bump openai version to ^1.0.0

coverage.xml

 <?xml version="1.0" ?>
-<coverage version="7.3.2" timestamp="1702463592393" lines-valid="334" lines-covered="334" line-rate="1" branches-covered="0" branches-valid="0" branch-rate="0" complexity="0">
+<coverage version="7.3.2" timestamp="1702485028371" lines-valid="344" lines-covered="344" line-rate="1" branches-covered="0" branches-valid="0" branch-rate="0" complexity="0">
 	<!-- Generated by coverage.py: https://coverage.readthedocs.io/en/7.3.2 -->
 	<!-- Based on https://raw.githubusercontent.com/cobertura/web/master/htdocs/xml/coverage-04.dtd -->
 	<sources>
@@ -353,27 +353,37 @@
 						<line number="8" hits="1"/>
 						<line number="11" hits="1"/>
 						<line number="12" hits="1"/>
-						<line number="13" hits="1"/>
 						<line number="14" hits="1"/>
-						<line number="15" hits="1"/>
-						<line number="16" hits="1"/>
-						<line number="18" hits="1"/>
+						<line number="19" hits="1"/>
+						<line number="20" hits="1"/>
+						<line number="21" hits="1"/>
 						<line number="22" hits="1"/>
 						<line number="23" hits="1"/>
+						<line number="24" hits="1"/>
+						<line number="25" hits="1"/>
 						<line number="26" hits="1"/>
-						<line number="27" hits="1"/>
 						<line number="28" hits="1"/>
 						<line number="29" hits="1"/>
 						<line number="30" hits="1"/>
 						<line number="31" hits="1"/>
 						<line number="32" hits="1"/>
-						<line number="33" hits="1"/>
-						<line number="34" hits="1"/>
 						<line number="35" hits="1"/>
 						<line number="36" hits="1"/>
 						<line number="37" hits="1"/>
+						<line number="38" hits="1"/>
 						<line number="39" hits="1"/>
 						<line number="40" hits="1"/>
+						<line number="41" hits="1"/>
+						<line number="42" hits="1"/>
+						<line number="43" hits="1"/>
+						<line number="44" hits="1"/>
+						<line number="45" hits="1"/>
+						<line number="46" hits="1"/>
+						<line number="47" hits="1"/>
+						<line number="49" hits="1"/>
+						<line number="50" hits="1"/>
+						<line number="52" hits="1"/>
+						<line number="53" hits="1"/>
 					</lines>
 				</class>
 			</classes>

pyproject.toml

@@ -14,7 +14,7 @@ readme = "README.md"
 [tool.poetry.dependencies]
 python = "^3.10"
 pydantic = "^1.8.2"
-openai = "^0.28.1"
+openai = "^1.0.0"
 cohere = "^4.32"
 numpy = "^1.25.2"
 pinecone-text = "^0.7.0"

semantic_router/encoders/openai.py

@@ -2,39 +2,52 @@ import os
 from time import sleep
 
 import openai
-from openai.error import OpenAIError, RateLimitError, ServiceUnavailableError
+from openai import OpenAIError
 
 from semantic_router.encoders import BaseEncoder
 from semantic_router.utils.logger import logger
 
 
 class OpenAIEncoder(BaseEncoder):
-    def __init__(self, name: str, openai_api_key: str | None = None):
+    client: openai.Client | None
+
+    def __init__(
+        self,
+        name: str = os.getenv("OPENAI_MODEL_NAME", "text-embedding-ada-002"),
+        openai_api_key: str | None = None,
+    ):
         super().__init__(name=name)
         openai.api_key = openai_api_key or os.getenv("OPENAI_API_KEY")
         if openai.api_key is None:
             raise ValueError("OpenAI API key cannot be 'None'.")
+        try:
+            self.client = openai.Client()
+        except Exception as e:
+            raise ValueError(f"OpenAI API client failed to initialize. Error: {e}")
 
     def __call__(self, docs: list[str]) -> list[list[float]]:
-        """Encode a list of texts using the OpenAI API. Returns a list of
-        vector embeddings.
-        """
-        res = None
+        if self.client is None:
+            raise ValueError("OpenAI client is not initialized.")
+        embeds = None
         error_message = ""
 
