@alvera-ai/platform-sdk 0.17.0 → 0.18.0

package/.agent/cookbook/dunning-sms-for-delinquent.md

@@ -1,659 +0,0 @@
----
-title: Send a payment-reminder SMS to delinquent customers with a pay-invoice link
-summary: End-to-end standard workflow — provision a Dunning SMS workflow over the AR customer dataset, ingest two Stripe customer rows (one phone + tax-id verified, one missing tax-id) through a customer interop contract, run the workflow so the SMS-reachable-and-KYC-complete filter routes each row, read the rendered SMS back from the message dataset, and close the connected-app reply loop. No agent, no LLM.
-industry: subscription
-slug: dunning-sms-for-delinquent
-vitest_source:
-  - integration-tests/tests/subscription/dunning-sms-workflow.test.ts
-  - integration-tests/tests/subscription/interoperability-contracts.test.ts
-  - integration-tests/tests/subscription/run-dac-single.test.ts
-  - integration-tests/tests/subscription/create-dac.test.ts
-  - integration-tests/tests/subscription/tools.test.ts
-  - integration-tests/tests/subscription/data-sources.test.ts
-  - integration-tests/tests/subscription/bootstrap.test.ts
-status: green
----
-
-# Problem
-
-When an invoice goes unpaid, the billing team wants to send a
-payment-reminder ("dunning") SMS with a self-serve pay link — but
-only to customers it is both *allowed* and *able* to contact. The
-customer must be SMS-reachable (a phone on file) and KYC-complete
-(a verified tax id). A customer missing either precondition should
-be skipped, not errored, and certainly not chased.
-
-The classic implementation queries the customers table, branches
-on phone-present and tax-id-present, and hands the survivors to an
-SMS sender — with both preconditions living in application code.
-
-The Alvera platform's standard workflow primitive keeps both
-checks inside the platform. An AR `customer` dataset is the
-workflow's event source; a pure-Liquid filter gates on
-`customer.phone and customer.tax_id`; a single SMS action carries
-a deep-link to a connected-app payment portal. The
-reachable-and-verified decision is a one-line filter, not a code
-branch.
-
-This cookbook walks the **whole** scenario, not just the
-provisioning: it creates the workflow, ingests two Stripe-shaped
-customer rows through the production data-activation chain (one
-phone + tax-id verified, one missing the tax id), runs the
-workflow so the filter routes each row, reads the rendered dunning
-SMS back out of the message dataset, and finally resolves the
-connected-app deep-link the SMS carries — closing the SMS → reply
-loop.
-
-The scenario is anchored to
-`platform/integration-tests/tests/subscription/dunning-sms-workflow.test.ts`
-— a green end-to-end test (§1–§8) that exercises this exact shape.
-This cookbook is a prose re-presentation of what that test walks.
-The setup file `_setup/subscription.md` already provisioned
-the tenant, datalake, and industry-admin client; this cookbook
-starts from there.
-
-# Composition
-
-| Resource provisioned             | Owner       |
-|----------------------------------|-------------|
-| SMS tool (SNS-backed)            | build       |
-| Dunning payment-portal connected app | build   |
-| Dunning SMS workflow             | build       |
-| Stripe data source               | build       |
-| Manual Upload tool               | build       |
-| Stripe Customer interop contract | build       |
-| Manual-upload DAC                | build       |
-
-The setup file `_setup/subscription.md` already
-provisioned the tenant + datalake + tenant-scoped client; this
-cookbook starts from there.
-
-# Walkthrough
-
-## 001 — create the SMS tool
-
-The Dunning SMS workflow's action invokes an SMS tool. The tool's
-`body.tool_body_type: 'sns'` means it routes via AWS SNS; local
-dev points it at LocalStack on `http://localhost:4566` via
-`endpoint_url` so no real AWS credentials are needed. The
-`intent: 'sms'` tags this tool for workflow actions that send SMS
-(versus `data_exchange` for ingestion tools).
