LLMs.txt

MinimalDoc generates LLM-friendly output for AI assistants.

Generated Files

File	Purpose
llms.txt	Index with links to individual pages
llms-full.txt	All documentation in a single file
*.md	Companion markdown files alongside HTML

Purpose

LLM-friendly format enables:

• AI chatbots to answer questions about your project
• IDE assistants (Copilot, Cursor) to understand your docs
• Custom AI integrations
• Documentation search tools

Configuration

Enabled by default:

enable_llms: true

Disable:

enable_llms: false

Or via CLI:

minimaldoc build --llms=false

File Location

Generated at output root:

public/
├── index.html
├── llms.txt              # Index with page links
├── llms-full.txt         # All content in one file
├── getting-started/
│   ├── installation.html
│   └── installation.md   # Companion markdown
├── sitemap.xml
└── ...

llms.txt Format

Index file with links to documentation:

# Site Title

> Site description

## Documentation

- [Installation](/getting-started/installation.md): Install MinimalDoc
- [Quick Start](/getting-started/quick-start.md): Create your first site

## Optional

- [Complete Documentation](/llms-full.txt): All documentation in a single file

llms-full.txt Format

All content concatenated:

# Site Title

> Site description

---

# Installation

Full page content here...

---

# Quick Start

Full page content here...

Companion .md Files

Each HTML page gets a markdown companion for direct LLM access:

/getting-started/installation.html  →  /getting-started/installation.md

Use Cases

ChatGPT / Claude

Upload llms.txt as context:

Based on the attached documentation, how do I configure dark mode?

GitHub Copilot

Reference in your codebase:

// See llms.txt for MinimalDoc configuration options
const config = {
  // ...
};

Custom AI Integration

import openai

with open('llms.txt', 'r') as f:
    docs = f.read()

response = openai.ChatCompletion.create(
    model="gpt-4",
    messages=[
        {"role": "system", "content": f"Documentation:\n{docs}"},
        {"role": "user", "content": "How do I add a new page?"}
    ]
)

RAG (Retrieval Augmented Generation)

Split and embed for vector search:

from langchain.text_splitter import MarkdownTextSplitter
from langchain.vectorstores import Chroma

splitter = MarkdownTextSplitter(chunk_size=1000)
chunks = splitter.split_text(docs)

vectorstore = Chroma.from_texts(chunks, embeddings)

Optimization

Exclude Content

Hidden pages are excluded:

---
title: Internal Notes
hidden: true
---

Prioritize Content

Structure pages so important content appears early:

• Clear, concise introductions
• Key information first
• Examples after explanations

Token Efficiency

LLM context windows are limited. Keep documentation:

• Focused and relevant
• Free of redundancy
• Well-organized

File Size

llms-full.txt size depends on documentation volume. Most LLMs handle 100KB+ easily.

Comparison

Format	Use Case
llms.txt	Index with links to pages
llms-full.txt	All content for full context
*.md companions	Individual page access
sitemap.xml	Search engines
search-index.json	Client-side search