fix(hub): hub and api keep --host; the test provider is called Sandbox

docs/code.md

@@ -20,7 +20,7 @@ the integration (`code:execute` for `invoke`).
 | `norbix code qualifiers --integration <int_…> --function <id>` | the aliases and versions: name, kind (Alias / Version), the version an alias points at | `GET /qualifiers` |
 | `norbix code invoke --integration <int_…> --function <id> [--qualifier <alias\|version>] [--payload '<json>' \| --payload-file <file>] [--async] [--run-as <usr_…>] [--config '<json>'] [--dry-run]` | Sync (default): the status code, the duration in ms, the function's output and the provider's request id. `--async`: that the call was accepted | `POST /invoke` |
 | `norbix code event-sample --source <database\|membership\|files\|payments\|ai\|scheduler\|manual> [--event <name>] [--schema <sch_…>]` | the full envelope with sample data — the fixture for `sam local invoke`, `func start` or the Functions Framework | `GET /event-sample` |
-| `norbix code runs --trigger <trg_…> \| --task <tsk_…> [--limit <1-50>]` | the last runs, newest first: fired at, event, succeeded, status code, duration, error, request id — a run through the Fake provider also its `input`, the envelope the function received (`context.runAs.accessToken` masked, cut at 64 KB), pretty-printed under the row | `GET /runs` |
+| `norbix code runs --trigger <trg_…> \| --task <tsk_…> [--limit <1-50>]` | the last runs, newest first: fired at, event, succeeded, status code, duration, error, request id — a run through the Sandbox provider (API value `Fake`) also its `input`, the envelope the function received (`context.runAs.accessToken` masked, cut at 64 KB), pretty-printed under the row | `GET /runs` |
 
 `--json` on any of them prints the raw answer. `invoke` sends the function a
 `manual.invoke` envelope with the payload in `data.payload`; `context.runAs` is
@@ -79,7 +79,7 @@ All paths below are under `/{version}/code/marketplace`.
 | `norbix code marketplace functions add --connection <int_…> --key <key> [--name <text>] [--arguments '<json>'] [--arg k=v]… [--input name[:type[:required]]]…` | adds a function and prints the minted id | `POST /integrations/{id}/functions` |
 | `norbix code marketplace functions enable\|disable\|delete <func_…> --connection <int_…>` | turns it on / off, deletes it | `POST …/enable`, `POST …/disable`, `DELETE …` |
 | `norbix code marketplace functions invoke <func_…> [--payload '<json>' \| --payload-file <file>]` | calls it as you; prints `isSuccess`, `output`, `vendorRequestId` | `POST /functions/{func}/invoke` |
-| `norbix code marketplace runs --function <func_…> \| --trigger <trg_…> \| --task <tsk_…> \| --connection <int_…> [--limit <1-50>] [--details]` | the last runs, newest first: `--function` every run of one function from every source (invoke, trigger, scheduler; 50 kept), `--trigger` / `--task` the runs it started (20 kept), `--connection` the tool calls an AI chat made on an MCP connection without a saved function (rows carry `integrationId` and `toolName`). A row has `status` (`ok` / `failed` / `skipped`), `durationMs`, `source`, `correlationId`, `hasDetails`; a Fake-vendor run also its `input` (secrets `***`), pretty-printed under the row. `--details` adds each run's log lines (input, steps, answer or error) for rows with `hasDetails: true`; any other row says "details need the project's logs on" | `GET /functions/runs`; `--details`: `GET /logs/audit?targetCorrelationId=…` per row |
+| `norbix code marketplace runs --function <func_…> \| --trigger <trg_…> \| --task <tsk_…> \| --connection <int_…> [--limit <1-50>] [--details]` | the last runs, newest first: `--function` every run of one function from every source (invoke, trigger, scheduler; 50 kept), `--trigger` / `--task` the runs it started (20 kept), `--connection` the tool calls an AI chat made on an MCP connection without a saved function (rows carry `integrationId` and `toolName`). A row has `status` (`ok` / `failed` / `skipped`), `durationMs`, `source`, `correlationId`, `hasDetails`; a Sandbox-vendor run (API value `Fake`) also its `input` (secrets `***`), pretty-printed under the row. `--details` adds each run's log lines (input, steps, answer or error) for rows with `hasDetails: true`; any other row says "details need the project's logs on" | `GET /functions/runs`; `--details`: `GET /logs/audit?targetCorrelationId=…` per row |
 
