fix(memory): use sqlite-vec knn (MATCH+k) for vector search

src/memory/manager-search.test.ts

@@ -0,0 +1,40 @@
+import { describe, expect, it, vi } from "vitest";
+
+import { searchVector } from "./manager-search.js";
+
+describe("memory vector search SQL", () => {
+  it("uses sqlite-vec knn query (MATCH + k) when available", async () => {
+    const rows = [
+      {
+        id: "id-1",
+        path: "MEMORY.md",
+        start_line: 1,
+        end_line: 1,
+        text: "hello",
+        source: "memory",
+        dist: 0.1,
+      },
+    ];
+    const all = vi.fn((..._args: unknown[]) => rows);
+    const prepare = vi.fn((_sql: string) => ({ all }));
+    const db = { prepare } as unknown as Parameters<typeof searchVector>[0]["db"];
+
+    const result = await searchVector({
+      db,
+      vectorTable: "chunks_vec",
+      providerModel: "mock-model",
+      queryVec: [1, 2, 3],
+      limit: 5,
+      snippetMaxChars: 100,
+      ensureVectorReady: async () => true,
+      sourceFilterVec: { sql: "", params: [] },
+      sourceFilterChunks: { sql: "", params: [] },
+    });
+
+    expect(result).toHaveLength(1);
+    expect(prepare).toHaveBeenCalledTimes(1);
+    const sql = String(prepare.mock.calls[0]?.[0] ?? "");
+    expect(sql).toContain("embedding MATCH ? AND k = ?");
+    expect(sql).toContain("WITH knn AS");
+  });
+});

src/memory/manager-search.ts

@@ -31,19 +31,28 @@ export async function searchVector(params: {
 }): Promise<SearchRowResult[]> {
   if (params.queryVec.length === 0 || params.limit <= 0) return [];
   if (await params.ensureVectorReady(params.queryVec.length)) {
+    const query = vectorToBlob(params.queryVec);
     const rows = params.db
       .prepare(
-        `SELECT c.id, c.path, c.start_line, c.end_line, c.text,\n` +
+        `WITH knn AS (\n` +
+          `  SELECT id\n` +
+          `    FROM ${params.vectorTable}\n` +
+          `   WHERE embedding MATCH ? AND k = ?\n` +
+          `)\n` +
+          `SELECT c.id, c.path, c.start_line, c.end_line, c.text,\n` +
           `       c.source,\n` +
           `       vec_distance_cosine(v.embedding, ?) AS dist\n` +
-          `  FROM ${params.vectorTable} v\n` +
-          `  JOIN chunks c ON c.id = v.id\n` +
+          `  FROM knn\n` +
+          `  JOIN chunks c ON c.id = knn.id\n` +
+          `  JOIN ${params.vectorTable} v ON v.id = knn.id\n` +
           ` WHERE c.model = ?${params.sourceFilterVec.sql}\n` +
           ` ORDER BY dist ASC\n` +
           ` LIMIT ?`,
       )
       .all(
-        vectorToBlob(params.queryVec),
+        query,
+        params.limit,
+        query,
         params.providerModel,
         ...params.sourceFilterVec.params,
         params.limit,