Environment-Scoped Routing

Route REST, metadata, automation, AI, and package calls through /api/v1/environments/:environmentId/....

Environment-scoped routing makes the target runtime explicit in the URL:

/api/v1/environments/:environmentId/data/:object
/api/v1/environments/:environmentId/meta
/api/v1/environments/:environmentId/automation/...
/api/v1/environments/:environmentId/ai/...

The scoped URL wins over hostname, header, session, and default-environment resolution. Use it when a caller must address a specific environment without depending on request context.

Server configuration

Environment scoping is a host decision, not an application one. The two keys are declared on the stack's top-level api block, but the host that boots the stack is what decides whether they take effect:

import { defineStack } from '@objectstack/spec';

export default defineStack({
  // ...your manifest, objects, apis, etc.
  api: {
    enableProjectScoping: true,
    projectResolution: 'auto',
  },
});

On the default os serve path this block does not turn scoping on. A bare defineStack() config carries no instantiated plugins, so the CLI boots createStandaloneStack() and merges that boot result over the authored config (mergeBootConfig). The boot builder ships api: { enableProjectScoping: false, projectResolution: 'auto' } and wins on exactly those two keys — deliberately, because scoping is not the author's call on a standalone host.

The two keys are not overridden the same way:

enableProjectScoping: true is contradicted. The value forwarded to the REST and dispatcher plugins is false, and no scoped routes are registered.
projectResolution: 'auto' is redundant. The boot builder pins the same value, so writing it changes nothing; declaring any other strategy here is silently replaced by 'auto'.

Every other authored api key — enforceProjectMembership, for example — survives the merge untouched.

The block above is live when the CLI does not boot the standalone stack: when the exported config is host-shaped, meaning its plugins array already carries instantiated plugin objects. That is the shape a multi-environment host assembles, and this open-core CLI supports the standalone boot mode only — cloud / multi-environment hosts ship from a separate distribution. Setting OS_MODE=off (or bootMode: 'off' on the exported config) also skips the standalone boot, but it drops the CLI to its legacy lightweight assembler rather than producing a scoped standalone host; bootMode is additionally not a declared stack key, so defineStack() rejects it as an unrecognized key.

api is a declared top-level field on ObjectStackDefinitionSchema, so it survives defineStack's strict parsing — you can pass it directly inside the defineStack({ ... }) call as shown above. (Older stacks that instead spread it onto the exported config object, e.g. export default { ...stack, api: {...} }, still work the same way.) The CLI reads the resolved value from the exported config (config.api) when registering the REST and dispatcher plugins — but it reads it after the boot result has been merged in, which is why the standalone path forwards the boot builder's scoping decision rather than the author's.

The option names are historical for compatibility with existing config files; the route, header, env var, and request context all use environment.

Once scoping is enabled by the host, projectResolution selects the route surface:

Strategy	Behavior	When to use
`auto`	Registers both unscoped `/api/v1/...` and scoped `/api/v1/environments/:environmentId/...` routes.	Default migration mode.
`optional`	Same route surface as `auto`; resolution may come from URL, hostname, header, session, or default.	Multi-environment hosts that still accept unscoped callers.
`required`	Registers only environment-scoped routes for data/meta/AI/automation/package handlers.	Hardened clients that always pass an environment id.

Client SDK

The client keeps the legacy factory name project(id), but the value is an environment id and the generated URLs are environment-scoped:

import { ObjectStackClient } from '@objectstack/client';

const client = new ObjectStackClient({
  baseUrl: 'https://api.example.com',
});

const env = client.environment('env_prod');

await env.data.find('customer', { top: 20 });
await env.meta.getItems('object');
await env.packages.list();

The top-level client can also set a default environment header:

const client = new ObjectStackClient({
  baseUrl: 'https://api.example.com',
  environmentId: 'env_prod',
});

That sends X-Environment-Id: env_prod on unscoped requests.

Resolution order

The open-source HttpDispatcher only parses the scoped URL for an environment-id hint (extractEnvironmentIdFromPath) and forwards it via prepareResolverHints. Actual resolution is owned by the host's KernelResolver (resolveKernel, per ADR-0006 Phase 5), which resolves the environment in this order:

:environmentId from /api/v1/environments/:environmentId/...
Hostname through the configured environment registry
X-Environment-Id
session.activeEnvironmentId
A configured default environment
A single available environment, when the registry can prove the choice is unambiguous

Control-plane endpoints remain unscoped. Requests such as /api/v1/cloud/environments/:id manage environments rather than running inside one, so they are intentionally excluded from data-plane resolution.

Migration checklist

Replace /api/v1/projects/:projectId/... with /api/v1/environments/:environmentId/....
Replace X-Project-Id with X-Environment-Id.
Replace OS_PROJECT_ID with OS_ENVIRONMENT_ID.
Store runtime rows under environment_id.
Replace the SDK client.project(id) call with client.environment(id). It was a compatibility method name until ADR-0006 D2; there is no alias, so the old spelling does not resolve.

Environment-Scoped Routing

Server configuration

Client SDK

Resolution order

Migration checklist

Related

On this page