 Notes:
 

docs/email.md

@@ -7,7 +7,7 @@ Every command is covered by `tests/email-routes.test.ts`, which runs it through
 the real SDK with `fetch` replaced and checks the verb and path it sends (so a
 wrong route token fails a test). `tests/email.test.ts` and
 `tests/email-existing.test.ts` also check the request fields. No test contacts
-a mail service; the only provider a test saves is **Fake**.
+a mail service; the only provider a test saves is **Sandbox** (API value `Fake`).
 
 All paths below are under `/{version}/notifications/email`, except
 `campaign messages` (the gateway spells that route `/notifications/emails/…`).
@@ -39,16 +39,15 @@ first two, for scripts that treat every module alike.
 | `norbix email integration confirm-delivery <id>` | confirm the test e-mail reached a person | `POST /integrations/confirm-human-delivery` |
 | `norbix email integration domain-health <id>` | check SPF / DKIM / DMARC of the sending domain (read only) | `POST /integrations/domain-health` |
 
-`--provider` is one of `Fake`, `Smtp`, `SendGrid`, `MailGun`, `AwsSes`. **Fake**
-needs nothing else: it accepts every send and contacts no mail service, so
+`--provider` is one of `Sandbox`, `Smtp`, `SendGrid`, `MailGun`, `AwsSes`. **Sandbox** (sent to the API as `Fake`) needs nothing else: it accepts every send and contacts no mail service, so
 nothing reaches a real inbox — use it while you develop. A project can hold one
-Fake and it has no update path: delete it and add it again. The real providers
+Sandbox and it has no update path: delete it and add it again. The real providers
 need a sender (`--from`) and their credentials: put the provider-specific
 fields in `--config` as a JSON object (inline, `@file.json`, or `-` for stdin)
 so a key does not land in your shell history.
 
 ```bash
-norbix email integration save --provider Fake
+norbix email integration save --provider Sandbox
 norbix email integration save --provider SendGrid --name SendGrid --from hello@example.com --config @sendgrid.json
 norbix email integration test --integration int_123 --to dev@example.com
 norbix email integration confirm-delivery int_123

docs/push.md

@@ -6,7 +6,7 @@ Every command is covered by `tests/push-routes.test.ts`, which runs it through
 the real SDK with `fetch` replaced and checks the verb and path it sends (so a
 wrong route token fails a test). `tests/push.test.ts` and
 `tests/push-existing.test.ts` also check the request fields. No test contacts a
-push provider; the only provider a test saves is **Fake**.
+push provider; the only provider a test saves is **Sandbox** (API value `Fake`).
 
 All paths below are under `/{version}/notifications/push`.
 
@@ -33,15 +33,15 @@ All paths below are under `/{version}/notifications/push`.
 | `norbix push integration test --integration <id> [--token <t>] [--family <f>]` | send a test push | `POST /integrations/test` |
 | `norbix push integration confirm-delivery <id>` | confirm the test push reached a person | `POST /integrations/confirm-human-delivery` |
 
-`--provider` is one of `Fake`, `AppleApns`, `AndroidFirebase`, `SafariPush`,
-`ChromeWeb`, `FirefoxWeb`, `EdgeWeb`, `ChromePush`. **Fake** needs nothing else:
+`--provider` is one of `Sandbox`, `AppleApns`, `AndroidFirebase`, `SafariPush`,
+`ChromeWeb`, `FirefoxWeb`, `EdgeWeb`, `ChromePush`. **Sandbox** (sent to the API as `Fake`) needs nothing else:
 it accepts every send and contacts no push service, so nothing reaches a real
 device — use it while you develop. The real providers need their credentials:
 put the provider-specific fields in `--config` as a JSON object (inline,
 `@file.json`, or `-` for stdin).
 
