CI (SIAX Cloud) / integration (push) Failing after 22s
CI (SIAX Cloud) / security (push) Successful in 22s
Security Scan (SIAX Cloud) / security (push) Successful in 21s
CI (SIAX Cloud) / quality (push) Successful in 43s
CI (SIAX Cloud) / integration (pull_request) Failing after 3s
CI (SIAX Cloud) / security (pull_request) Successful in 17s
Security Scan (SIAX Cloud) / security (pull_request) Successful in 16s
CI (SIAX Cloud) / quality (pull_request) Successful in 43s
- tests/integration/: 10 tests against REAL OpenSearch 2.11 (tenant isolation, purpose/rights filters, generation rebuild/rollback, projection-disaster rebuild §68) + REAL ClickHouse 24.8 (tenant-isolated metric DSL, TTL schema, injection-surface refusals). Honest skip without env - and CI fails hard if the suites skip instead of running. - .gitea/workflows/ci.yml: new blocking 'integration' job with self-identifying service containers (st0re ci.yml's measured alias-collision discipline: run-unique cluster.name / credentials, getent-enumerated IP selection, no-match = hard fail). - STATUS.md: full 8-level Master-DoD classification per capability incl. the honest gaps blocking PRODUCTION_READY. - dependency-mode: NEXT_PHASE=phase-production-build is not a serving runtime (next build page-data collection must compile; guard re-arms at boot) - pinned by contract test. - docs/02-architecture.md: superseded the accurate-but-now-fixed 'every client runs against its mock' paragraph; .env.example: client URLs now actually read from env + new P0 backend vars. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
91 lines
3.7 KiB
TypeScript
91 lines
3.7 KiB
TypeScript
/**
|
|
* src/platform/dependency-mode.ts
|
|
* ================================
|
|
* THE production mock-fallback kill switch (Master-DoD §3: "Ingen
|
|
* production-path faller tyst tillbaka till mock/in-memory", §10: "Production
|
|
* kan inte falla tillbaka till mock").
|
|
*
|
|
* Before this module, every platform client in src/clients/* documented the
|
|
* same convention: "when no base URL is configured the client runs in
|
|
* LOCAL/MOCK/NOOP mode". That convention is fine for unit tests and local
|
|
* dev, and it is EXACTLY the silent-fallback failure mode the Master-DoD
|
|
* forbids in production. This module centralizes the decision so every
|
|
* factory makes it the same way:
|
|
*
|
|
* - production runtime + no URL → throw SiaxError DEPENDENCY_UNAVAILABLE
|
|
* at CONSTRUCTION time (fail hard, fail early — not on first call).
|
|
* - non-production + no URL → the factory may return its local/mock
|
|
* implementation (hermetic tests, local dev).
|
|
* - URL configured → real HTTP client, any runtime.
|
|
*
|
|
* "Production runtime" means NODE_ENV=production, the same switch the
|
|
* existing auth layer uses (src/auth/context.ts refuses AUTH_ENFORCE=false
|
|
* under NODE_ENV=production). There is deliberately NO override that lets
|
|
* production run on mocks: if a deployment genuinely wants a dependency off,
|
|
* that is a topology decision expressed by NOT calling the capability, never
|
|
* by a mock silently answering it.
|
|
*/
|
|
import { SiaxError } from "./errors";
|
|
|
|
/**
|
|
* True when this process is a production runtime — i.e. it can SERVE
|
|
* requests. `next build` runs page-data collection under NODE_ENV=production
|
|
* while serving nothing; failing the BUILD for unset dependency URLs would
|
|
* block the compile step, not protect any request path, so the Next build
|
|
* phase (NEXT_PHASE=phase-production-build) is explicitly not a production
|
|
* runtime. The guard re-arms the moment the built app boots (`next start`
|
|
* clears NEXT_PHASE).
|
|
*/
|
|
export function isProductionRuntime(env: NodeJS.ProcessEnv = process.env): boolean {
|
|
if (env.NEXT_PHASE === "phase-production-build") return false;
|
|
return env.NODE_ENV === "production";
|
|
}
|
|
|
|
/**
|
|
* Resolve a dependency base URL for a client factory.
|
|
*
|
|
* Returns the URL (trimmed, no trailing slash) when configured. Returns
|
|
* `undefined` when unconfigured in a non-production runtime — the caller may
|
|
* then use its documented local/mock implementation. THROWS
|
|
* `SiaxError("DEPENDENCY_UNAVAILABLE")` when unconfigured in production.
|
|
*/
|
|
export function resolveDependencyBaseUrl(
|
|
dependency: string,
|
|
configuredUrl: string | undefined,
|
|
env: NodeJS.ProcessEnv = process.env
|
|
): string | undefined {
|
|
const url = configuredUrl?.trim();
|
|
if (url) return url.replace(/\/+$/, "");
|
|
if (isProductionRuntime(env)) {
|
|
throw new SiaxError(
|
|
"DEPENDENCY_UNAVAILABLE",
|
|
dependency,
|
|
`no base URL configured for required dependency "${dependency}" in a production runtime; ` +
|
|
`refusing to fall back to a mock/in-memory implementation (Master-DoD §3/§10)`
|
|
);
|
|
}
|
|
return undefined;
|
|
}
|
|
|
|
/**
|
|
* Same decision for stateful stores (Postgres-backed registries): production
|
|
* with no DATABASE_URL must fail hard rather than serve an in-memory Map.
|
|
*/
|
|
export function resolveStoreMode(
|
|
store: string,
|
|
databaseUrl: string | undefined,
|
|
env: NodeJS.ProcessEnv = process.env
|
|
): { mode: "postgres"; url: string } | { mode: "memory" } {
|
|
const url = databaseUrl?.trim();
|
|
if (url) return { mode: "postgres", url };
|
|
if (isProductionRuntime(env)) {
|
|
throw new SiaxError(
|
|
"DEPENDENCY_UNAVAILABLE",
|
|
store,
|
|
`no database configured for persistent store "${store}" in a production runtime; ` +
|
|
`refusing the in-memory fallback (Master-DoD §3/§10)`
|
|
);
|
|
}
|
|
return { mode: "memory" };
|
|
}
|