Service De Channel Activate

Activating a Messaging Channel

What this skill does

Given a {CHANNEL_ID}, reads the channel's MessagingChannelUsage.Id, then fires PATCH /services/data/v{V}/sobjects/MessagingChannelUsage/{MCU_ID} with body {"DeploymentStatus":"Provisioning"}. The server-side chain:

1. MessagingChannelUsageFunctions.validateBeforeSave runs validateDeploymentStatus → validateChannelReadinessOnProvisioning. For WhatsApp this confirms consent is configured. Rejected writes return HTTP 400 FIELD_INTEGRITY_EXCEPTION.
2. MessagingChannelUsageFunctions.saveHook_PostStmtExecuteOnce fires unconditionally after the UPDATE statement. It calls MessagingChannelUsageFunctionsHelper.handlePostSave which registers a post-commit TransactionObserver.
3. At commit, the observer calls ConversationChannelUsageDeploymentStatusService.handleProvisioning (inherited from AbstractChannelUsageDeploymentStatusService), which:
   • Calls runProvisioning — switch-dispatches by MessageType for the external callout:
     • WHATS_APP → registerCsotWhatsAppNumber → LiveMessageSetupApi.registerWhatsAppNumber → Meta /register + status verification. 15-21s wall-clock.
     • FACEBOOK → metaGraphApiService.subscribeFacebookPage.
     • TEXT → registerCsotSms.
     • AppleBusinessChat, Line, everything else → default branch, no external callout, no network wait.
   • On success: writes DeploymentStatus = 'Active' via PLSQL.
   • On failure: writes DeploymentStatus = 'Error' plus ErrorReason / ErrorDetails.
4. Inside the same observer, a second pass syncs MessagingChannel.IsActive = true once MCU reaches Active (the isTransitioningStatus flag skips the flip while status is still Provisioning).

All synchronous within the PATCH request — the 204 response only comes back after the full chain completes. WhatsApp: ~15-21s (Meta /register round-trip). Apple / Line: ~1s (no external call; just the local save-hook + observer + PLSQL write). Verified on wadtesting 2026-04-30.

Reference File Index

Why REST PATCH instead of Apex?

A direct REST PATCH produces the identical save-hook chain as the old activateChannelUsage Apex method, with substantially less machinery — no CSRF cookie acquisition, no bootstrap fetch, no Aura response parsing, no double-wrapped returnValue. REST semantics are honest: 204 means the transition succeeded; 4xx means it didn't.

Code proof: MessagingChannelUsageFunctions.saveHook_PostStmtExecuteOnce fires on any DML path (REST, SOAP, Apex, Metadata API) — there is no Apex-specific gate. The entity XML (MessagingChannelUsage.entity.xml) marks DeploymentStatus as editAccess="always" with no <readonly> attribute. The transition validator (getValidAPIStatusTransitions) allows Disabled → Provisioning (and New → Provisioning, and Error → Provisioning | Deprovisioning, and Active → Deprovisioning). The DB-only transitions Provisioning → Active | Error are reserved for the observer's PLSQL call — that's why we write Provisioning and let the server pick the terminal state.

When NOT to use this skill

• The channel is already IsActive=true. Re-firing is blocked by the API transition validator (Active → Provisioning is not in getValidAPIStatusTransitions()) — the PATCH would return 400. The Stage 1 precondition check catches this and emits noop:true.
• Routing isn't configured. activateChannelUsage used to fail with LiveMessageSetupException / nullQueueId at the Apex entry point. With the REST path the same guard lives in validateChannelReadinessOnProvisioning — write with SessionHandlerId=null && FallbackQueueId=null → 400 FIELD_INTEGRITY_EXCEPTION. Run service-de-channel-routing-configure first. The Stage 1 check still runs defensively.
• The MCU doesn't exist. Can't PATCH a row that's missing. Run the insertion skill first — it always creates the MCU as a side-effect of addChannel.

Inputs (from caller)

• {CHANNEL_ID} — a 15- or 18-char MessagingChannel.Id (prefix 0Mj). The channel must already exist with a non-null SessionHandlerId or FallbackQueueId.
• {ORG_ALIAS} — optional; the sf CLI target-org alias. Default: whatever sf config get target-org returns. Used for OAuth and SOQL reads.
• {API_VERSION} — optional; REST API version. Default: 68.0. Any version where MessagingChannelUsage is addressable as a standard sobject is fine (v50+ should work; not exhaustively tested).

