ModelCachePath

ModelCachePath is the folder where the engine stores downloaded model files. First-run downloads from Hugging Face (or the Aspose catalog) land here; subsequent runs load from this cache.

Quick reference


Type	`string?`
Default	`<LocalAppData>/Aspose.LLM/models`
Category	Engine
Field on	`EngineParameters.ModelCachePath`

What it does

On Windows: %LOCALAPPDATA%\Aspose.LLM\models.
On Linux / macOS: the equivalent Environment.SpecialFolder.LocalApplicationData path, joined with Aspose.LLM/models.
Overriding sets a custom directory.

At each AsposeLLMApi.Create, the engine checks this folder for the preset’s model file. If present, load. If absent, download from the resolved source (BaseModelSourceParameters), save into this folder, and load.

When to change it

Scenario	Value
Default	`null` or unset
Shared cache across multiple applications	`/srv/aspose-llm/models` (or equivalent)
Faster storage (SSD when `LocalAppData` is on HDD)	Path to SSD folder
Docker / container with mounted volume	`/models` (volume mount)
Pre-populated offline deployment	Path to the pre-populated cache

Example

var preset = new Qwen25Preset();
preset.EngineParameters.ModelCachePath = @"D:\models";
// Large models will land on drive D instead of C.

using var api = AsposeLLMApi.Create(preset);

Container pattern:

preset.EngineParameters.ModelCachePath = "/models";  // Docker volume mount

Interactions

BaseModelSourceParameters — resolution order; the cached file must match the requested source.
BinaryManagerParameters.BinaryPath — separate cache for native binaries; different folder.

What’s next

Offline deployment use case — pre-populating the cache.
Model source parameters — what gets cached.
Engine parameters hub — all engine knobs.

EnableDebugLogging