From 2a74deacbd5a111cff550d1d3c0f6ca126e7d20c Mon Sep 17 00:00:00 2001
From: Dan Anderson <anderson@familymedia.org>
Date: Wed, 30 Sep 2026 01:33:48 -0400
Subject: [PATCH] Bound vault sweep classification reads to selected vaults

---
 apps/api/src/indexer/vault-contents.ts | 4 +++-
 apps/api/tests/vault-contents.test.ts  | 6 ++++--
 2 files changed, 7 insertions(+), 3 deletions(-)

diff --git a/apps/api/src/indexer/vault-contents.ts b/apps/api/src/indexer/vault-contents.ts
index 2ba0266..51129e6 100644
--- a/apps/api/src/indexer/vault-contents.ts
+++ b/apps/api/src/indexer/vault-contents.ts
@@ -150,13 +150,15 @@ export async function classifyVaults(env: Env): Promise<Record<string, unknown>>
 
     // Outbound SWEEPS are ALSO a crack — a sweep moves EVERYTHING out of the address at once (no per-asset
     // row), so any outbound sweep from a funded vault empties it. Merge with sends; earliest wins.
+    // Keep selected vaults outermost: scanning sweeps first makes SQLite probe every selected rowid
+    // for each sweep. A 1,000-vault production batch read 1.5M rows instead of 3k indexed probes.
     const sweepRows =
       (
         await env.CORE_DB.prepare(
           `WITH ${selected}
          SELECT source.address,sweep.block_time,destination.address destination
          FROM selected vault
-         JOIN sweeps sweep ON sweep.source_id=vault.btc_address_id
+         CROSS JOIN sweeps sweep INDEXED BY idx_sweeps_source ON sweep.source_id=vault.btc_address_id
          JOIN address_dictionary source ON source.address_id=sweep.source_id
          JOIN address_dictionary destination ON destination.address_id=sweep.destination_id
          ORDER BY sweep.source_id,sweep.block_time`,
diff --git a/apps/api/tests/vault-contents.test.ts b/apps/api/tests/vault-contents.test.ts
index 7ea2a56..e544332 100644
--- a/apps/api/tests/vault-contents.test.ts
+++ b/apps/api/tests/vault-contents.test.ts
@@ -56,6 +56,7 @@ test("vault classification derives compact identities, contents, and crack recip
       WHERE destination_address_id IS NOT NULL;
     CREATE TABLE balances(address_id INTEGER,asset_id INTEGER,quantity TEXT,quantity_normalized TEXT);
     CREATE TABLE sweeps(source_id INTEGER,destination_id INTEGER,block_time INTEGER);
+    CREATE INDEX idx_sweeps_source ON sweeps(source_id);
     INSERT INTO address_dictionary VALUES
       (10,'vault-single'),(11,'vault-multi'),(12,'vault-foreign'),(20,'funder'),(30,'cracker'),(40,'contract');
     INSERT INTO asset_dictionary VALUES(1,'XCP'),(2,'CARD'),(3,'OTHER');
@@ -69,6 +70,7 @@ test("vault classification derives compact identities, contents, and crack recip
       (4,20,11,NULL,NULL,3,'100000000','1',1,101),
       (5,20,99,NULL,10,2,'100000000','1',1,110),
       (6,99,99,10,30,2,'100000000','1',3,190);
+    INSERT INTO sweeps VALUES(11,30,150),(12,30,160),(20,30,170);
   `);
   const result = await classifyVaults({ CORE_DB: d1(sqlite) } as never);
   assert.equal(result.classified, 3);
@@ -105,8 +107,8 @@ test("vault classification derives compact identities, contents, and crack recip
         funded: 1,
         contents_asset: null,
         contents_qty: null,
-        cracked_at: null,
-        cracker: null,
+        cracked_at: 150,
+        cracker: "cracker",
       },
       {
         token_id: "3",