Skip to content
DocuBot documentation

How it works

DocuBot is more than a simple text generator; it’s a documentation engine designed to understand the relationships between your codebases and your users. By analyzing your source code and applying a layered style system, DocuBot creates structured, static help centers that read as if they were written by a human technical writer.

Multi-repository aggregation and roles

DocuBot treats your documentation site as a unified product, even if that product spans multiple GitHub repositories. When you connect your sources, the engine assigns a functional role to each repository to guide the AI’s focus:

  • Primary UI: This is the main application experience. DocuBot prioritizes this code to understand the core user journey.
  • UI App: These are secondary experiences, such as an admin console or a mobile app, that complement the primary product.
  • API Reference: Repositories containing OpenAPI or Swagger specifications are treated as technical data sources for developer documentation.
  • Reference: Supporting codebases that provide context but aren’t the primary focus for end-user guides.

You can override these roles at any time to ensure the AI understands exactly which parts of your codebase deserve the most attention.

The documentation generation pipeline

The lifecycle of a documentation sync follows a rigorous process to ensure accuracy and prevent the “hallucinations” common in standard AI tools.

  1. Repository scanning: DocuBot identifies UI signals, functional modules, and API specifications within your connected repositories.
  2. Structural planning: Before writing any content, the engine creates a detailed plan for every document type you selected. This plan identifies the intent of each section and the specific evidence in the code that supports it.
  3. Content writing: Using the structural plan as a map, the engine writes the documentation using your selected voice and tone.
  4. Static compilation: The final content is compiled into high-performance HTML pages, ensuring your documentation is ready for public consumption immediately.

Layered style fusion engine

To keep your documentation cohesive, DocuBot uses a “style fusion” engine that combines three distinct layers of guidance:

  • General Styles: This is your overall brand voice. Whether you choose “Precision Technical” for developers or “Approachable Guided” for new users, this layer controls the vocabulary and attitude of the writing.
  • Quadrant Guidelines: Based on the Diátaxis framework, this layer determines the structure of the page. It ensures that a tutorial feels like a lesson, while a reference page stays focused on facts.
  • Document Type Templates: These are specific rules for pages like “Quick Start Guides” or “FAQs.” They ensure that every page of a certain type contains the information users expect to find.

AI context management and accuracy

DocuBot maintains technical accuracy through a system called context budgeting. Because codebases can be massive, the engine calculates a “budget” for each repository based on its assigned role.

The primary application receives the largest share of the budget, allowing the AI to see the most detail. Shared context packs then allow the documentation for one repository to understand dependencies in another. This means your “Integration Guide” can accurately describe how your frontend interacts with your backend API without losing track of the details in either.

Design rationale: Static vs. runtime delivery

DocuBot generates static documentation rather than rendering content on demand. This approach provides several key benefits for your team and your users:

  • Performance: Pre-generated HTML pages load nearly instantly, providing a better experience for users seeking quick answers.
  • SEO optimization: Because the pages are static, search engines can easily crawl and index every guide, making your help center easier to find via Google.
  • Reliable linking: Every page has a real, permanent URL. Your support team can share links to specific sections with confidence, knowing the content is always there.

Interactive API rendering

When DocuBot detects OpenAPI or Swagger specifications, it handles them differently than standard prose. Instead of just writing about your API, it integrates an interactive reference viewer. This allows developers to explore endpoints, view data models, and even test live API calls directly from your documentation site. These technical references are automatically placed in a dedicated section of your sidebar for easy access.

If you encounter issues with your documentation sync or need help refining your style settings, contact our team at support@ademero.com or visit https://www.ademero.com.