Self-hosted diagnostics and executable E2E harness
Origin
This proposal follows the self-hosted server, Aone provider, evidence, webhook, runner and browser E2E dogfooding performed for #160, #161, #162 and implementation PR #163. The resulting durable behavior is archived by #165.
Why
Validating a self-hosted deployment currently requires manually composing PostgreSQL, server configuration, operator provider registries, profiles, source bindings, external code evidence, webhook receivers, browser sessions and platform-specific tests. Failures such as API/Web origin mismatch, provider-registry precedence, stale evidence, webhook suppression or an inaccessible browser route are difficult to explain without reading databases and implementation code.
What Changes
- Add profile and onboarding diagnostics that explain server identity, origins, transport posture, permissions, provider registry and planned init mutations.
- Add gate-oriented evidence explanation and synchronization diagnostics for missing, stale, wrong-revision or unauthorized evidence.
- Provide a reproducible self-hosted E2E environment lifecycle with seed data, runner, provider and browser suites and durable evidence output.
- Add webhook test delivery and policy explanation using a built-in capture fixture.
- Define platform-aware required checks and route/security matrices so local and hosted evidence can be evaluated consistently at an exact revision.
Non-Goals
- Compiling vendor-specific source or CI adapters into core.
- Shipping a hosted testing service.
- Persisting real provider credentials in fixtures or evidence bundles.
- Replacing focused unit, integration, race or browser tests.
Compatibility
Diagnostics are additive and read-only unless the user explicitly runs init, evidence synchronization or E2E seed operations. Provider-specific behavior continues through operator-registered neutral adapters.
Success Criteria
- Operators can explain profile, origin, realm, permission and provider failures before init mutates state.
- Verify failures identify the exact missing or unusable evidence fact and remediation path.
- A clean checkout can start, seed, exercise and tear down the production self-hosted stack reproducibly.
- Webhook filtering, signing, redaction, retry and replay can be tested without a custom receiver.
- Browser route, public/private authorization and platform-specific check coverage are emitted as exact-revision evidence.
Self-hosted diagnostics and executable E2E harness
Origin
This proposal follows the self-hosted server, Aone provider, evidence, webhook, runner and browser E2E dogfooding performed for #160, #161, #162 and implementation PR #163. The resulting durable behavior is archived by #165.
Why
Validating a self-hosted deployment currently requires manually composing PostgreSQL, server configuration, operator provider registries, profiles, source bindings, external code evidence, webhook receivers, browser sessions and platform-specific tests. Failures such as API/Web origin mismatch, provider-registry precedence, stale evidence, webhook suppression or an inaccessible browser route are difficult to explain without reading databases and implementation code.
What Changes
Non-Goals
Compatibility
Diagnostics are additive and read-only unless the user explicitly runs init, evidence synchronization or E2E seed operations. Provider-specific behavior continues through operator-registered neutral adapters.
Success Criteria