Configure platform log fields (#2147)

Author: RhysSullivan

apps/cloud/docs/platform-logs.md

@@ -0,0 +1,17 @@
+# Platform logs
+
+Apply `apps/cloud/logpush.json` to the account's `workers_trace_events` Logpush
+job as the JSON body of `PUT /accounts/{account_id}/logpush/jobs/{job_id}`.
+This updates output fields without changing the destination or event selection.
+Wrangler's `logpush: true` enables export for the Worker; it does not manage
+the account-level job configuration.
+
+The field list retains outcomes, logs, exceptions, timings and script metadata.
+It excludes `Event`, which contains request URLs and other source-event data.
+Worker observability query redaction is separate from Logpush output selection.
+Redacted request paths and HTTP statuses remain available in application traces.
+
+After applying, read back the job's output options and check newly delivered
+records at the destination. Allow for configuration propagation and in-flight
+batches. Verify that `Event` is absent and retained diagnostic fields still arrive.
+This does not remove historical records or sanitize arbitrary console messages.

apps/cloud/logpush.json

@@ -0,0 +1,20 @@
+{
+  "output_options": {
+    "output_type": "ndjson",
+    "timestamp_format": "rfc3339",
+    "field_names": [
+      "CPUTimeMs",
+      "DispatchNamespace",
+      "Entrypoint",
+      "EventTimestampMs",
+      "EventType",
+      "Exceptions",
+      "Logs",
+      "Outcome",
+      "ScriptName",
+      "ScriptTags",
+      "ScriptVersion",
+      "WallTimeMs"
+    ]
+  }
+}

apps/cloud/wrangler.jsonc

@@ -41,6 +41,8 @@
   // Axiom. Pinned here because the setting lives on the script: a deploy that
   // omits it can reset it to false, which is how the export silently died in
   // the July deploy-tooling migration.
+  // Account-level output fields are configured separately via logpush.json;
+  // see docs/platform-logs.md.
   "logpush": true,
   "durable_objects": {
     "bindings": [