File Formats¶

RavenRustRAG supports loading documents in several formats. The loader is selected automatically based on file extension.

Supported Formats¶

Extension	Loader	Notes
`.txt`	Text	Plain text, also used as fallback for unknown extensions
`.md`	Markdown	YAML frontmatter parsed into metadata
`.csv`	CSV	Each row becomes a document; headers become metadata keys
`.json`	JSON	Pretty-printed content for chunking
`.jsonl`	JSONL	Each line treated as a separate record
`.html`	HTML	Tags stripped, scripts/styles removed

Markdown files are loaded with full content preserved. If the file contains YAML frontmatter, it is parsed into document metadata:

---
title: My Document
author: John Doe
tags: [rust, rag]
---

# Content starts here

The frontmatter fields (title, author, tags) become searchable metadata.

Each row in a CSV file becomes a separate document. Column headers are used as metadata keys:

title,content,category
Introduction,Welcome to the guide,basics
Setup,Install with cargo,getting-started

JSON files are pretty-printed and treated as a single document for chunking. Nested structures are preserved in the text representation.

HTML files have all tags stripped. <script> and <style> blocks are completely removed. The resulting plain text is then chunked normally.

When indexing a directory, use --extensions to limit which files are processed:

# Only index markdown and text files
ravenrag index ./docs --extensions md,txt

# Include HTML files
ravenrag index ./web --extensions html,md,txt

The default extensions are txt,md.