ModelManager

Interface for managing remote face recognition model downloads and caching. Provides functionality for downloading models from CDNs like Cloudflare R2, verifying checksums, and managing local cache.

Methods

downloadModel

Download a model from a remote URL with progress callbacks. Downloads to a temporary location, verifies the checksum, then moves to cache.

downloadModel(
  url: string,
  checksum: string,
): Promise<string>

Parameters

Name	Type	Description
url	string	Remote URL to download model from
checksum	string	Expected SHA-256 checksum (lowercase hex string, 64 chars)

Returns

• Promise<string> - Promise resolving to local file path of cached model

Throws

• Error - If download fails, checksum mismatch, or disk full

Example

const path = await ModelManager.downloadModel(
  'https://r2.cloudflare.com/models/Megatron',
  'a3f7c8d2e1b4f9a6...',
  (bytes, total) => console.log(`${bytes}/${total}`)
);

---

getCachedModel

Check if a model exists in cache and verify its checksum. If the cached file's checksum doesn't match, it's deleted and null is returned.

getCachedModel(checksum: string): string | null

Parameters

Name	Type	Description
checksum	string	SHA-256 checksum to look up

Returns

• string | null - Local file path if cached and valid, null otherwise

Example

const cachedPath = ModelManager.getCachedModel('a3f7c8d2...');
if (cachedPath) {
  // Model is cached and valid
  InspireFace.launch(cachedPath);
}