-
-```typescript
-const smsToolResp = await api.tools.create(tenantSlug, datalakeSlug, {
-  name: `Cookbook SMS Tool ${runSuffix}`,
-  description: 'SNS-backed SMS dispatcher for the Dunning SMS workflow, wired to LocalStack.',
-  intent: 'sms',
-  status: 'active',
-  datalake_id: ctx.datalakeId,
-  body: {
-    tool_body_type: 'sns',
-    auth_method: 'access_key',
-    region: 'us-east-1',
-    phone_number: '+15551234567',
-    endpoint_url: 'http://localhost:4566',
-    access_key_id: 'test',
-    secret_access_key: 'test',
-  },
-})
-toolId = smsToolResp.data.id!
-```
-
-## 002 — create the Dunning payment-portal connected app
-
-The workflow's SMS action carries a deep-link to a connected-app
-page so the customer can open the self-serve payment portal. The
-connected app is a thin registration of the portal's URL and
-mode; the actual page is hosted outside the platform
-(`mode: 'self_hosted'`). The server-derived `slug` is captured for
-the resolve-page call in §013.
-
-```typescript
-const connectedAppResp = await api.connectedApps.create(tenantSlug, datalakeSlug, {
-  name: `Cookbook Dunning Payment Portal ${runSuffix}`,
-  description: 'Self-serve invoice payment portal linked from the outbound dunning SMS.',
-  mode: 'self_hosted',
-  urls: [
-    {
-      url: 'https://dunning.example.local',
-      is_primary: true,
-      label: 'production',
-    },
-  ],
-})
-connectedAppId = connectedAppResp.data.id!
-ctx.connectedAppSlug = connectedAppResp.data.slug!
-```
-
-## 003 — create the Dunning SMS workflow
-
-The workflow has the standard shape: a filter (Liquid; passes
-customers who are both SMS-reachable *and* KYC-complete —
-`customer.phone and customer.tax_id`), a decision (a literal
-Liquid array naming one decision key), and one SMS action. The
-SMS body references `{{ customer.customer_number }}` so the
-recipient sees which account is overdue, and
-`{{ connected_app_form_url }}` so the platform can bake in the
-`/t/<token>` pay-link.
-
-Two details matter. `dataset_type: 'customer'` makes the AR
-customer dataset the event source; the workflow context injects
-the regulated customer row under the `customer` key, so the
-filter reads `customer.phone` and `customer.tax_id` directly. A
-`context_datasets` entry queries the `message` dataset for any
-dunning SMS already sent to this customer in the last thirty days
-— the platform's built-in guard against dunning the same customer
-twice in a billing cycle. On a fresh tenant that lookup returns
-empty, which is harmless.
-
-```typescript
-const FILTER_BODY = '{% if customer.phone and customer.tax_id %}true{% endif %}'
-const DECISION_KEY = 'send_dunning_sms'
-const DECISION_BODY = `["${DECISION_KEY}"]`
-const DECISION_OUTPUT_SCHEMA = { type: 'array', items: { type: 'string' } }
-const SMS_TO_TEMPLATE = '{{ mdm_output.regulated_customer.phone }}'
-const SMS_BODY_TEMPLATE =
-  'Hi {{ mdm_output.regulated_customer.name }}, an invoice on account ' +
-  '{{ customer.customer_number }} needs your attention. ' +
-  'Pay now: {{ connected_app_form_url }}'
-
-const workflowResp = await api.workflows.create(tenantSlug, datalakeSlug, {
-  name: `Cookbook Dunning SMS Workflow ${runSuffix}`,
-  description: 'Sends a payment-reminder SMS to delinquent contracted customers with a self-serve pay link.',
-  dataset_type: 'customer',
-  status: 'live',
-  tags: ['billing', 'dunning'],
-  filter_config: {
-    type: 'custom',
-    body: FILTER_BODY,
-    output_schema: { type: 'boolean' },
-  },
-  decision_config: {
-    type: 'custom',
-    body: DECISION_BODY,
-    output_schema: DECISION_OUTPUT_SCHEMA,
-  },
-  context_datasets: [
-    {
-      dataset_type: 'message',
-      where_clause:
-        `rm.customer_id = '{{ customer_id }}' AND rm.decision_key = '${DECISION_KEY}' ` +
-        "AND rm.sent_at > NOW() - INTERVAL '30 days'",
-      limit: 1,
-      position: 0,
-    },
-  ],
-  actions: [
-    {
-      action_type: 'sms',
-      tool_id: toolId,
-      decision_key: DECISION_KEY,
-      position: 0,
-      trigger_template: 'now',
-      idempotency_template: '{{ customer_id }}-{{ decision_key }}',
-      connected_app_id: connectedAppId,
-      connected_app_route: '/portal/pay',
-      connected_app_metadata_template:
-        '{"customer_id":"{{ mdm_output.customer.id }}","customer_number":"{{ customer.customer_number }}"}',
-      tool_call: {
-        tool_call_type: 'sms_request',
-        to: { type: 'custom', body: SMS_TO_TEMPLATE },
-        body: { type: 'custom', body: SMS_BODY_TEMPLATE },
-        sms_type: 'transactional',
-      },
-    },
-  ],
-})
-workflowId = workflowResp.data.id!