 ```bash
-norbix push integration save --provider Fake
+norbix push integration save --provider Sandbox
 norbix push integration save --provider AndroidFirebase --name Android --config @firebase.json
 norbix push integration test --integration int_123 --token dGVzdA== --family Ios
 norbix push integration confirm-delivery int_123

docs/sms.md

@@ -6,7 +6,7 @@ Every command is covered by `tests/sms-routes.test.ts`, which runs it through
 the real SDK with `fetch` replaced and checks the verb and path it sends (so a
 wrong route token fails a test). `tests/sms.test.ts` and
 `tests/sms-existing.test.ts` also check the request fields. No test contacts an
-SMS provider; the only provider a test saves is **Fake**.
+SMS provider; the only provider a test saves is **Sandbox** (API value `Fake`).
 
 All paths below are under `/{version}/notifications/sms`.
 
@@ -36,16 +36,16 @@ first two, for scripts that treat every module alike.
 | `norbix sms integration test --integration <id> [--to <phone>]` | send a test SMS | `POST /integrations/test` |
 | `norbix sms integration confirm-delivery <id>` | confirm the test SMS reached a person | `POST /integrations/confirm-human-delivery` |
 
-`--provider` is one of `Fake`, `Twilio`, `Vonage`, `Plivo`, `Telnyx`, `Bird`,
-`Telesign`, `Sinch`. **Fake** needs nothing else: it accepts every send and
+`--provider` is one of `Sandbox`, `Twilio`, `Vonage`, `Plivo`, `Telnyx`, `Bird`,
+`Telesign`, `Sinch`. **Sandbox** (sent to the API as `Fake`) needs nothing else: it accepts every send and
 contacts no SMS service, so nothing reaches a real phone — use it while you
-develop. A project can hold one Fake and it has no update path: delete it and
+develop. A project can hold one Sandbox and it has no update path: delete it and
 add it again. The real providers need their credentials: put the
 provider-specific fields in `--config` as a JSON object (inline, `@file.json`,
 or `-` for stdin).
 
 ```bash
-norbix sms integration save --provider Fake
+norbix sms integration save --provider Sandbox
 norbix sms integration save --provider Twilio --name Twilio --config @twilio.json
 norbix sms integration test --integration int_123 --to +37060000000
 norbix sms integration confirm-delivery int_123

docs/tasks/host-flag-and-sandbox.md

@@ -0,0 +1,74 @@
+# CLI: --host on hub/api, Fake → Sandbox
+This file: /Users/djovaisas/Projects/norbix/worktrees/sdks/cli/fix/host-flag/docs/tasks/host-flag-and-sandbox.md (branch fix/host-flag)
+
+## Goal
+`norbix hub` / `norbix api` honour `--host` (and every other global flag), and the CLI calls the built-in test provider "Sandbox" everywhere a user reads it.
+Not in scope: the gateway's enum / API value `Fake`, generated request-field data (src/generated), MCP tool names. Tracking file of the whole batch: /Users/djovaisas/Projects/norbix/worktrees/gateway/fix/small-batch/docs/tasks/small-fixes-batch.md (branch fix/small-batch) — owned by the tracking session.
+
+## Plan
+1. [done] fix(cli:hub:host): `hub` and `api` keep --host, --language and --prod as global flags; discovery runs for the --host target — owner item 6
+2. [done] feat(cli:integrations:labels): `--provider Sandbox` replaces `--provider Fake` for email / SMS / push; the request still sends `Fake` — owner item 7
+3. [doing] release(cli): ship through nbx-ship (rebase merge) and cut a release
+
+## Changes
+| file (absolute, branch fix/host-flag) | what changed | step |
+|------|--------------|------|
+| /Users/djovaisas/Projects/norbix/worktrees/sdks/cli/fix/host-flag/src/lib/namespace-command.ts | `knownFlags()` derives the global / command flag names; `discoveryArgv()` returns the split global part | 1 |
+| /Users/djovaisas/Projects/norbix/worktrees/sdks/cli/fix/host-flag/src/base.ts | `init()` parses `discoveryArgv()` and gives back the argv oclif overwrote | 1 |
+| /Users/djovaisas/Projects/norbix/worktrees/sdks/cli/fix/host-flag/tests/namespace-host.test.ts | new: two fake gateways ([default] profile host vs --host); 6 tests | 1 |
