Key Principles of Documentation Creation
Publication date: 2025-02-12Terminology
- Documentation audience — a group of people and/or language models (LLMs) who use documentation to form cognitive context. Includes end users, developers, analysts, and other stakeholders.
- Cognitive context — the combination of knowledge, experience, expectations, and motivation that determines information perception and interpretation. Includes both prior knowledge and current goals.
- Documentation subject — the product, process, or concept being described in the document: software, methodology, business process, etc.
Introduction
Documentation is the process of synchronizing the cognitive context of the audience (people and LLMs). The goal is to provide a comprehensive understanding of the subject, starting with general concepts and moving to details. Information is presented hierarchically, without duplication, allowing both quick familiarization and deep dive.
Reader's Context Immersion
The documentation structure should facilitate the reader's quick immersion into the cognitive context of the subject. This is achieved through logic, sequence, and sufficient depth of presentation.
Hierarchy and Structure
Information is presented by levels: from general to specific. Each level complements the previous one and forms a coherent sequence focused on the reader's current tasks.
Documentation Levels
- Title — a brief name reflecting the essence.
- Summary — a very brief description of the meaning (mental anchor).
- Introduction — document purpose, its audience, context setting.
- Overview — a brief summary allowing relevance assessment.
- Main section — detailed description, structured by subject area.
Volume Recommendations
- Title: 5–10 words, maximum < 20
- Summary: 30–50 words, maximum < 150
- Introduction: 10–20 words, maximum < 50
- Overview: 50–100 words, maximum < 250
- Main text: 1000–3000 tokens, maximum < 8000 (tokenizer)
Navigation Accessibility
Documentation is an interconnected structure where each document covers one aspect. Hyperlinks, tables of contents, and cross-references should aid navigation. The reader should always understand their position in the structure and have the ability to jump to related topics or return to the overview.
LLM Orientation
Documentation should account for language models' characteristics and be formatted in Markdown. Key information should be at the beginning of sections. It's important to use consistent terminology and avoid ambiguity. The structure should be logical, hierarchical, with clear heading and list markup. Sections should be self-contained but interconnected.