-ctx.workflowSlug = workflowResp.data.slug!
-```
-
-## 004 — create the Stripe data source
-
-A workflow runs on rows; rows arrive through the data-activation
-chain. The chain's first link is a `DataSource` — a registration
-of where the rows originate. The `uri` is the system-of-record
-address; it flows into each ingested row's `source_uri`, which
-the customer template renders onto the MDM identifier `system`
-URN.
-
-```typescript
-const dataSourceResp = await api.dataSources.create(tenantSlug, datalakeSlug, {
-  name: `Cookbook Stripe Source ${runSuffix}`,
-  uri: 'stripe.example.com',
-  description: 'Stripe billing system — origin of the customer rows the dunning workflow runs on.',
-  status: 'active',
-  is_default: false,
-})
-dataSourceId = dataSourceResp.data.id!
-```
-
-## 005 — create the Manual Upload tool
-
-The Data Activation Client needs a tool. For inline-JSON ingest a
-`manual_upload` tool is the minimal choice — `intent:
-'data_exchange'` distinguishes it from the SMS tool, and
-`tool_body_type: 'manual_upload'` needs no endpoint or credential
-wiring (the rows arrive in the ingest call body, not by the tool
-fetching them).
-
-```typescript
-const manualUploadToolResp = await api.tools.create(tenantSlug, datalakeSlug, {
-  name: `Cookbook Manual Upload Tool ${runSuffix}`,
-  description: 'Manual-upload data-exchange tool — backs the DAC that ingests Stripe customer rows.',
-  intent: 'data_exchange',
-  status: 'active',
-  datalake_id: ctx.datalakeId,
-  data_source_id: dataSourceId,
-  body: { tool_body_type: 'manual_upload' },
-})
-ctx.manualUploadToolId = manualUploadToolResp.data.id!
-```
-
-## 006 — create the Stripe Customer interoperability contract
-
-The interoperability contract is the row-shaping rule: a Liquid
-template that maps an inbound Stripe customer row into an AR
-`Customer` upsert. This cookbook loads the production Stripe
-customer + MDM templates from the vendored fixtures directory
-rather than inlining the Liquid.
-
-Two template slots matter. `template_config` shapes the Customer
-resource — including the nested `regulated_customer` object that
-carries the PII fields (`name`, `email`, `phone`, `tax_id`).
-`mdm_input_config` shapes the MDM input — the identifier
-(`customer_number`), name, and contact fields the platform's
-master-data resolution keys on to decide whether this is a new
-customer or an existing one.
-
-```typescript
-const { readFileSync } = await import('node:fs')
-const { join } = await import('node:path')
-const customerTemplate = readFileSync(
-  join(process.env.COOKBOOK_FIXTURES_DIR!, 'subscription/_customers_subscription_customer.liquid'),
-  'utf8',
-)
-const mdmTemplate = readFileSync(
-  join(process.env.COOKBOOK_FIXTURES_DIR!, 'subscription/_customers_subscription_mdm.liquid'),
-  'utf8',
-)
-
-const contractResp = await api.interoperabilityContracts.create(tenantSlug, datalakeSlug, {
-  name: `Cookbook Stripe Customer Contract ${runSuffix}`,
-  description: 'Stripe customers → Subscription Customer (custom Liquid + MDM input).',
-  resource_type: 'customer',
-  template_config: { type: 'custom', body: customerTemplate },
-  mdm_input_config: { type: 'custom', body: mdmTemplate },
-  generic_table_id: null,
-})
-interopContractId = contractResp.data.id!
