tomevault-io EchoSharp architecture skill
Use this when changing architecture, adding components, or reviewing code in this repository.
EchoSharp separates provider-neutral orchestration from provider-specific integrations:
src\EchoSharp\Audio: reusable audio sources and sample conversion utilities.src\EchoSharp\SpeechTranscription: STT contracts, realtime orchestration, and transcript models.src\EchoSharp\VoiceActivityDetection: VAD contracts and segment models.src\EchoSharp\Provisioning: model download, integrity, hashing, memory/file model storage, and unarchive helpers.components\*: provider-specific adapters that plug into core contracts.tests\EchoSharp.Tests: integration and unit coverage for core and component behavior.
When adding a component, mirror the closest existing component before inventing a new pattern. The usual shape is:
- Public config class for provisioner settings.
- Public options class for factory/runtime settings.
- Public factory implementing the core factory interface.
- Internal detector/transcriptor implementing the core runtime interface.
- Public provisioner implementing the matching provisioning interface.
- Static model catalog when models can be downloaded with integrity checks.
Keep dependencies flowing inward: components depend on core; core never depends on components.
Source: sandrohanea/echosharp — distributed by TomeVault.