+| /Users/djovaisas/Projects/norbix/worktrees/sdks/cli/fix/host-flag/src/lib/providers.ts | new: `providerFlag()` — options Sandbox + real providers, Sandbox → `Fake` on the wire | 2 |
+| src/commands/{email,sms,push}/integration/{save,test}.ts | help, examples and flag texts say Sandbox | 2 |
+| src/commands/code/runs.ts, src/commands/code/marketplace/runs.ts, src/lib/code.ts, src/lib/marketplace.ts | "Fake provider/vendor" → "Sandbox provider/vendor (API value Fake)" | 2 |
+| docs/{email,sms,push,code}.md | same rename | 2 |
+| tests/{email,sms,push}-routes.test.ts | Sandbox sends `Fake`; `--provider Fake` is refused with a list that names Sandbox | 2 |
+
+## Findings
+
+fix(cli:hub:host): `--host` (also `--language`, `--prod`) was not in the hand-kept global flag list, so it became a request field and the [default] profile host was used — done
+    where: /Users/djovaisas/Projects/norbix/worktrees/sdks/cli/fix/host-flag/src/lib/namespace-command.ts:59 (branch fix/host-flag)
+```ts
+// before — src/lib/namespace-command.ts:39-41 (origin/main 7dc7298)
+  protected splitArgv(): {flagArgv: string[]; rest: string[]; help: boolean} {
+    const VALUE_FLAGS = new Set(['--project', '--env', '--region', '--api-key', '--account', '--profile', '--body'])   // <-- missing: --host, --language
+    const BOOL_FLAGS = new Set(['--json', '--yes', '-y', '--dry-run'])                                                  // <-- missing: --prod
+```
+```ts
+// after — src/lib/namespace-command.ts:59-61 (fix/host-flag)
+  protected splitArgv(): {flagArgv: string[]; rest: string[]; help: boolean} {
+    const ctor = this.constructor as typeof NamespaceCommand
+    const {VALUE_FLAGS, BOOL_FLAGS} = knownFlags({...BaseCommand.baseFlags, ...ctor.flags})
+```
+
+fix(cli:base:discovery): `init()` parsed the full argv of `hub`/`api`, failed on the request fields, swallowed the error and skipped host discovery; and oclif's `parse` overwrites `this.argv` — done
+    where: /Users/djovaisas/Projects/norbix/worktrees/sdks/cli/fix/host-flag/src/base.ts:370 (branch fix/host-flag)
+```ts
+// after — src/base.ts:374-384 (fix/host-flag)
+    let flags: GlobalFlags
+    // oclif's parse replaces this.argv with what it parsed; `run` needs the whole argv back.
+    const argv = this.argv
+    try {
+      flags = (await this.parse(ctor as unknown as Parameters<typeof this.parse>[0], this.discoveryArgv())).flags as GlobalFlags
+    } catch {
+      return
+    } finally {
+      this.argv = argv      // <-- without this, `hub scheduler tasks get` lost its words and listed the modules
+    }
+```
+```ts
+// the test that proves it — tests/namespace-host.test.ts:45-51 (fix/host-flag); fails on origin/main, passes here
+  it('hub <request> --host X goes to X, not to the [default] profile host', async () => {
+    const r = await cli(home, ['hub', 'scheduler', 'tasks', 'get', ...target(), '--json'])
+    expect(r.code, r.stderr).toBe(0)
+    expect(byDefault.hits).toHaveLength(0)
+    expect(byFlag.hits).toHaveLength(1)
+```
+
+## Rejected / moved out
+- decision(cli:integrations): keep `Fake` as a hidden alias of `--provider` — rejected — reason: oclif checks `options` before `parse`, so an alias means dropping `options` and `norbix schema` would lose the provider list agents read; Norbix has no tenants yet (greenfield), so no back-compat. `--provider Fake` now fails with "Expected --provider=Fake to be one of: Sandbox, …".
+- chore(cli): main checkout /Users/djovaisas/Projects/norbix/sdks/cli is on fix/sdk-1.3-and-deps with 1 changed file — not mine, untouched.
+
+## Needs you
+- none
+
+## Open questions
+- none

src/base.ts

@@ -372,15 +372,24 @@ export abstract class BaseCommand extends Command {
     const ctor = this.constructor as typeof BaseCommand
     if (!ctor.discoversHost) return
     let flags: GlobalFlags
+    // oclif's parse replaces this.argv with what it parsed; `run` needs the whole argv back.
+    const argv = this.argv
     try {
-      flags = (await this.parse(ctor as unknown as Parameters<typeof this.parse>[0])).flags as GlobalFlags
