---
title: "LLMind Python CLI — LRFS implementation | LLMind"
description: "LLMind Python CLI: an LRFS-conformant implementation in Python. L3-full · 0/10 vectors · MIT."
url: https://llmind.org/spec/implementations/llmind-python/
source_format: html
---
# LLMind Python CLI

Python · L3-full · 0/10 vectors · MIT

Status: **stable** · released 2026-04-09 · verified 2026-04-24 · [repository](https://github.com/dmitryrollins/LLMind)

LLMind Python CLI is the first implementation of the [LLM-Ready File Specification (LRFS)](https://llmind.org/spec/). It is a command-line tool (`pipx install 'llmind-cli[all]'`, `llmind <command>`) that reads and writes the full LRFS payload format, produces HMAC-SHA256 and ed25519 signatures per the signing scheme, and embeds structured metadata layers directly into PDF, JPEG, PNG, and audio (MP3/WAV/M4A) files as XMP metadata — following the conformance-level behaviors defined in [/spec/conformance/](https://llmind.org/spec/conformance/).

## Location

The CLI lives in the `llmind-cli/` directory on the [`feature/llmind-cli` branch](https://github.com/dmitryrollins/LLMind/tree/feature/llmind-cli/llmind-cli) of the LLMind repository. See `pyproject.toml` for the package definition (name: `llmind-cli`, entry point: `llmind = llmind.cli:main`).

## Install

Via PyPI (end users) — pipx is the recommended installer because it sandboxes the CLI in its own virtualenv:

```
pipx install 'llmind-cli[all]'
```

Available extras let you choose which vision/transcription provider gets bundled: `[anthropic]`, `[openai]`, `[gemini]`, `[whisper-local]`, or `[all]`. Requires Python 3.11+. PDF rendering also needs `poppler-utils` on the system (`brew install poppler` on macOS, `apt install poppler-utils` on Debian/Ubuntu).

For contributors, clone the LLMind repository and run `pip install -e '.[all,dev]'` from the `llmind-cli/` directory.

## What it does

LLMind takes a standard PDF, image, or audio file and enriches it with a structured metadata layer containing extracted text, a natural-language content description, and document/audio structure — all embedded as XMP metadata inside the file’s binary with no external database or sidecar file. Each enrichment layer is cryptographically signed with HMAC-SHA256 (default) or ed25519 (optional) and bound to the file via SHA-256 checksum.

## Conformance

-   L1 (reader): claimed — parses any LRFS payload, validates signatures, extracts layers.
-   L2 (writer): claimed — writes new layers with valid RDF canonicalization and HMAC-SHA256 signatures.
-   L3 (full): claimed — implements the full version-chain algorithm (immutable layer history) per [/spec/conformance/](https://llmind.org/spec/conformance/).

Conformance self-verification pending — the implementation has not yet been formally re-run against the [LRFS test vectors](https://llmind.org/spec/test-vectors/) and `test_vectors_passing` is `0` until that check is performed. The `conformance_level` reflects the implementation’s design intent, not a verified pass count. When verification is run and passes, this entry will be updated to `status: reference` and `test_vectors_passing: 10`.

## Related

-   [LLM-Ready File Specification](https://llmind.org/spec/)
-   [Conformance levels](https://llmind.org/spec/conformance/)
-   [LRFS test vectors](https://llmind.org/spec/test-vectors/)
-   [All implementations](https://llmind.org/spec/implementations/)
-   [LRFS — glossary entry](https://llmind.org/glossary/lrfs/)