Reference

Reference for Modelfile and MIR specifications.

Modelfile Reference

Main Structure

Represents a parsed Modelfile:

pub struct Modelfile {
    pub from: Option<String>,
    pub parameters: Vec<Parameter>,
    pub template: Option<String>,
    pub adapter: Option<String>,
    pub system: Option<String>,
    pub license: Option<String>,
    pub messages: Vec<Message>,
    pub data: Vec<String>,      // Original lines for display
    pub errors: Vec<String>,    // Parsing errors
}

Parameter

pub struct Parameter {
    pub param_type: String,
    pub value: ParamValue,
}

ParamValue

pub enum ParamValue {
    Int(i32),
    Float(f32),
    Str(String),
}

Message

pub struct Message {
    role: Role,
    message: String,
}

API

Parsing Functions

parse(input: &str) -> Result<Modelfile, String>
Parses a Modelfile from a string. Returns a Modelfile struct on success, or an error string on failure.
parse_from_file(path: &str) -> Result<Modelfile, String>
Parses a Modelfile from a file path. Reads the file and calls parse() on its contents.

Builder API

new() -> Modelfile — Creates a new empty Modelfile
add_from(value: &str) -> Result<(), String> — Adds a FROM instruction
add_parameter(param_type: &str, param_value: &str) -> Result<(), String> — Adds a PARAMETER
add_template(value: &str) -> Result<(), String> — Adds a TEMPLATE instruction
add_system(value: &str) -> Result<(), String> — Adds a SYSTEM instruction
add_adapter(value: &str) -> Result<(), String> — Adds an ADAPTER instruction
add_license(value: &str) -> Result<(), String> — Adds a LICENSE instruction
add_message(role: &str, message: &str) -> Result<(), String> — Adds a MESSAGE instruction
add_comment(value: &str) -> Result<(), String> — Adds a comment
build() -> Result<(), String> — Validates the Modelfile (checks for required FROM)

Traits

FromStr — Allows parsing from string: modelfile.parse()
Display — Allows conversion to string: modelfile.to_string()
Default — Provides default empty Modelfile

Validation Rules

Required FROM: A Modelfile must contain exactly one FROM instruction.
Single-instruction limits: The following can only appear once: FROM, TEMPLATE, SYSTEM, ADAPTER, LICENSE.
Parameter validation: Parameters are validated for correct types (int/float/string).
Role validation: MESSAGE roles must be one of: system, user, assistant.
Case insensitivity: All instruction and parameter names are case-insensitive.

Error Handling

Parsing errors are collected in the errors field.
parse() returns an error string containing all error messages if any errors occur.
The parser continues parsing after errors to collect all issues.

Implementation Details

Uses the nom parser combinator library for parsing.
Instructions are parsed case-insensitively.
Whitespace is trimmed from argument values.
The parser preserves original lines in the data field for display.
Comments are preserved when converting back to string.

Example Usage

Parsing from file

use tiles::core::modelfile::parse_from_file;
 
let modelfile = parse_from_file("path/to/Modelfile")?;

Parsing from string

use tiles::core::modelfile::parse;
 
let content = r#"
FROM llama3.2
PARAMETER temperature 0.7
SYSTEM "You are helpful"
"#;
 
let modelfile = parse(content)?;

Building programmatically

use tiles::core::modelfile::Modelfile;
 
let mut modelfile = Modelfile::new();
modelfile.add_from("llama3.2")?;
modelfile.add_parameter("temperature", "0.7")?;
modelfile.add_system("You are helpful")?;
modelfile.build()?;

Notes

Implementation is based on the Ollama Modelfile specification.
Not all Ollama Modelfile features may be implemented.
The parser is compatible with Ollama-generated Modelfiles.
Future enhancements may validate GGUF file paths in FROM instructions.

MIR Reference

Copied verbatim from: github.com/darkshapes/MIR (opens in a new tab):

MIR (Machine Intelligence Resource)

A naming schema for AIGC/ML work.

