From e5f15485871bdbedf50369b20c3be3909983bcaa Mon Sep 17 00:00:00 2001
From: Pluviobyte <Pluviobyte@users.noreply.github.com>
Date: Fri, 29 May 2026 06:07:17 +0000
Subject: [PATCH] fix(openai-adapters): read Gemini embeddings from correct
 response field
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

The Gemini adapter's embed() reads `data.batchEmbedContents` from the
`:batchEmbedContents` response, but that field does not exist — the Gemini
API returns the vectors under `embeddings`. Every embeddings request
through the Gemini adapter therefore throws "Cannot read properties of
undefined (reading 'map')", breaking @codebase / docs indexing for anyone
using Gemini as their embeddings provider.

The core Gemini implementation (core/llm/llms/Gemini.ts) already reads
`data.embeddings`; this aligns the adapter with it and drops the usage
block that read two more non-existent fields. Adds a regression test that
feeds the real response shape and asserts the parsed vectors.

Co-authored-by: Cursor <cursoragent@cursor.com>
---
 .../openai-adapters/src/apis/Gemini.test.ts   | 50 ++++++++++++++++++-
 packages/openai-adapters/src/apis/Gemini.ts   |  6 +--
 2 files changed, 50 insertions(+), 6 deletions(-)

diff --git a/packages/openai-adapters/src/apis/Gemini.test.ts b/packages/openai-adapters/src/apis/Gemini.test.ts
index 3b47620bc71..e1e81d6a743 100644
--- a/packages/openai-adapters/src/apis/Gemini.test.ts
+++ b/packages/openai-adapters/src/apis/Gemini.test.ts
@@ -1,7 +1,16 @@
-import { describe, expect, it } from "vitest";
+import { afterEach, describe, expect, it, vi } from "vitest";
 
 import { GeminiApi } from "./Gemini.js";
 
+// Mock the fetch package so the embeddings test can stub the HTTP response.
+vi.mock("@continuedev/fetch", async () => {
+  const actual = await vi.importActual("@continuedev/fetch");
+  return {
+    ...actual,
+    fetchwithRequestOptions: vi.fn(),
+  };
+});
+
 describe("GeminiApi", () => {
   const api = new GeminiApi({
     provider: "gemini",
@@ -133,4 +142,43 @@ describe("GeminiApi", () => {
       expect(result.contents[2].role).toBe("user");
     });
   });
+
+  describe("embed", () => {
+    afterEach(() => {
+      vi.clearAllMocks();
+      vi.unstubAllGlobals();
+    });
+
+    it("parses the 'embeddings' field from the batchEmbedContents response", async () => {
+      const mockFetch = vi.fn().mockResolvedValue(
+        new Response(
+          JSON.stringify({
+            embeddings: [
+              { values: [0.1, 0.2, 0.3] },
+              { values: [0.4, 0.5, 0.6] },
+            ],
+          }),
+          { headers: { "Content-Type": "application/json" } },
+        ),
+      );
+      vi.stubGlobal("fetch", mockFetch);
+      const fetchPackage = await import("@continuedev/fetch");
+      vi.mocked(fetchPackage.fetchwithRequestOptions).mockImplementation(
+        mockFetch as any,
+      );
+
+      const response = await api.embed({
+        model: "gemini-embedding-001",
+        input: ["Hello", "World"],
+      });
+
+      expect(mockFetch).toHaveBeenCalledTimes(1);
+      const [url] = mockFetch.mock.calls[0];
+      expect(url.toString()).toContain(":batchEmbedContents");
+      expect(response.data.map((d) => d.embedding)).toEqual([
+        [0.1, 0.2, 0.3],
+        [0.4, 0.5, 0.6],
+      ]);
+    });
+  });
 });
diff --git a/packages/openai-adapters/src/apis/Gemini.ts b/packages/openai-adapters/src/apis/Gemini.ts
index f7298267849..fc573b76598 100644
--- a/packages/openai-adapters/src/apis/Gemini.ts
+++ b/packages/openai-adapters/src/apis/Gemini.ts
@@ -494,11 +494,7 @@ export class GeminiApi implements BaseLlmApi {
     const data = (await response.json()) as any;
     return embedding({
       model: body.model,
-      usage: {
-        total_tokens: data.total_tokens,
-        prompt_tokens: data.prompt_tokens,
-      },
-      data: data.batchEmbedContents.map((embedding: any) => embedding.values),
+      data: data.embeddings.map((embedding: any) => embedding.values),
     });
   }