+      flags = (await this.parse(ctor as unknown as Parameters<typeof this.parse>[0], this.discoveryArgv())).flags as GlobalFlags
     } catch {
       return
+    } finally {
+      this.argv = argv
     }
 
     await this.prepareHost(flags)
   }
 
+  /** The argv `init` parses to find the host; commands that split their own argv (hub, api) pass the global part only. */
+  protected discoveryArgv(): string[] {
+    return this.argv
+  }
+
   /** Discover the host `flags` point at, so `resolveContext` knows its Hub. */
   protected async prepareHost(flags: GlobalFlags, opts: {refresh?: boolean} = {}): Promise<void> {
     const target = this.pickTarget(flags)

src/commands/code/marketplace/runs.ts

@@ -15,7 +15,7 @@ const ONE_OF = ['function', 'trigger', 'task', 'connection']
 export default class CodeMarketplaceRuns extends BaseCommand {
   static description = `The last runs of a marketplace function, of a trigger, of a scheduler task or of an MCP connection's AI tool calls
 
-Newest first: when it ran, the event (invoke for a call by a user, an API key or the AI), the source (user, api-key, ai, database, membership, files, payments, scheduler), the function, the run-as user, the status (ok, failed, skipped), the duration in ms, the error, the vendor's request id and the correlation id. --function lists every run of one function from every source (up to 50 kept); --trigger and --task list the runs that trigger or task started (20 kept). --connection lists the tool calls an AI chat made on an MCP connection without a saved function (each row names the toolName). A run through the Fake vendor also carries input — what the vendor received, secrets masked — printed as JSON under its row (a string with --json).
+Newest first: when it ran, the event (invoke for a call by a user, an API key or the AI), the source (user, api-key, ai, database, membership, files, payments, scheduler), the function, the run-as user, the status (ok, failed, skipped), the duration in ms, the error, the vendor's request id and the correlation id. --function lists every run of one function from every source (up to 50 kept); --trigger and --task list the runs that trigger or task started (20 kept). --connection lists the tool calls an AI chat made on an MCP connection without a saved function (each row names the toolName). A run through the Sandbox vendor (API value Fake) also carries input — what the vendor received, secrets masked — printed as JSON under its row (a string with --json).
 
 --details also reads each run's log lines — input, steps, the answer or the error — by its correlation id (GET /logs/audit). They exist only for a run made while the project's logs were on in that environment (hasDetails: true); any other row says "${RUN_DETAILS_NEED_LOGS}". With --json each row gains "details" (the lines, or null) and, when null, "detailsNote".`
 

src/commands/code/runs.ts

@@ -6,7 +6,7 @@ import {formatRuns} from '../../lib/code.js'
 export default class CodeRuns extends BaseCommand {
   static description = `The last cloud-function runs of a trigger or a scheduler task
 
-Newest first: when it fired, the event, whether it succeeded, the status code, the duration in ms, the error and the provider's request id. A run through the Fake provider also carries input — the envelope the function received, context.runAs.accessToken masked, cut at 64 KB — printed as JSON under its row (a string with --json). The same list as the Last runs panel of the trigger's or task's edit screen.`
+Newest first: when it fired, the event, whether it succeeded, the status code, the duration in ms, the error and the provider's request id. A run through the Sandbox provider (API value Fake) also carries input — the envelope the function received, context.runAs.accessToken masked, cut at 64 KB — printed as JSON under its row (a string with --json). The same list as the Last runs panel of the trigger's or task's edit screen.`
 
   static examples = [
     '<%= config.bin %> code runs --trigger trg_1p4R3gTobGZpD6fXvHT4rd',