-```
-
-## 007 — create the manual-upload DAC
-
-The Data Activation Client binds the three preceding pieces — the
-manual-upload tool, the data source, and the interop contract —
-into one ingestion endpoint. `tool_call.tool_call_type:
-'manual_upload'` selects the inline-JSON ingest path. The
-server-derived `slug` is the handle §008 ingests rows against.
-
-```typescript
-const dacResp = await api.dataActivationClients.create(tenantSlug, datalakeSlug, {
-  name: `Cookbook Stripe Customer DAC ${runSuffix}`,
-  description: 'Manual-upload DAC — ingests Stripe customer rows into Customer via the interop contract.',
-  tool_id: ctx.manualUploadToolId,
-  data_source_id: dataSourceId,
-  tool_call: { tool_call_type: 'manual_upload' },
-  interoperability_contract_ids: [interopContractId],
-})
-dacId = dacResp.data.id!
-ctx.dacSlug = dacResp.data.slug!
-```
-
-## 008 — ingest two Stripe customer rows
-
-Two rows, ingested as inline JSON through the manual-upload DAC.
-Both carry a phone number; they differ in `tax_id`. The first row
-is KYC-complete — it has a verified tax id — so the filter
-(`customer.phone and customer.tax_id`) will pass it. The second
-omits `tax_id` entirely; even though it has a phone, the filter
-must reject it. Each ingest call gets its own batch id; both are
-pinned for the run scope in §010.
-
-```typescript
-const phone4 = String(Date.now()).slice(-4)
-
-const verifiedRow = {
-  customer_number: `CUST-KYC-${runSuffix}`,
-  customer_type: 'individual',
-  status: 'contracted',
-  name: 'Priya Anand',
-  email: `priya-${runSuffix}@example.com`,
-  phone: `+1202555${phone4}`,
-  tax_id: '123-45-6789',
-  currency: 'USD',
-  delinquent: 'true',
-}
-const unverifiedRow = {
-  customer_number: `CUST-NOKYC-${runSuffix}`,
-  customer_type: 'individual',
-  status: 'contracted',
-  name: 'Daniel Foster',
-  email: `daniel-${runSuffix}@example.com`,
-  phone: `+1203555${phone4}`,
-  currency: 'USD',
-  delinquent: 'true',
-  // tax_id intentionally absent — the workflow filter must reject this row
-}
-
-const [verifiedIngest, unverifiedIngest] = await Promise.all([
-  api.dataActivationClients.ingest(tenantSlug, datalakeSlug, ctx.dacSlug, { data: verifiedRow }),
-  api.dataActivationClients.ingest(tenantSlug, datalakeSlug, ctx.dacSlug, { data: unverifiedRow }),
-])
-ctx.batchVerified = verifiedIngest.data.batch_id!
-ctx.batchUnverified = unverifiedIngest.data.batch_id!
-```
-
-## 009 — wait for both ingest batches to reach steady-state
-
-Ingestion is async — the DAC enqueues per-row jobs that the
-`BatchMergeWorker` drains into the regulated customer table. Poll
-the DAC's activation logs until both batches show a row with
-`rows_ingested >= 1` and a non-empty `output_files` array (the
-merged Parquet landed in object storage). Only then is it safe to
-run the workflow against these rows.
-
-```typescript
-const targetBatches = new Set([ctx.batchVerified, ctx.batchUnverified])
-const deadline = Date.now() + 90_000
-let greenCount = 0
-while (Date.now() < deadline) {
-  const { data } = await api.dataActivationClients.logs.list(tenantSlug, datalakeSlug, ctx.dacSlug)
-  const green = new Set<string>()
-  for (const row of (data.data ?? []) as Array<Record<string, unknown>>) {
-    const b = row.batch_id
-    if (typeof b !== 'string' || !targetBatches.has(b)) continue
-    if (typeof row.rows_ingested !== 'number' || row.rows_ingested < 1) continue
-    const files = row.output_files
-    if (!Array.isArray(files) || files.length === 0) continue
-    green.add(b)
-  }
-  greenCount = green.size
-  if (greenCount === targetBatches.size) break
-  await new Promise((r) => setTimeout(r, 1_000))
-}
-if (greenCount !== targetBatches.size) {
-  throw new Error(`only ${greenCount}/2 customer batches reached steady-state within 90s`)
-}
-```
-
-## 010 — run the workflow against the two batches
-
-`workflows.run` with `manual_override: false` evaluates the
-filter, so the `customer.phone and customer.tax_id` filter
-genuinely routes each row. The SQL where-clause scopes the run to
-exactly the two batches §008 ingested (`ra` is the
-regulated-customer alias the run-query exposes).
