured context files that AI agents crave.

// docmd.config.js
module.exports = {
  plugins: {
    llms: { 
      fullContext: true // Generates llms-full.txt (highly recommended)
    }
  }
}

2. Semantic Heading Hierarchy

AI models use headings to build a mental map of your documentation.

• Don’t skip levels: Always go H1 → H2 → H3.
• Be Descriptive: Instead of “Setup,” use “Installing CLI via NPM.”
• One H1 Per Page: Ensure your frontmatter title is descriptive, as docmd uses it as the primary H1.

3. Code Block Metadata

When providing code examples, always specify the language. This helps the LLM parser apply the correct syntax rules during context retrieval.

```typescript
// Good: Language is specified
const docmd = new Engine();
```

4. Using the llms-full.txt Pipeline

The llms-full.txt file is a concatenated version of your entire documentation.

• Prompting Tip: Tell your AI: “Use the structure in /llms.txt and the full content in /llms-full.txt to answer my questions about this project.”
• Customization: Use llms: false in frontmatter to exclude private or internal-only pages from this public AI context file.

5. Descriptive Image Alt-Text

While AI is getting better at vision, text is still the most reliable way to provide context. descriptive alt text in your images ensures that even if the AI doesn’t “see” the image, it understands its purpose in the build.