feat(datadog): Monitor and ServiceLevelObjective resources

[gwkline]

New Datadog provider with Monitor and ServiceLevelObjective resources, for declaring alerting and SLO targets (e.g. latency/availability of tRPC endpoints or backend workflows) alongside the infrastructure they observe.

Built on @distilled.cloud/datadog — companion PR: alchemy-run/distilled#456 (new SDK package generated from Datadog's official v1 OpenAPI spec, Monitors + Service Level Objectives services). This PR bumps the distilled submodule to that branch; the pointer needs retargeting after the distilled PR merges.

import * as Datadog from "alchemy/Datadog";

const monitor = yield* Datadog.Monitor("checkout-errors", {
  name: "tRPC checkout.create error rate",
  type: "query alert",
  query:
    "sum(last_10m):sum:trace.express.request.errors{env:prod,resource_name:checkout.create}.as_count() / sum:trace.express.request.hits{env:prod,resource_name:checkout.create}.as_count() > 0.05",
  message: "checkout.create is failing for >5% of requests. @slack-alerts",
  options: { thresholds: { critical: 0.05, warning: 0.01 } },
});

yield* Datadog.ServiceLevelObjective("checkout-uptime", {
  name: "checkout.create uptime",
  type: "monitor",
  monitor_ids: [monitor.id.as<number>()],
  thresholds: [{ timeframe: "30d", target: 99.9 }],
  timeframe: "30d",
  target_threshold: 99.9,
});

• Props are the distilled request types verbatim (CreateMonitorRequest / CreateSLORequest, snake_case wire names) — same shape as the Axiom Monitor resource. Queries built in the Datadog UI port over directly; metric, monitor, and time_slice SLOs are all supported.
• Reconcile observes by cached id via getMonitor/getSLO, catches the typed NotFound tag to re-create after out-of-band deletion, PUTs in place otherwise. type changes replace (except the equivalent metric alert/query alert pair, which Datadog normalizes between). list paginates for account-wide enumeration.
• Credentials resolve via the Alchemy AuthProvider (alchemy login, registered in the CLI's builtin auth providers) or env (DD_API_KEY / DD_APP_KEY / optional DD_SITE for regional sites).
• Live-cloud tests (test/Datadog/) cover create → list → update → destroy for monitors, metric SLOs, and a monitor-backed SLO referencing the monitor's output id; they skipIf-gate when DD_API_KEY/DD_APP_KEY are absent.

🤖 Generated with Claude Code

[sam-goodwin]

You'll need to add data dog to distilled first.

Cc @Mkassabov

[gwkline]

Done — added @distilled.cloud/datadog in alchemy-run/distilled#456 (Monitors + Service Level Objectives services, generated from Datadog's official v1 OpenAPI spec via the github-style tag-bucketed pipeline). This hand-rolled client is gone; the provider now consumes the generated ops (createMonitor, getSLO, …) with typed catchTag("NotFound") handling, per the Axiom shape. The submodule pointer here targets my fork's branch until the distilled PR lands.

[sam-goodwin]

Doe this cover 100% of data dog's features? We should run a workflow to cover all instead of just doing a few