LRFS

LRFS (LLM-Ready File Specification) is LLMind's published open specification for embedding semantic metadata inside files and cryptographically signing it. It defines exactly how metadata is structured in RDF/XML, stored in XMP packets, canonicalized, and signed so that any compliant reader — whether built by LLMind or a third party — can verify and extract the metadata. LRFS is language-agnostic and format-agnostic: it works with JPEG, PNG, PDF, TIFF, audio, video, and any format that supports XMP.

What LRFS defines

The spec covers the XMP packet structure, the custom namespace (https://llmind.org/ns/1.0/), the RDF schema for each semantic layer (description, entities, structure, transcription, lineage), and the signing scheme. For signing, LRFS defaults to HMAC-SHA256 for private verification and supports ed25519 for public-key scenarios. The spec also defines canonicalization rules so that identical metadata produces identical signatures across tools and languages, and it specifies conformance levels so partial implementations can still interoperate.

Versioning and adoption

LRFS v1.x is the current major version. Breaking changes require a v2.0 release with a new namespace URI, preserving backward compatibility for legacy files. By publishing an open spec and maintaining backward compatibility, LLMind enables third parties to build conformant readers and writers — turning LRFS from a single tool into a category. Organizations can enrich their own files without vendor lock-in.