-
-The SMS action's `trigger_template: 'now'` dispatches the action
-immediately rather than deferring it, so once the run fires it
-reaches a terminal status on its own — poll `batchLogs.refresh`
-until it leaves `:pending`. The run itself is still scheduled:
-`workflows.run` records it and returns, which is why the setup
-file's `ctx.waitForFiredRun` sits between the call and the log id. A `:partial` status is expected and fine here:
-one row passed and one was filtered.
-
-```typescript
-const runResp = await api.workflows.run(tenantSlug, datalakeSlug, ctx.workflowSlug, {
-  sql_where_clause: `ra.batch_id IN ('${ctx.batchVerified}', '${ctx.batchUnverified}')`,
-  mode: 'live',
-  manual_override: false,
-})
-// run-workflow only SCHEDULES the run. The log id and batch id are
-// written when it fires, so read them back via workflowRuns.get.
-const fired = await ctx.waitForFiredRun(datalakeSlug, runResp.data.workflow_run_id)
-ctx.runLogId = fired.workflowRunLogId
-ctx.runBatchId = fired.batchId!
-
-const deadline = Date.now() + 120_000
-let status: string | null = null
-while (Date.now() < deadline) {
-  const { data: log } = await api.workflows.batchLogs.refresh(tenantSlug, datalakeSlug, ctx.workflowSlug, ctx.runLogId)
-  status = log.status ?? null
-  if (status && status !== 'pending') break
-  await new Promise((r) => setTimeout(r, 2_000))
-}
-if (status === 'failed') throw new Error('workflow run reached :failed')
-if (!status || status === 'pending') {
-  throw new Error('workflow run did not leave :pending within 120s')
-}
-```
-
-## 011 — verify the filter routed: verified passes, unverified is filtered
-
-Each row produced a Workflow Execution Log. The KYC-complete
-customer (phone + tax id) passes the filter, so its WEL is
-`:executing` or `:completed`. The customer missing the tax id
-fails the filter, so its WEL is `:filtered`. A run where both
-passed — or both were filtered — would mean the filter is not
-actually evaluating both `customer.phone` and `customer.tax_id`.
-
-```typescript
-const { data: wfLogs } = await api.workflows.workflowLogs.list(tenantSlug, datalakeSlug, ctx.workflowSlug)
-const ourWels = (wfLogs.data ?? []).filter(
-  (w) => (w as { batch_id?: string }).batch_id === ctx.runBatchId,
-)
-if (ourWels.length !== 2) {
-  throw new Error(`expected 2 WELs for the run, got ${ourWels.length}`)
-}
-
-const byStatus: Record<string, number> = {}
-for (const w of ourWels) {
-  const st = (w as { status?: string }).status ?? 'unknown'
-  byStatus[st] = (byStatus[st] ?? 0) + 1
-}
-const passCount = (byStatus.executing ?? 0) + (byStatus.completed ?? 0)
-if (passCount !== 1) {
-  throw new Error(`expected 1 pass-branch WEL, got ${passCount} — distribution ${JSON.stringify(byStatus)}`)
-}
-if ((byStatus.filtered ?? 0) !== 1) {
-  throw new Error(
-    `expected 1 :filtered WEL (the tax-id-missing customer) — distribution ${JSON.stringify(byStatus)}`,
-  )
-}
-```
-
-## 012 — read the rendered dunning SMS back from the message dataset
-
-A `trigger_template: 'now'` action that passed the filter fires
-immediately and persists a row in the regulated `message`
-dataset, carrying the fully-rendered SMS body. Search that
-dataset scoped to this workflow, poll until the row appears, and
-extract the `/t/<token>` shortlink the platform baked into the
-body when it minted the connected-app page token.
