CatalogController gains platform-admin-gated writes: create a plan,
create a draft version (with its plan items, transactionally), publish
a version, create an addon — plus a plan-detail endpoint exposing every
version. Pricing stays versioned: create_version always makes a new
draft, publish retires the prior active version, existing subscriptions
are untouched. The Catalog context functions already existed; this just
exposes them over HTTP.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
POST /api/v1/integrations/llm-usage stores priced LLM usage events
(idempotent on gateway_request_id) in llm_usage_events. The gateway is
the LLM-pricing authority — arcadia-cloud trusts the charge it sends
rather than re-pricing.
The monthly invoice rollup now appends an llm_usage line per deployment
alongside the infra quote lines; the exact decimal charges are summed
and rounded to cents once. Closes the gateway→cloud billing loop.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
ArcadiaCloud.Analytics — the operator margin view. Revenue (tenant
invoices, ex-GST) vs COGS (DO cost lines), margin = revenue - COGS.
- margin_summary/1 — overall P&L + per-tenant + per-deployment margin
for a month.
- by_kind/1 — revenue and COGS broken down by resource kind (separate
axes; billing kinds and DO kinds don't 1:1).
- live_accrual/0 — current-month unbilled: runs the quote engine with
partial metered usage per active subscription. "What tenants are
racking up right now before the rollup."
COGS-to-tenant attribution uses COALESCE(deployment.tenant_id,
resource.tenant_id) — a resource in a deployment bills to the
deployment's tenant; standalone resources fall back to their own
tenant_id (skyai-internal infra).
API (platform_admin only):
- GET /api/v1/dashboard/margin?period=YYYY-MM-DD
- GET /api/v1/dashboard/accrual
Schema fix: cloud_resources.tenant_id and cloud_projects.tenant_id were
binary_id (UUID) while cloud_deployments / tenant_invoices use string.
Migrated both to text — a UUID is a valid string, arcadia also uses
non-UUID tenant slugs ("platform-admin"), and the type alignment lets
the analytics COALESCE join work. Side benefit: kills the phase-1 bug
where a non-UUID tenant_id claim crashed the inventory query.
Smoke verified against real ingested April DO COGS: pilot deployment on
Studio with 5 droplets, metered + rolled up — by_tenant shows
dashboard-pilot rev $71.12 / COGS $30.53 / margin $40.59 (57.1%);
overall P&L revenue $71.12 vs all-April COGS $86.92 = -$15.80 (the
pilot's revenue doesn't cover Sky AI's full April infra — correct, the
rest is internal/unattributed).
Phase 3 complete: catalog, deployments+subscriptions, quote engine,
metering, invoice rollup, cost-vs-revenue dashboard.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Month-end engine — turns a period of metered usage into tenant invoices
(revenue side). Distinct from cloud_invoices, which are DO's bills to
Sky AI (COGS).
tenant_invoices — one per (tenant, period). subtotal/tax/total cents,
status draft/issued/paid/void. unique (tenant_id, period_start).
tenant_invoice_lines — kind plan_base/addon/overage/tax, tagged with
deployment_id (NULL for tenant-level lines like GST) + resource_kind,
so the cost-vs-revenue dashboard can group by deployment and by kind.
ArcadiaCloud.Invoicing.roll_up_period/3:
- groups active subscriptions by tenant
- one tenant_invoice per tenant; per subscription, runs the quote engine
with the deployment's ACTUAL metered usage (Metering.usage_for_period)
and persists the recurring + overage lines tagged with the deployment
- appends a tenant-level GST line (AU 10%, per project_skyai_australia)
- idempotent on (tenant_id, period_start); re-run skips unless force:true
Because the same quote engine serves provisioning-time projection and
month-end invoicing, a tenant's quoted price and invoiced price are
computed identically.
InvoiceRollupWorker — Oban cron, 1st of month 03:00 UTC, invoices the
month just ended.
API (platform_admin sees all; tenants scoped to own):
- GET /api/v1/invoices — tenant invoice list
- GET /api/v1/invoices/:id — invoice with lines
Also: SubscriptionAddon now preloads its :addon so quote/invoice lines
read "Addon: storage_50gb" rather than the addon UUID.
Smoke verified: pilot deployment on Studio + storage_50gb, 3 droplets
metered across all 30 days of April (2160 droplet_hours vs 1488
included) — rollup produced an invoice with plan_base $50 + addon $7.50
+ droplet_hours overage $6.72 (672h x 1c) = $64.22 subtotal, GST $6.42,
total $70.64. Re-run without force correctly skipped.
