Namespace InferenceEngine.Core.ModelProviders
Classes
- DownloadingModelProvider
Model provider that downloads models from URLs with local caching. If the model already exists in the cache, returns the cached path immediately. Downloads are hardened (security assessment MEDIUM-1): model names are validated against path traversal, files are downloaded atomically via a temp file, an optional SHA-256 checksum is enforced, a size cap aborts runaway downloads, and redirects are capped.
- LocalModelProvider
Model provider for locally stored model files. Validates that the file exists but does not perform any network operations.
- ModelRegistry
Registry of known models and their download URLs from popular model hubs. Allows engines to specify only a model name for commonly known models.
- ModelRegistryData
Data class for the registry JSON structure.
- ModelRegistryEntry
Entry in the model registry.
Interfaces
- IModelProvider
Abstraction for model acquisition. Implementations can load models from local files, download from URLs, fetch from cloud storage, etc.