-
-```typescript
-const { data: userSearch } = await api.datasets.createUserSearch(tenantSlug, datalakeSlug, 'message', {
-  search_query: `rm.workflow_id = '${workflowId}'`,
-})
-if (userSearch.status !== 'completed') {
-  throw new Error(`message user-search status=${userSearch.status} error=${userSearch.error_message ?? '(none)'}`)
-}
-
-const deadline = Date.now() + 45_000
-let messages: Array<Record<string, unknown>> = []
-while (Date.now() < deadline && messages.length === 0) {
-  const { data } = await api.datasets.search(tenantSlug, datalakeSlug, 'message', {
-    userSearchId: userSearch.id!,
-    dataAccessMode: 'regulated',
-  })
-  messages = (data.data ?? []) as Array<Record<string, unknown>>
-  if (messages.length === 0) await new Promise((r) => setTimeout(r, 1_000))
-}
-if (messages.length === 0) {
-  throw new Error('no dunning SMS message persisted for the workflow within 45s')
-}
-
-const withLink = messages
-  .map((m) => String(m.body ?? ''))
-  .find((body) => body.includes('/t/') && body.includes('needs your attention'))
-if (!withLink) {
-  throw new Error(`no rendered dunning body with a /t/ link — bodies: ${JSON.stringify(messages.map((m) => m.body))}`)
-}
-const tokenMatch = withLink.match(/\/t\/([A-Za-z0-9_-]+)/)
-if (!tokenMatch) {
-  throw new Error(`no /t/<token> in rendered body: ${withLink}`)
-}
-ctx.dunningShortPath = tokenMatch[1]!
-```
-
-## 013 — resolve the connected-app deep-link and post tracking
-
-The `/t/<token>` shortlink the SMS carries resolves via
-`connectedApps.resolvePage` — the `route_path` must match the
-action's `connected_app_route`. Posting `opened_at` +
-`form_submitted_at` via `updateMessageTracking` then mirrors what
-the connected-app frontend does when the customer opens the
-payment portal, closing the SMS → reply loop end-to-end.
-
-```typescript
-const { data: resolved } = await api.connectedApps.resolvePage(tenantSlug, datalakeSlug, ctx.connectedAppSlug, {
-  short_path: ctx.dunningShortPath,
-  user_agent: 'cookbook-doctest/dunning-sms',
-})
-if (resolved.route_path !== '/portal/pay') {
-  throw new Error(`resolvePage route_path mismatch: ${resolved.route_path}`)
-}
-if (!(resolved.message?.body ?? '').includes('needs your attention')) {
-  throw new Error('resolved page message body missing the dunning copy')
-}
-
-const now = new Date().toISOString()
-const { data: tracked } = await api.connectedApps.updateMessageTracking(tenantSlug, datalakeSlug, ctx.connectedAppSlug, {
-  short_path: ctx.dunningShortPath,
-  opened_at: now,
-  form_submitted_at: now,
-})
-if (!tracked.message?.opened_at || !tracked.message?.form_submitted_at) {
-  throw new Error('message tracking did not persist opened_at + form_submitted_at')
-}
-```
-
-## 014 — write the integration test
-
-End the build with a test you keep: re-read the workflow and prove the
-pipeline still executes — without a side effect. `mode: 'dry_run'` with a
-never-matching selection runs the FULL pipeline (selection → filter →
-decision) and intercepts only the final action call, so no message
-leaves, yet the acknowledgement proves the workflow is runnable. This
-block runs live under `make validate-cookbook`.
-
-```typescript
-// Re-GET — the workflow must still be live, or nothing will run.
-const { data: wfRow } = await api.workflows.get(tenantSlug, datalakeSlug, workflowId)
-if (wfRow.status !== 'live') {
-  throw new Error(`workflow regressed from live: ${wfRow.status}`)
-}
-// Behavioural probe — a dry run against a selection no row can match:
-// the pipeline executes end-to-end, the final action call is
-// intercepted, and the acknowledgement carries the scheduled run id.
-const { data: probeRun } = await api.workflows.run(tenantSlug, datalakeSlug, ctx.workflowSlug, {
-  sql_where_clause: "ra.batch_id = 'test-never-matching-batch'",
-  mode: 'dry_run',
-  manual_override: false,
-})
-// The run-log id does not exist until the run fires — wait, do not read a null.
-const probeFired = await ctx.waitForFiredRun(datalakeSlug, probeRun.workflow_run_id)
-if (probeFired.workflowRunLogId.length === 0) {
-  throw new Error('dry-run probe never produced a workflow_run_log_id')
-}
-```
-
-If the probe fails in production, escalate with the run response as
-evidence — don't flip the workflow's status or rewrite its configs to
-chase the error.