NOT in this chunk: pushing tenant invoices to skyai-finance as AR —
that needs an income-side endpoint on skyai-finance (the phase-1 push
endpoint creates vendor expense invoices, wrong direction). Deferred to
its own chunk.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
ArcadiaCloud.Quoting.quote/3 — turns a plan version + addons + usage into
itemized pricing. The same function serves both provisioning-time
projection (with projected usage) and month-end invoicing (with actual
metered usage), so the two can never disagree on the math.
A quote has:
- recurring — plan base + addon lines (fixed monthly)
- first_invoice — recurring prorated to a mid-month start date
(fraction = days_remaining / days_in_month)
- overage — usage beyond plan+addon allowance, billed in arrears
- all_in_monthly_cents
Overage computation: per plan_item, effective allowance =
included_qty + summed addon qty for the same resource_kind. Overage =
max(0, used - allowance), clamped by hard_cap_qty. overage_price_cents
is per overage_unit; a @unit_factor map converts base-unit usage to
overage units (e.g. 1k_tokens divides token counts by 1000).
Lines are uniform maps (kind/resource_kind/description/qty/unit/
unit_price_cents/amount_cents/meta) — the invoice rollup chunk will
persist them near-verbatim as invoice_lines.
addon_fields/1 accepts either a catalog Addon (projection from the
catalog) or a SubscriptionAddon (actual, using snapshotted price/qty).
API: POST /api/v1/quote { plan_code, addons[], projected_usage{},
starts_on } — tenant-facing pre-provisioning price preview.
Smoke verified against seeded plans:
- Studio bare: $50/mo, no overage.
- Studio + storage_50gb, mid-month start 2026-05-20, projected usage:
recurring $57.50; first invoice prorated 12/31 -> $22.25; overage
LLM 500k-token $5.00 + Spaces 50GB $1.00 (the addon correctly lifted
the Spaces allowance 100->150 GB); droplet_hours exactly at allowance
produced no line; all-in $63.50/mo.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
cloud_deployments — the billable unit (one app instance). A tenant has
1..N deployments; cloud_resources.deployment_id ties resources to one.
Fields: tenant_id, slug (unique per tenant), display_name, region,
state, llm_mode, billing_action_suspended (operator override),
template_code/version (nullable — formal templates land in phase 4).
Lifecycle state machine in ArcadiaCloud.Deployments — states trial /
active / past_due / paused / suspended / cancelled / archived. Every
transition is validated against an explicit @transitions map and
recorded in cloud_deployment_events. create_deployment defaults to
`active` (trial is wired but no flow enters it yet).
subscriptions — one per deployment, binds it to a plan_version. status
active/paused/cancelled, current period dates, trial_ends_at.
subscription_addons — addons attached to a subscription with price + qty
SNAPSHOTTED at attach time, so a later catalog price change can't
retroactively reprice an existing subscriber.
ArcadiaCloud.Subscriptions context: create_subscription (period defaults
to current calendar month), attach_addon (snapshots from the live Addon),
change_plan_version (migrate to a new version — price changes / up-down
grades), get_subscription_for_deployment.
API (platform_admin sees all tenants; others scoped to own tenant_id):
- GET/POST /api/v1/deployments
- GET /api/v1/deployments/:id (with subscription + events)
- POST /api/v1/deployments/:id/transition
- POST /api/v1/deployments/:id/subscribe (plan_code + optional addons)
Smoke verified: created a deployment, transitioned active->paused
(events logged with actor), rejected an invalid paused->archived
transition (422), subscribed to Studio with the storage_50gb addon —
addon price snapshotted at 750c/qty 50; show returns deployment +
subscription + event history.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Five catalog tables:
- plans — plan identity (code, name); the stable thing.
- plan_versions — versioned pricing (base_price_cents, currency,
status draft/active/retired). A subscription binds
to a version; raising prices = publish a new
version, existing subs unaffected until migrated.
- plan_items — what a version includes per resource_kind, plus
overage terms (overage_unit, overage_price_cents,
hard_cap_qty).
- addons — a la carte upgrades (code, resource_kind, qty,
price_cents).
- resource_prices — effective-dated fallback per-unit pricing for
ad-hoc items not covered by a plan.
ArcadiaCloud.Catalog context: plan + version CRUD, active_version/1
(what a new signup gets), publish_version/1 (retires the prior active
version transactionally then activates the new one),
current_resource_price/2 (effective-dated lookup).
Seed (priv/repo/seeds/catalog_seed.exs, idempotent) creates three AUD
plans — Starter $20, Studio $50, Pro $120/mo — with included
droplet_hours / spaces_gb_month / snapshot_gb_month / bandwidth_gb /
dns_zones (and LLM token allowances on Studio + Pro), plus three storage
/ LLM addons. Prices are placeholders to tune against real DO COGS once
the cost-vs-revenue dashboard lands.