-        # exponential backoff
-        for j in range(5):
+        # Exponential backoff
+        for j in range(3):
             try:
                 logger.info(f"Encoding {len(docs)} documents...")
-                res = openai.Embedding.create(input=docs, engine=self.name)
-                if isinstance(res, dict) and "data" in res:
+                embeds = self.client.embeddings.create(input=docs, model=self.name)
+                if isinstance(embeds, dict) and "data" in embeds:
                     break
-            except (RateLimitError, ServiceUnavailableError, OpenAIError) as e:
-                logger.warning(f"Retrying in {2**j} seconds...")
+            except OpenAIError as e:
                 sleep(2**j)
                 error_message = str(e)
-        if not res or not isinstance(res, dict) or "data" not in res:
-            raise ValueError(f"OpenAI API call failed. Error: {error_message}")
+                logger.warning(f"Retrying in {2**j} seconds...")
+            except Exception as e:
+                logger.error(f"OpenAI API call failed. Error: {error_message}")
+                raise ValueError(f"OpenAI API call failed. Error: {e}")
+
+        if not embeds or not isinstance(embeds, dict) or "data" not in embeds:
+            raise ValueError(f"No embeddings returned. Error: {error_message}")
 
-        embeds = [r["embedding"] for r in res["data"]]
-        return embeds
+        embeddings = [r["embedding"] for r in embeds["data"]]
+        return embeddings

tests/unit/encoders/test_openai.py

-import openai
 import pytest
-from openai.error import RateLimitError
+from openai import OpenAIError
 
 from semantic_router.encoders import OpenAIEncoder
 
 
 @pytest.fixture
 def openai_encoder(mocker):
-    mocker.patch("openai.Embedding.create")
-    return OpenAIEncoder(name="test-engine", openai_api_key="test_api_key")
+    mocker.patch("openai.Client")
+    return OpenAIEncoder(openai_api_key="test_api_key")
 
 
 class TestOpenAIEncoder:
-    def test_initialization_with_api_key(self, openai_encoder):
-        assert openai.api_key == "test_api_key", "API key should be set correctly"
-        assert openai_encoder.name == "test-engine", "Engine name not set correctly"
-
-    def test_initialization_without_api_key(self, mocker, monkeypatch):
-        monkeypatch.delenv("OPENAI_API_KEY", raising=False)
-        mocker.patch("openai.Embedding.create")
-        with pytest.raises(ValueError):
-            OpenAIEncoder(name="test-engine")
-
-    def test_call_method_success(self, openai_encoder, mocker):
-        mocker.patch(
-            "openai.Embedding.create",
-            return_value={"data": [{"embedding": [0.1, 0.2, 0.3]}]},
+    def test_openai_encoder_init_success(self, mocker):
+        mocker.patch("os.getenv", return_value="fake-api-key")
+        encoder = OpenAIEncoder()
+        assert encoder.client is not None
+
+    def test_openai_encoder_init_no_api_key(self, mocker):
+        mocker.patch("os.getenv", return_value=None)
+        with pytest.raises(ValueError) as e:
+            OpenAIEncoder()
+        assert "OpenAI API key cannot be 'None'." in str(e.value)
+
+    def test_openai_encoder_call_uninitialized_client(self, openai_encoder):
+        # Set the client to None to simulate an uninitialized client
+        openai_encoder.client = None
+        with pytest.raises(ValueError) as e:
+            openai_encoder(["test document"])
+        assert "OpenAI client is not initialized." in str(e.value)
+
+    def test_openai_encoder_init_exception(self, mocker):
+        mocker.patch("os.getenv", return_value="fake-api-key")
+        mocker.patch("openai.Client", side_effect=Exception("Initialization error"))
+        with pytest.raises(ValueError) as e:
+            OpenAIEncoder()
+        assert (
+            "OpenAI API client failed to initialize. Error: Initialization error"
+            in str(e.value)
         )
 