-
-# Branches
-
-- **The unverified customer is filtered, not failed** — §011
-  asserts the tax-id-missing customer's WEL is `:filtered`, a
-  distinct terminal status from `:failed`. A filtered row is a
-  *correct* outcome: the workflow looked at it, the
-  `customer.phone and customer.tax_id` filter rendered empty
-  because one conjunct was falsy, and the platform recorded the
-  row as intentionally skipped. No SMS runs for a filtered row —
-  exactly the business rule (never dun a customer whose KYC is
-  incomplete).
-- **Both conjuncts are load-bearing** — the filter is an `and` of
-  two preconditions. The §008 unverified row has a phone, so it
-  is the *tax-id* conjunct that fails it. A row missing the phone
-  instead would fail the same filter on the other conjunct. The
-  anchor vitest's §8 proves the opposite extreme: a filter gated
-  on a sentinel no customer can satisfy filters *every* row.
-- **The thirty-day recency guard** — §003's `context_datasets`
-  entry queries the `message` dataset for a prior dunning SMS to
-  the same customer in the last 30 days. On this fresh tenant it
-  returns empty, so the action runs. On a tenant with history, a
-  recent prior dunning message would make the action skip — the
-  platform's built-in guard against dunning the same customer
-  repeatedly within one billing cycle.
-
-# Rollback
-
-The cookbook doctest harness does not currently tear down created
-resources. The `_setup/subscription.md` setup file's
-runSuffix-scoped tenant / datalake / user names mean each run is
-naturally isolated; the seeded local DB is cheap to reset
-(`mix ecto.reset` on the platform repo).
-
-# Outcome
-
-After this cookbook's thirteen steps run green:
-
-- An subscription tenant exists with an
-  subscription-domain datalake
-- An SMS tool, a Dunning payment-portal connected app, and a
-  Dunning SMS standard workflow (status `live`, `customer`
-  dataset) are registered
-- A Stripe data source, a Manual Upload tool, a Stripe Customer
-  interop contract, and a manual-upload DAC form a working
-  ingestion chain
-- Two customers have been ingested through that chain — one
-  KYC-complete, one missing a tax id
-- Running the workflow routed them correctly: the verified
-  customer passed the phone + tax-id filter and dispatched an
-  SMS; the unverified customer was `:filtered`
-- The rendered dunning SMS is readable from the `message`
-  dataset, carrying the "needs your attention" copy with the
-  overdue account number and a `/t/` deep-link
-- The connected-app deep-link the SMS carried resolves, and
-  message tracking records the open + form-submit timestamps
-
-The business outcome — a payment-reminder SMS sent to a reachable
-and KYC-complete delinquent customer, with a working pay-invoice
-link — is demonstrated end-to-end, not merely provisioned.
-
-# See also
-
-- `_setup/subscription.md` — the inlined bootstrap that
-  provisions the tenant + datalake this cookbook starts from
-- `welcome-sms-for-customers.md` — the sibling AR cookbook; same
-  customer-dataset shape, a simpler phone-only filter
-- `.agent/tools.md` — SMS (`tool_body_type: sns`) and
-  manual-upload tool body shapes; intent classification
-- `.agent/connected_apps.md` — connected-app registration + the
-  resolve-page / message-tracking reply loop
-- `.agent/interoperability_contracts.md` — custom contract shape,
-  `template_config` vs `mdm_input_config`
-- `.agent/data_activation_clients.md` — data source → tool →
-  interop contract → DAC ingestion chain
-- `.agent/workflows.md` — standard workflow primitive (filter +
-  decision + actions), `context_datasets`, `workflows.run`
-- `.agent/cookbook/_fixtures/subscription/` — the vendored
-  Stripe customer / MDM Liquid templates §006 loads
-- `integration-tests/tests/subscription/dunning-sms-workflow.test.ts` —
-  the anchor green test (§1–§8) these snippets are lifted from
-- `integration-tests/tests/subscription/interoperability-contracts.test.ts`,
-  `run-dac-single.test.ts`, `create-dac.test.ts`, `tools.test.ts`,
-  `data-sources.test.ts` — the per-resource create + ingest
-  snippets §004–§009 are lifted from