Skip to content

armgit / mem0

Public Imported from GitHub
0

The Memory Layer for AI Agents - Drop-in memory infrastructure for AI agents and apps. Context that persists. Built for production.

fix(ts-oss/baidu): convert L2 distance to similarity score in search() (#6485)

Abhay Singhf590c9596aa2
diff --git a/mem0-ts/src/oss/src/vector_stores/baidu.ts b/mem0-ts/src/oss/src/vector_stores/baidu.ts
index 645910daa..34e54c5fe 100644
--- a/mem0-ts/src/oss/src/vector_stores/baidu.ts
+++ b/mem0-ts/src/oss/src/vector_stores/baidu.ts
@@ -453,7 +453,13 @@ export class BaiduDB implements VectorStore {
return (response.rows ?? []).map((result) => ({
id: String(result.row.id),
payload: resultPayload(result.row),
- score: result.score,
+ // L2 is a distance (lower = closer); the VectorStore contract wants higher = better.
+ score:
+ this.metricType === "L2"
+ ? result.score != null
+ ? 1 / (1 + result.score)
+ : undefined
+ : result.score,
}));
}
diff --git a/mem0-ts/src/oss/tests/baidu.test.ts b/mem0-ts/src/oss/tests/baidu.test.ts
index 912758ef9..a2a50013d 100644
--- a/mem0-ts/src/oss/tests/baidu.test.ts
+++ b/mem0-ts/src/oss/tests/baidu.test.ts
@@ -431,9 +431,12 @@ describe("BaiduDB reads", () => {
],
});
- const results = await makeStore(client).search([1, 2, 3], 5, {
- userId: "alice",
- });
+ // Non-L2 metrics already return a higher-is-better score and pass through untouched.
+ const results = await makeStore(client, { metricType: "COSINE" }).search(
+ [1, 2, 3],
+ 5,
+ { userId: "alice" },
+ );
expect(results).toEqual([{ id: "m1", payload: { data: "x" }, score: 0.8 }]);
const { request } = client.vectorSearch.mock.calls[0][0];
@@ -445,6 +448,35 @@ describe("BaiduDB reads", () => {
expect(request.config.params).toEqual({ ef: 200 });
});
+ it("converts an L2 distance into a similarity score (higher = better)", async () => {
+ // Mirrors the Python provider (#6435): 1 / (1 + distance), so closer scores higher.
+ const client = fakeClient();
+ client.vectorSearch.mockResolvedValue({
+ ...OK,
+ rows: [
+ { row: { id: "near", metadata: {} }, score: 0.5 },
+ { row: { id: "far", metadata: {} }, score: 2.0 },
+ ],
+ });
+
+ const results = await makeStore(client).search([1, 2, 3], 2);
+
+ expect(results[0].score).toBeCloseTo(1 / 1.5, 10);
+ expect(results[1].score).toBeCloseTo(1 / 3.0, 10);
+ });
+
+ it("leaves an unscored L2 row's score undefined instead of treating it as the closest match", async () => {
+ const client = fakeClient();
+ client.vectorSearch.mockResolvedValue({
+ ...OK,
+ rows: [{ row: { id: "unscored", metadata: {} } }],
+ });
+
+ const results = await makeStore(client).search([1, 2, 3], 1);
+
+ expect(results[0].score).toBeUndefined();
+ });
+
it("omits the filter when no filters are supplied", async () => {
const client = fakeClient();
client.vectorSearch.mockResolvedValue({ ...OK, rows: [] });