API (authenticated tenants — the catalog is what they pick from):
- GET /api/v1/catalog/plans — plans with active version + items
- GET /api/v1/catalog/addons
Smoke verified: seed creates 3 plans + 3 addons; endpoints return the
shaped catalog.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Compares cloud_provisioned.spec (what we asked DO for) against the live
cloud_resources row (what DO actually has). Any divergence becomes an
operator-resolvable drift record.
cloud_drift table: one row per drifted field. status open/accepted/
reverted/stale. Partial unique index keeps at most one OPEN drift per
(resource, field); resolved rows are retained as history.
ArcadiaCloud.Drift context:
- detect_all/0 — sweeps every provisioned resource. Per field in spec,
resolves the actual value (top-level schema field first, then attrs),
compares with loose equality (stringified scalars; lists as sets so
JSON round-trips don't false-positive). Mismatches upsert a cloud_drift
row + emit a drift_detected event in the resource event log.
- close_stale_drift — an open drift whose field no longer mismatches
(fixed elsewhere) closes as "stale" on the next sweep.
- accept_drift/2 — the live value becomes the new desired-state: parent
cloud_provisioned.spec is updated, spec_version bumped, drift closed
"accepted". Revert (mutating live infra back to spec) is intentionally
NOT here — it needs a saga and lands with the droplet-resize work.
DriftDetectionWorker — Oban cron at :20 past the hour, offset past the
:15 resource syncs so it compares against fresh inventory.
Provisioning.record_provisioned/3 — populates cloud_provisioned desired-
state (upsert on resource_id, bumps spec_version). Future provisioning
sagas call this; for now it's how drift gets something to detect.
API (platform_admin only):
- GET /api/v1/drift — open drift inbox
- POST /api/v1/drift/:id/accept — adopt live value as desired-state
Smoke verified: recorded a droplet's desired spec with a deliberately
wrong size_slug + a correct region; detect_all flagged only size_slug,
wrote the drift_detected event; accept updated the spec to the live
value and closed the drift; re-detect found zero drift.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Three new schemas:
- cloud_balance_snapshots — hourly MTD balance/usage poll for live-accrual.
- cloud_invoices — header per provider invoice, with ingest status flags.
- cloud_cost_lines — per-line-item COGS, FK to cloud_resources where matched.
Three new Oban workers (queue: cloud_billing):
- BalanceWorker (hourly) records a snapshot.
- BillingHistoryWorker (daily) discovers invoices via /v2/customers/my/
billing_history, upserts headers, enqueues an InvoiceIngestWorker for
each not-yet-ingested invoice.
- InvoiceIngestWorker (per-invoice) fetches /invoices/:uuid/csv, parses
with NimbleCSV (header-keyed so column order shifts don't break us),
replaces the invoice's line set, then matches lines to cloud_resources
by (kind, name) — case-insensitive, name extracted from "name (size)"
description format.
DigitalOcean.Client gains get_balance / list_billing_history /
get_invoice_summary / fetch_invoice_csv. The CSV endpoint returns text/csv
so we bypass Req's body decoder.
Cron additions: BalanceWorker hourly at :07, BillingHistoryWorker daily
at 02:23.
API:
- GET /api/v1/billing/balance — latest snapshot, platform_admin only.
- GET /api/v1/billing/cost-lines?period=YYYY-MM-DD&kind&limit — per-line
COGS, platform_admin only.
Live smoke against real DO billing API surfaced and fixed three CSV-format
gotchas: column headers use underscores not spaces (group_description,
project_name), USD column has $ prefix, dates use "YYYY-MM-DD HH:MM:SS
+0000" format (space separator + RFC822 offset).
Verified: 137 historical invoices discovered going back to 2014;
April 2026 invoice (33 lines, $86.92 total) ingested with 6/33 lines
matched to current cloud_resources. Unmatched lines are correctly
historic droplets, Spaces buckets (not yet synced), and GST.
NimbleCSV ~> 1.2 added as a dep.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
API-only Phoenix 1.8 project for cloud-ops, inventory, billing, and
provisioning sagas. Validates arcadia JWTs via shared Guardian secret
(verify-only; arcadia-app remains the issuer).
Deps beyond default Phoenix: guardian, cors_plug, oban, req.
Postgres on local port 5433 per arcadia stack convention.
Endpoint runs on :4005.
Endpoints:
- GET /api/health — public, returns service identifier
- GET /api/v1/inventory — auth-gated, returns empty list (phase 0 stub)
Oban configured with the queues phase 1+ will need:
provisioning / cloud_sync_fast|full|slow / cloud_billing / metering.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