-        result = openai_encoder(["test"])
-        assert isinstance(result, list), "Result should be a list"
-        assert len(result) == 1 and len(result[0]) == 3, "Result list size is incorrect"
-
-    def test_call_method_rate_limit_error__raises_value_error_after_max_retries(
-        self, openai_encoder, mocker
-    ):
-        mocker.patch("semantic_router.encoders.openai.sleep")
-        mocker.patch(
-            "openai.Embedding.create",
-            side_effect=RateLimitError(message="rate limit exceeded", http_status=429),
+    def test_openai_encoder_call_success(self, openai_encoder, mocker):
+        mocker.patch("os.getenv", return_value="fake-api-key")
+        mocker.patch.object(
+            openai_encoder.client.embeddings,
+            "create",
+            return_value={"data": [{"embedding": [0.1, 0.2]}]},
         )
-
-        with pytest.raises(ValueError):
-            openai_encoder(["test"])
-
-    def test_call_method_failure(self, openai_encoder, mocker):
-        mocker.patch("openai.Embedding.create", return_value={})
-
-        with pytest.raises(ValueError):
-            openai_encoder(["test"])
-
-    def test_call_method_rate_limit_error__exponential_backoff_single_retry(
-        self, openai_encoder, mocker
-    ):
-        mock_sleep = mocker.patch("semantic_router.encoders.openai.sleep")
-        mocker.patch(
-            "openai.Embedding.create",
-            side_effect=[
-                RateLimitError("rate limit exceeded"),
-                {"data": [{"embedding": [1, 2, 3]}]},
-            ],
+        embeddings = openai_encoder(["test document"])
+        assert embeddings == [[0.1, 0.2]]
+
+    def test_openai_encoder_call_with_retries(self, openai_encoder, mocker):
+        mocker.patch("os.getenv", return_value="fake-api-key")
+        mocker.patch("time.sleep", return_value=None)  # To speed up the test
+        mocker.patch.object(
+            openai_encoder.client.embeddings,
+            "create",
+            side_effect=OpenAIError("Test error"),
         )
-
-        openai_encoder(["sample text"])
-
-        mock_sleep.assert_called_once_with(1)  # 2**0
-
-    def test_call_method_rate_limit_error__exponential_backoff_multiple_retries(
-        self, openai_encoder, mocker
-    ):
-        mock_sleep = mocker.patch("semantic_router.encoders.openai.sleep")
-        mocker.patch(
-            "openai.Embedding.create",
-            side_effect=[
-                RateLimitError("rate limit exceeded"),
-                RateLimitError("rate limit exceeded"),
-                {"data": [{"embedding": [1, 2, 3]}]},
-            ],
+        with pytest.raises(ValueError) as e:
+            openai_encoder(["test document"])
+        assert "No embeddings returned. Error: Test error" in str(e.value)
+
+    def test_openai_encoder_call_failure_non_openai_error(self, openai_encoder, mocker):
+        mocker.patch("os.getenv", return_value="fake-api-key")
+        mocker.patch("time.sleep", return_value=None)  # To speed up the test
+        mocker.patch.object(
+            openai_encoder.client.embeddings,
+            "create",
+            side_effect=Exception("Non-OpenAIError"),
         )
-
-        openai_encoder(["sample text"])
-
-        assert mock_sleep.call_count == 2
-        mock_sleep.assert_any_call(1)  # 2**0
-        mock_sleep.assert_any_call(2)  # 2**1
-
-    def test_call_method_rate_limit_error__exponential_backoff_max_retries_exceeded(
-        self, openai_encoder, mocker
-    ):
-        mock_sleep = mocker.patch("semantic_router.encoders.openai.sleep")
-        mocker.patch(
-            "openai.Embedding.create", side_effect=RateLimitError("rate limit exceeded")
+        with pytest.raises(ValueError) as e:
+            openai_encoder(["test document"])
+        assert "OpenAI API call failed. Error: Non-OpenAIError" in str(e.value)
+
+    def test_openai_encoder_call_successful_retry(self, openai_encoder, mocker):
+        mocker.patch("os.getenv", return_value="fake-api-key")
+        mocker.patch("time.sleep", return_value=None)  # To speed up the test
+        responses = [OpenAIError("Test error"), {"data": [{"embedding": [0.1, 0.2]}]}]
+        mocker.patch.object(
+            openai_encoder.client.embeddings, "create", side_effect=responses
         )
-
-        with pytest.raises(ValueError):
-            openai_encoder(["sample text"])
-
-        assert mock_sleep.call_count == 5  # Assuming 5 retries
-        mock_sleep.assert_any_call(1)  # 2**0
-        mock_sleep.assert_any_call(2)  # 2**1
-        mock_sleep.assert_any_call(4)  # 2**2
-        mock_sleep.assert_any_call(8)  # 2**3
-        mock_sleep.assert_any_call(16)  # 2**4
-
-    def test_call_method_rate_limit_error__exponential_backoff_successful(
-        self, openai_encoder, mocker
-    ):
-        mock_sleep = mocker.patch("semantic_router.encoders.openai.sleep")
-        mocker.patch(
-            "openai.Embedding.create",
-            side_effect=[
-                RateLimitError("rate limit exceeded"),
-                RateLimitError("rate limit exceeded"),
-                {"data": [{"embedding": [1, 2, 3]}]},
-            ],
-        )
-
-        embeddings = openai_encoder(["sample text"])
-
-        assert mock_sleep.call_count == 2
-        assert embeddings == [[1, 2, 3]]
+        embeddings = openai_encoder(["test document"])
+        assert embeddings == [[0.1, 0.2]]

walkthrough.ipynb

@@ -128,7 +128,7 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "from semantic_router.router import RouteLayer\n",
+    "from semantic_router.layer import RouteLayer\n",
     "\n",
     "dl = RouteLayer(encoder=encoder, routes=routes)"
    ]