Spotter

Spotter is a no-auth, synthetic-data V1 of an AI-assisted coaching system. It generates grounded training programmes, evaluates them with deterministic safety rules and independent quality checks, routes them through coach approval, and exposes only approved work to the client experience.

The implementation follows the canonical contract in DOCS/. The repository contains:

• frontend/: Next.js coach console and thin REST adapters.
• backend/: server-only domain, application, persistence, agent, evaluation, job, shared-package, infrastructure, deterministic-resource, and verification-tooling modules.
• client-frontend/: Expo client application.

Backend-owned supporting assets are grouped by responsibility:

• backend/packages/: versioned cross-surface contracts, configuration, design tokens, and test utilities.
• backend/resources/: committed synthetic demo, real-model recordings, pending calibration labels, and deterministic safety data.
• backend/infra/: PostgreSQL initialization and verification schema declarations.
• backend/tooling/: repository verification and operational commands.

Generated reports, coverage, local state, dependencies, and credentials remain untracked. No real client data or credentials are required for the deterministic V1 path.

For Docker startup, Expo launch, hands-on acceptance, automated E2E commands, and cleanup, follow the local and staging playbook.

Product screenshots

These are application-build screens rendered with deterministic synthetic fixtures.

Coach console

See the coach console gallery for client context, traceability, and evaluation evidence.

Client experience

See the client experience gallery for client selection and session logging.