---
title: "Learn: file enrichment, LLM-ready files, semantic metadata | LLMind"
description: "Concepts, how-tos, and explainers on file enrichment, LLM-ready files, and signed semantic metadata. Written for developers shipping AI products."
url: https://llmind.org/learn/
source_format: html
---
# Learn

Concepts, how-tos, and explainers for developers shipping AI products.

-   ## [What is file enrichment?](https://llmind.org/learn/what-is-file-enrichment/)
    
    The AI-ready file pattern explained — how enrichment differs from OCR, parsing, and RAG.
    
-   ## [What is an LLM-ready file?](https://llmind.org/learn/llm-ready-files/)
    
    The file format property that lets any AI tool read meaning without re-parsing.
    
-   ## [Enrichment vs. chunking](https://llmind.org/learn/enrichment-vs-chunking/)
    
    Why embedding semantic metadata inside a file is not the same as splitting it for a vector DB.
    
-   ## [OCR once, read forever](https://llmind.org/learn/ocr-once-read-forever/)
    
    Stop paying to OCR the same file every time a new AI tool touches it.
    
-   ## [Semantic layer for files](https://llmind.org/learn/semantic-layer-files/)
    
    Borrows from BI semantic layers (dbt, Looker) — structured meaning that travels with every file.
    
-   ## [Enrichment vs parsing](https://llmind.org/learn/enrichment-vs-parsing/)
    
    Parsing extracts. Enrichment persists. LLMind lives at the layer above parsers.
    
-   ## [XMP metadata in Python](https://llmind.org/learn/xmp-metadata-python/)
    
    Read and write XMP metadata programmatically. Examples using LLMind and Piexif.
    
-   ## [Custom XMP namespace](https://llmind.org/learn/custom-xmp-namespace/)
    
    Define your own XMP metadata namespace. Portable, file-native, tamper-evident.
    
-   ## [How to give Claude access to files](https://llmind.org/learn/give-claude-access-to-files/)
    
    Three ways: upload in chat, filesystem MCP, or LLMind-enriched directory. Tradeoffs for each.
    
-   ## [AI agent file access patterns](https://llmind.org/learn/agent-file-access-patterns/)
    
    4 patterns compared: context dump, RAG stack, MCP-only, and MCP + enriched files.
    

## Explore more

-   [Use-cases](https://llmind.org/use-cases/)
-   [Glossary](https://llmind.org/glossary/)
-   [Spec](https://llmind.org/spec/)