Custom ML Model / API Integration

Summary Add the ability to connect a user-supplied model or API so that transcription and processing run on the user's own private tokens.

Problem / Motivation Currently, all transcription and processing is handled through our default backend. Users who want greater control over cost, privacy, or model choice have no way to route this work through their own provider.

Proposed Solution Allow users to configure a custom model endpoint and supply their own API key/token. All transcription and processing requests would then be sent through that connection rather than the default service.