Enfyra Server v2.2.3

Enfyra Server v2.2.3

Breaking Changes

The file public-access field on enfyra_file changed from isPublished to isPublic. Existing databases are migrated through data/snapshot-migration.json, but dynamic scripts, API clients, GraphQL selections, file upload/update payloads, and integrations that read or write enfyra_file.isPublished must migrate to enfyra_file.isPublic.

The server now requires Node 24.x for install, build, start, dev, and test commands. package.json, .node-version, .nvmrc, and scripts/check-node-version.js enforce the runtime so environments on older or newer Node majors must switch to Node 24 before running Enfyra Server.

POST /auth/token/exchange now returns short-lived access JWTs capped at 60 seconds or the API token expiration, whichever is earlier. Clients that treated exchanged JWTs as long-lived credentials must re-exchange the API token when the access JWT expires.

Features

• Added process-local user and role caching in src/shared/utils/load-user-with-role.util.ts, reducing repeated database/cache reads across HTTP, GraphQL, WebSocket, OAuth, and private file access paths.
• Added overlap-aware metadata migration in src/engines/bootstrap/services/metadata-migration.service.ts and src/engines/bootstrap/utils/metadata-physical-migration.util.ts, allowing legacy and canonical system rows to merge while preserving custom fields.
• Added static admin route metadata to data/data-migration.json for reload, script validation, test run, flow trigger, and Redis admin endpoints so upgraded databases receive the same built-in route definitions.
• Added Node and Yarn runtime metadata in package.json, .node-version, .nvmrc, and .yarnrc.yml to align local, CI, and production execution.

Operational Flow

Before: API token exchange could issue an access JWT whose expiration followed the API token itself. After: each exchange returns a short-lived access JWT, so API-token clients keep the long-lived token private and request fresh access credentials as needed.

Before: user and role snapshots were repeatedly loaded through shared cache paths during authenticated request handling. After: each server process keeps a short local snapshot and invalidates it through the existing revocation flow.

Before: metadata migration stopped or risked losing custom values when legacy and canonical system tables or collections both existed. After: migration reconciles overlapping rows, backfills legacy-only fields, and keeps canonical records as the active target.

Before: existing databases could miss metadata for built-in admin endpoints such as /admin/script/validate or /admin/test/run. After: data/data-migration.json seeds those route records during upgrade.

Bug Fixes

• Fixed src/http/middlewares/dynamic-interceptor.middleware.ts so dynamic success post-hooks no longer run on error responses from built-in routes.
• Fixed src/modules/me/services/me.service.ts to use secure repository access for /me and OAuth account reads, applying field-permission scoping to self-service profile responses.
• Fixed file route and file-management payload handling in src/http/routes/file.routes.ts and src/modules/file-management/services/file-management.service.ts to use the isPublic field consistently.
• Improved metadata migration test coverage in test/domain/metadata-migration.service.spec.ts, test/domain/metadata-migration.real-db.spec.ts, and test/domain/metadata-physical-migration.util.spec.ts for overlapping SQL and MongoDB upgrade cases.