Module model_service 

Model management service for downloading and caching AI models from HuggingFace Hub

This service provides a unified interface for:

• Downloading models from HuggingFace Hub using the model catalog
• Caching models to the project’s models/ directory
• Listing available and installed models
• Managing model lifecycle (pull, remove, info)

Design

• Models are defined in model_catalog.toml embedded in the CLI binary
• Models are downloaded from HuggingFace Hub on first use (lazy loading)
• Models are cached to <project_root>/models/<name>.onnx
• HF cache is also used (~/.cache/huggingface/hub/) for offline support

Usage

use mecha10_cli::services::ModelService;

let service = ModelService::new()?;

// List catalog models
let catalog = service.list_catalog()?;

// Download a model
service.pull("yolov8n", None)?;

// List installed models
let installed = service.list_installed()?;

Structs

InstalledModel

Installed model metadata

ModelCatalogEntry

Model catalog entry from model_catalog.toml

ModelInfo

Combined model information (catalog + installed)

ModelService

Model management service

QuantizeConfig

Quantization configuration for automatic INT8 conversion

Enums

PreprocessingPreset

Preprocessing presets for common model families