Unlike the old Aura-based version of this skill, there are no {POLL_TIMEOUT_S} / {POLL_INTERVAL_S} inputs — the PATCH is synchronous end-to-end.

Output (to caller)

Success — channel is live:

{"ok": true, "channelId": "0Mj...", "mcuId": "0gL...", "deploymentStatus": "Active", "isActive": true, "messageType": "WhatsApp", "durationMs": 20934}

Success — no-op (already active):

{"ok": true, "noop": true, "channelId": "0Mj...", "message": "Channel already active"}

Failure — precondition not met:

{"ok": false, "kind": "no-routing",     "hint": "run service-de-channel-routing-configure first"}
{"ok": false, "kind": "no-mcu",         "hint": "no MessagingChannelUsage row for this channel — run the insertion skill first"}
{"ok": false, "kind": "channel-missing","hint": "MessagingChannel id not found"}

Failure — validator or server-side provisioning error:

{"ok": false, "kind": "readiness-failed", "errorCode": "FIELD_INTEGRITY_EXCEPTION",
 "message": "...", "hint": "validateChannelReadinessOnProvisioning rejected the write — most commonly missing consent; run service-de-channel-settings-configure first"}
{"ok": false, "kind": "provisioning-error", "mcuId": "0gL...", "errorReason": "MetaRegistrationFailed", "errorDetails": "..."}
{"ok": false, "kind": "verification-failed", "hint": "WhatsApp phone number verification failed or was declined by user"}
{"ok": false, "kind": "verification-request-failed", "errorCode": "...", "message": "...", "hint": "Could not request verification code from Meta"}

Failure — auth / transport:

{"ok": false, "kind": "auth",      "hint": "OAuth token invalid / expired — run 'sf org login web'"}
{"ok": false, "kind": "transport", "status": 500, "message": "..."}

Stage 1: Precondition checks

Read the channel, then its MCU, as two separate SOQL calls. A combined subquery would be cheaper but fails on orgs where the child relationship is unnameable (see gotcha #4).

sf data query --target-org '{ORG_ALIAS}' \
  --query "SELECT Id, DeveloperName, MessageType, IsActive, SessionHandlerId, FallbackQueueId, MessagingPlatformKey FROM MessagingChannel WHERE Id = '{CHANNEL_ID}'" \
  --json > /tmp/amc-precheck-channel.json

sf data query --target-org '{ORG_ALIAS}' \
  --query "SELECT Id, DeploymentStatus, DeploymentType, ErrorReason, ErrorDetails FROM MessagingChannelUsage WHERE MessagingChannelId = '{CHANNEL_ID}'" \
  --json > /tmp/amc-precheck-mcu.json

Let channel = /tmp/amc-precheck-channel.json records[0] and mcu = /tmp/amc-precheck-mcu.json records[0]:

Also record {T0} (epoch ms at start of Stage 2) so the final envelope can report durationMs.

Stage 1.1: Fast path for already-provisioning MCU

If {INITIAL_MCU_STATUS} === "Provisioning" — the MCU is already mid-flight from a prior call in this transaction window. Skip Stage 2 (firing the PATCH) entirely and jump to Stage 3 (verification). This is a rare race guard: the observer is synchronous with the PATCH, so by the time the caller sees the 204 the status is already terminal (Active or Error) — Provisioning should be invisible from outside. If we do see it in the precheck, something wrote Provisioning in a separate DML and the observer is still mid-flight — don't fire a second PATCH.

Stage 2: PATCH MessagingChannelUsage.DeploymentStatus = "Provisioning"

Use sf api request rest so authentication stays inside the CLI's transport — no OAuth token is ever extracted into shell state.

Fire the PATCH. This call can take 15-30 seconds for WhatsApp — the observer runs synchronously, including Meta's /register round trip. sf api request rest has no separate client-side timeout to raise; it waits on the underlying HTTP call.

sf api request rest \
  "/services/data/v{API_VERSION}/sobjects/MessagingChannelUsage/{MCU_ID}" \
  --method PATCH \
  --target-org '{ORG_ALIAS}' \
  --header 'Content-Type: application/json' \
  --body '{"DeploymentStatus":"Provisioning"}' \
  --include \
  > /tmp/amc-patch-response.txt 2>&1
HTTP_CODE=$(head -1 /tmp/amc-patch-response.txt | grep -oE '[0-9]{3}')

--include prints the HTTP status/headers block ahead of the (typically empty, on 204) body — read the status line from that block rather than a -w-style trailing marker.

Classify by HTTP status:

Known 400 FIELD_INTEGRITY_EXCEPTION messages: