diff --git a/sweepai/core/vector_db_test.py b/sweepai/core/vector_db_test.py
new file mode 100644
index 0000000000..382d6ea38d
--- /dev/null
+++ b/sweepai/core/vector_db_test.py
@@ -0,0 +1,86 @@
+import pytest
+import numpy as np
+from unittest.mock import patch, MagicMock
+from sweepai.core.vector_db import (
+    cosine_similarity,
+    embed_text_array,
+    openai_call_embedding,
+    openai_with_expo_backoff,
+    normalize_l2,
+)
+from sweepai.config.server import CACHE_DIRECTORY
+
+# Mock the Cache class
+@pytest.fixture
+def mock_cache():
+    with patch('sweepai.core.vector_db.Cache') as mock:
+        yield mock
+
+# Mock the openai client
+@pytest.fixture
+def mock_openai_client():
+    with patch('sweepai.core.vector_db.get_embeddings_client') as mock:
+        yield mock
+
+def test_embed_text_array(mock_openai_client):
+    mock_openai_client.return_value.embeddings.create.return_value.data = [
+        MagicMock(embedding=list(range(512))) for _ in range(2)
+    ]
+    
+    texts = ["Hello, world!", "Test embedding"]
+    result = embed_text_array(texts)
+    
+    assert len(result) == 1  # One batch
+    assert result[0].shape == (2, 512)  # Two embeddings, 512 dimensions each
+    mock_openai_client.return_value.embeddings.create.assert_called_once()
+
+def test_normalize_l2():
+    x = np.array([3, 4])
+    normalized = normalize_l2(x)
+    assert np.allclose(normalized, np.array([0.6, 0.8]))
+
+    X = np.array([[3, 4], [6, 8]])
+    normalized = normalize_l2(X)
+    expected = np.array([[0.6, 0.8], [0.6, 0.8]])
+    assert np.allclose(normalized, expected)
+
+def test_cosine_similarity():
+    a = np.array([[1, 0]])
+    B = np.array([[1, 0], [0, 1], [1, 1]])
+    result = cosine_similarity(a, B)
+    expected = np.array([[1, 0, 1/np.sqrt(2)]])
+    assert np.allclose(result, expected)
+
+def test_openai_with_expo_backoff(mock_cache):
+    mock_cache.return_value.get.side_effect = [None, np.array([1, 2, 3])]
+    mock_cache.return_value.set = MagicMock()
+
+    with patch('sweepai.core.vector_db.openai_call_embedding', return_value=np.array([[4, 5, 6]])):
+        result = openai_with_expo_backoff(["uncached text", "cached text"])
+
+    assert np.array_equal(result, np.array([[4, 5, 6], [1, 2, 3]]))
+    mock_cache.return_value.set.assert_called_once()
+
+def test_openai_call_embedding_token_limit():
+    long_text = "a" * 10000  # Assuming this exceeds the token limit
+    with patch('sweepai.core.vector_db.tiktoken_client.count', return_value=10000):
+        with patch('sweepai.core.vector_db.tiktoken_client.truncate_string', return_value="truncated"):
+            with patch('sweepai.core.vector_db.openai_call_embedding_router', side_effect=[
+                openai.BadRequestError("maximum context length"),
+                np.array([[1, 2, 3]])
+            ]) as mock_router:
+                result = openai_call_embedding([long_text])
+                
+                assert np.array_equal(result, np.array([[1, 2, 3]]))
+                assert mock_router.call_count == 2
+                mock_router.assert_called_with(["truncated"], "document")
+
+@pytest.mark.parametrize("exception,expected_calls", [
+    (requests.exceptions.Timeout(), 5),
+    (Exception("Unknown error"), 1)
+])
+def test_openai_with_expo_backoff_retries(exception, expected_calls):
+    with patch('sweepai.core.vector_db.openai_call_embedding', side_effect=exception):
+        with pytest.raises(Exception):
+            openai_with_expo_backoff(["test"])
+        assert openai_call_embedding.call_count == expected_calls
\ No newline at end of file