The MIR classification format seeks to standardize and complete a hyperlinked network of model information, improving accessibility and reproducibility across the AI community.

Example:

mir : model . transformer . clip-l : stable-diffusion-xl

 mir : model .    lora      .    hyper    :   flux-1
  ↑      ↑         ↑               ↑            ↑
 [URI]:[Domain].[Architecture].[Series]:[Compatibility]

Code for this project can be found at darkshapes/MIR (opens in a new tab) on GitHub.

Definitions

Like other URI schema, the order of the identifiers roughly indicates their specificity from left (broad) to right (narrow).

DOMAINS

↑Most Specific/Decentralized

Dev

Pre-release or under evaluation items without an identifier in an expected format
Anything in in-training, pre-public release, and items under evaluation

Meant to be created by anyone, derived from code and file analysis

Contextual
Layers of neural networks
Dynamic

Model

Publicly released machine learning models with an identifier in the database
Model weight tensors with arbitrary locations and quantitative dimensions

Meant to be created by file hosts, derived from research pre-prints

Contextual
Layers of neural networks
Fixed

Ops

References to specific optimization or manipulation techniques
Algorithms, optimizations and procedures for models

Meant to be created by code libraries, derived from research pre-prints

Universal
Attributes of neural networks
Dynamic

Info

Metadata of layer names or settings with an identifier in the database
Information about the model and tensor specifications

Meant to be created by standards community, derived from code and file analysis

Universal
Attributes of neural networks
Fixed

↓Most General/Centralized

ARCHITECTURE

Broad and general terms for system architectures:

Abbreviation	Description
GRU	Gated recurrent unit
RBM	Restricted Boltzmann machine
TAE	Tiny Autoencoder
VAE	Variable Autoencoder
LSTM	Long Short-Term Memory
RESNET	Residual Network
CNN	Convolutional Neural Network
RCNN	Region-based Convolutional Neural Network
RNN	Recurrent Neural Network
BRNN	Bi-directional Recurrent Neural Network
GAN	Generative Adversarial Model
SSM	State-Space Model
DETR	Detection Transformer
VIT	Vision Transformer
MOE	Mixture of Experts
AET	Autoencoding Transformer
STST	Sequence-to-Sequence Transformer
ART	Autoregressive Transformer
LORA	Low-Rank Adaptation
CONTROLNET	Controlnet
UNCLASSIFIED	Unknown

SERIES

Foundational network and technique types.

Rules

Lowercase, hyphen only
Remove parameter size, non-breaking semantic versioning, library names

Example: tencent-hunyuan/hunyuandiT-v1.2-diffusers
SERIES : hunyuandit-v1

Example: black-forest-labs/FLUX.1-dev
SERIES : flux1dev

In regex (roughly):

BREAKING*SUFFIX = r".*(?:-)(prior)$|.*(?:-)(diffusers)$|.\*[\*-](\d{3,4}px|-T2V$|-I2V$)"
PARAMETERS*SUFFIX = r"(\d{1,4}[KkMmBb]|[.*-]\d+[\._-]\d+[Bb][._-]).\*?$"
SEARCH*SUFFIX = r"\d+[.*-]?\d+[BbMmKk](it)?|[._-]\d+[BbMmKk](it)?"

COMPATIBILITY

Implementation details based on version-breaking changes, configuration inconsistencies, or other conflicting indicators that have practical application.

Rules

An additional SERIES label for identifying cross-compatibility

Notes

If you would like to regenerate or update the example file here, use nnll (opens in a new tab):

MIR is inspired by:

AIR-URN (opens in a new tab) project by CivitAI (opens in a new tab)
Spandrel (opens in a new tab) super-resolution registry by chaiNNer (opens in a new tab)
SDWebUI Model Toolkit (opens in a new tab) by silveroxides (opens in a new tab)

Implementation