From b157fc7dfafd4f2e46bafc2d069179012f382325 Mon Sep 17 00:00:00 2001
From: JSONbored <49853598+JSONbored@users.noreply.github.com>
Date: Sat, 13 Jun 2026 15:54:38 -0700
Subject: [PATCH] perf(db): index ai_usage_events(status, created_at) for the
 daily-budget query
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

sumAiEstimatedNeuronsSince runs SELECT sum(estimated_neurons) WHERE status='ok'
AND created_at >= ? — verified via EXPLAIN QUERY PLAN on prod to be a full
`SCAN ai_usage_events` (no covering index; existing indexes are on feature/actor).
It runs on every AI review/summary (the daily-budget check) and the table grows
with each AI call, so the scan cost grows unbounded once AI is enabled.

Add the covering index (status equality + created_at range). EXPLAIN confirms
the query now SEARCHes via it. Investigation found this is the ONLY un-indexed
read in the codebase — the per-webhook/gate hot path (issues/pull_requests/
files/bounties by repo, official_miner_detections by login, settings/repos by
PK) is already fully indexed, so the recent prod D1 overload is capacity/DB-size
related, not a missing-index problem on the hot path.

Migration numbered 0028 to avoid colliding with 0026/0027 reserved by #652.
---
 migrations/0028_ai_usage_status_index.sql |  6 ++++++
 src/db/schema.ts                          |  3 +++
 test/unit/ai-usage-index.test.ts          | 23 +++++++++++++++++++++++
 3 files changed, 32 insertions(+)
 create mode 100644 migrations/0028_ai_usage_status_index.sql
 create mode 100644 test/unit/ai-usage-index.test.ts

diff --git a/migrations/0028_ai_usage_status_index.sql b/migrations/0028_ai_usage_status_index.sql
new file mode 100644
index 0000000000..8d9a114bfb
--- /dev/null
+++ b/migrations/0028_ai_usage_status_index.sql
@@ -0,0 +1,6 @@
+-- Covers the daily AI-neuron budget query (sumAiEstimatedNeuronsSince):
+--   SELECT sum(estimated_neurons) FROM ai_usage_events WHERE status = 'ok' AND created_at >= ?
+-- Without this index that aggregate full-scans ai_usage_events, and it runs on every AI review/summary.
+-- Numbered 0028 to avoid colliding with migrations 0026/0027 reserved by the AI-review PR (#652); the
+-- D1 migration runner applies un-applied files in order and tolerates gaps.
+CREATE INDEX IF NOT EXISTS ai_usage_events_status_created_idx ON ai_usage_events (status, created_at);
diff --git a/src/db/schema.ts b/src/db/schema.ts
index 326a4f028d..9d0baa6827 100644
--- a/src/db/schema.ts
+++ b/src/db/schema.ts
@@ -923,5 +923,8 @@ export const aiUsageEvents = sqliteTable(
   (table) => ({
     featureCreated: index("ai_usage_events_feature_created_idx").on(table.feature, table.createdAt),
     actorCreated: index("ai_usage_events_actor_created_idx").on(table.actor, table.createdAt),
+    // Covers the daily-budget query (sumAiEstimatedNeuronsSince): WHERE status='ok' AND created_at >= ?.
+    // Without it that aggregate full-scans ai_usage_events, which runs on every AI review/summary.
+    statusCreated: index("ai_usage_events_status_created_idx").on(table.status, table.createdAt),
   }),
 );
diff --git a/test/unit/ai-usage-index.test.ts b/test/unit/ai-usage-index.test.ts
new file mode 100644
index 0000000000..61de4f065f
--- /dev/null
+++ b/test/unit/ai-usage-index.test.ts
@@ -0,0 +1,23 @@
+import { describe, expect, it } from "vitest";
+import { createTestEnv } from "../helpers/d1";
+
+describe("ai_usage_events budget index", () => {
+  it("creates the (status, created_at) index and the budget query uses it (SEARCH, not SCAN)", async () => {
+    const env = createTestEnv();
+
+    const idx = await env.DB.prepare("SELECT name FROM sqlite_master WHERE type='index' AND name = ?")
+      .bind("ai_usage_events_status_created_idx")
+      .first<{ name: string }>();
+    expect(idx?.name).toBe("ai_usage_events_status_created_idx");
+
+    // sumAiEstimatedNeuronsSince: WHERE status='ok' AND created_at >= ? — must SEARCH via the new index.
+    const plan = await env.DB.prepare(
+      "EXPLAIN QUERY PLAN SELECT coalesce(sum(estimated_neurons),0) FROM ai_usage_events WHERE created_at >= ? AND status = 'ok'",
+    )
+      .bind("2026-01-01T00:00:00.000Z")
+      .all<{ detail: string }>();
+    const detail = (plan.results ?? []).map((row) => row.detail).join(" ");
+    expect(detail).toContain("ai_usage_events_status_created_idx");
+    expect(detail).not.toContain("SCAN ai_usage_events ");
+  });
+});