AI-Ready Output

Every oxidoc build generates machine-readable files that make your documentation consumable by AI tools, LLMs, and RAG pipelines — with zero configuration.

What Gets Generated

llms.txt

A lightweight index of every page — title and URL, one per line:

dist/llms.txttext

- /docs/installation: Install Oxidoc
- /docs/quickstart: Quickstart
- /docs/configuration: Configuration Reference
- /docs/components: Built-in Components
- /docs/search: Search

AI tools use this to understand what your docs cover and which page to fetch for a given question.

llms-full.txt

The complete plain text content of every page, separated by --- markers:

dist/llms-full.txttext

---
# Install Oxidoc (docs/installation)

The recommended way to install Oxidoc is the install script...

---
# Quickstart (docs/quickstart)

Go from zero to a running documentation site in under five minutes...

This is the file AI tools ingest when they need full context. A single fetch gives the LLM your entire documentation in a format optimized for comprehension.

Zero Configuration

Both files are generated automatically on every oxidoc build. No config flags, no opt-in — they're always there in dist/.

How AI Tools Use These Files

Tool	How it uses llms.txt
ChatGPT / Claude	Users paste `llms-full.txt` as context to ask questions about your project
Cursor / Codex	Agents fetch the file to understand your API before writing code
RAG pipelines	Index `llms-full.txt` as a document source for retrieval
AI search	Use `llms.txt` as a sitemap for selective page fetching

Link it from your README

Add a note in your project README pointing AI users to your llms.txt URL. For example: https://your-docs-site.com/llms-full.txt

The llms.txt Standard

The llms.txt format is an emerging convention for making websites machine-readable. By generating these files automatically, Oxidoc ensures your documentation is ready for the AI-native web without any extra work from you.

Other documentation tools require plugins or manual setup. Oxidoc does it out of the box.

Pre-Computed Embeddings

When semantic search is enabled (semantic = true), Oxidoc also generates search-vectors.json — a JSON file containing a 384-dimensional embedding vector for every page, along with full metadata (title, path, text, heading positions).

This file is a portable artifact you can feed directly into any vector database (ChromaDB, Pinecone, Weaviate, Qdrant) or custom RAG pipeline.

See Semantic Search > Embedding Output Format for the schema and code examples in Python and JavaScript.

All AI Output Files

File	Always Generated	Description
`llms.txt`	Yes	Page index — title and URL per line
`llms-full.txt`	Yes	Full plain text of all pages
`search-vectors.json`	When `semantic = true`	Pre-computed embedding vectors + metadata
`search-model.gguf`	When `semantic = true`	The GGUF embedding model (for re-embedding queries)

Together, these files make your documentation fully RAG-ready. Users can build custom AI assistants that understand your project without scraping HTML or parsing Markdown.