diff --git a/docs/source/conf.py b/docs/source/conf.py
index a16c1325337ff130dc52e8600f424db4e04a86b9..7f3485122f1992618712a76ccfb24ee1e9a4bdb9 100644
--- a/docs/source/conf.py
+++ b/docs/source/conf.py
@@ -15,7 +15,7 @@ sys.path.insert(0, os.path.abspath("../..")) # Source code dir relative to this
project = "Semantic Router"
copyright = "2025, Aurelio AI"
author = "Aurelio AI"
-release = "0.1.0.dev10"
+release = "0.1.1"
# -- General configuration ---------------------------------------------------
# https://www.sphinx-doc.org/en/master/usage/configuration.html#general-configuration
diff --git a/pyproject.toml b/pyproject.toml
index 3375e9d4bb839ea050d25e3c37f94407c7188176..2a10684582c4384ff743a0229e0dbfd36ab3d9a9 100644
--- a/pyproject.toml
+++ b/pyproject.toml
@@ -1,6 +1,6 @@
[project]
name = "semantic-router"
-version = "0.1.0"
+version = "0.1.1"
description = "Super fast semantic router for AI decision making"
authors = [{ name = "Aurelio AI", email = "hello@aurelio.ai" }]
requires-python = ">=3.9,<3.14"
diff --git a/semantic_router/__init__.py b/semantic_router/__init__.py
index d9d8cba5af2fe6fdb0da5080315fc59915e9744f..3d7a4d7a867bfc2e120ac91dcaf5f4f70e99547b 100644
--- a/semantic_router/__init__.py
+++ b/semantic_router/__init__.py
@@ -3,4 +3,4 @@ from semantic_router.routers import HybridRouter, RouterConfig, SemanticRouter
__all__ = ["SemanticRouter", "HybridRouter", "Route", "RouterConfig"]
-__version__ = "0.1.0.dev10"
+__version__ = "0.1.1"
diff --git a/semantic_router/encoders/__init__.py b/semantic_router/encoders/__init__.py
index 5fdeb789098aa009996df1c6b264f1db1bbeaf82..764e7cafa30c0171d38955034684527e0921ba77 100644
--- a/semantic_router/encoders/__init__.py
+++ b/semantic_router/encoders/__init__.py
@@ -1,7 +1,8 @@
from typing import List, Optional
+from semantic_router.encoders.base import DenseEncoder, SparseEncoder # isort: skip
from semantic_router.encoders.aurelio import AurelioSparseEncoder
-from semantic_router.encoders.base import DenseEncoder, SparseEncoder
+from semantic_router.encoders.azure_openai import AzureOpenAIEncoder
from semantic_router.encoders.bedrock import BedrockEncoder
from semantic_router.encoders.bm25 import BM25Encoder
from semantic_router.encoders.clip import CLIPEncoder
@@ -13,7 +14,6 @@ from semantic_router.encoders.mistral import MistralEncoder
from semantic_router.encoders.openai import OpenAIEncoder
from semantic_router.encoders.tfidf import TfidfEncoder
from semantic_router.encoders.vit import VitEncoder
-from semantic_router.encoders.zure import AzureOpenAIEncoder
from semantic_router.schema import EncoderType, SparseEmbedding
__all__ = [
@@ -45,8 +45,7 @@ class AutoEncoder:
self.type = EncoderType(type)
self.name = name
if self.type == EncoderType.AZURE:
- # TODO should change `model` to `name` JB
- self.model = AzureOpenAIEncoder(model=name)
+ self.model = AzureOpenAIEncoder(name=name)
elif self.type == EncoderType.COHERE:
self.model = CohereEncoder(name=name)
elif self.type == EncoderType.OPENAI:
diff --git a/semantic_router/encoders/zure.py b/semantic_router/encoders/azure_openai.py
similarity index 51%
rename from semantic_router/encoders/zure.py
rename to semantic_router/encoders/azure_openai.py
index faab1c90d59336980fa9509620dd9ed2d70502ff..2ca45ada9a2559e657853b8f2f6af434379431fc 100644
--- a/semantic_router/encoders/zure.py
+++ b/semantic_router/encoders/azure_openai.py
@@ -1,8 +1,9 @@
import os
from asyncio import sleep as asleep
from time import sleep
-from typing import List, Optional, Union
+from typing import Any, Callable, Dict, List, Optional, Union
+import httpx
import openai
from openai import OpenAIError
from openai._types import NotGiven
@@ -24,100 +25,135 @@ class AzureOpenAIEncoder(DenseEncoder):
async_client: Optional[openai.AsyncAzureOpenAI] = None
dimensions: Union[int, NotGiven] = NotGiven()
type: str = "azure"
- api_key: Optional[str] = None
- deployment_name: Optional[str] = None
- azure_endpoint: Optional[str] = None
- api_version: Optional[str] = None
- model: Optional[str] = None
+ deployment_name: str | None = None
max_retries: int = 3
def __init__(
self,
- api_key: Optional[str] = None,
- deployment_name: Optional[str] = None,
- azure_endpoint: Optional[str] = None,
- api_version: Optional[str] = None,
- model: Optional[str] = None, # TODO we should change to `name` JB
+ name: Optional[str] = None,
+ azure_endpoint: str | None = None,
+ api_version: str | None = None,
+ api_key: str | None = None,
+ azure_ad_token: str | None = None,
+ azure_ad_token_provider: Callable[[], str] | None = None,
+ http_client_options: Optional[Dict[str, Any]] = None,
+ deployment_name: str = EncoderDefault.AZURE.value["deployment_name"],
score_threshold: float = 0.82,
dimensions: Union[int, NotGiven] = NotGiven(),
max_retries: int = 3,
):
"""Initialize the AzureOpenAIEncoder.
- :param api_key: The API key for the Azure OpenAI API.
- :type api_key: str
- :param deployment_name: The name of the deployment to use.
- :type deployment_name: str
:param azure_endpoint: The endpoint for the Azure OpenAI API.
- :type azure_endpoint: str
+ Example: ``https://accountname.openai.azure.com``
+ :type azure_endpoint: str, optional
+
:param api_version: The version of the API to use.
- :type api_version: str
- :param model: The model to use.
- :type model: str
- :param score_threshold: The score threshold for the embeddings.
- :type score_threshold: float
- :param dimensions: The dimensions of the embeddings.
- :type dimensions: int
- :param max_retries: The maximum number of retries for the API call.
- :type max_retries: int
+ Example: ``"2025-02-01-preview"``
+ :type api_version: str, optional
+
+ :param api_key: The API key for the Azure OpenAI API.
+ :type api_key: str, optional
+
+ :param azure_ad_token: The Azure AD/Entra ID token for authentication.
+ https://www.microsoft.com/en-us/security/business/identity-access/microsoft-entra-id
+ :type azure_ad_token: str, optional
+
+ :param azure_ad_token_provider: A callable function that returns an Azure AD/Entra ID token.
+ :type azure_ad_token_provider: Callable[[], str], optional
+
+ :param http_client_options: Dictionary of options to configure httpx client
+ Example:
+ {
+ "proxies": "http://proxy.server:8080",
+ "timeout": 20.0,
+ "headers": {"Authorization": "Bearer xyz"}
+ }
+ :type http_client_options: Dict[str, Any], optional
+
+ :param deployment_name: The name of the model deployment to use.
+ :type deployment_name: str, optional
+
+ :param score_threshold: The score threshold for filtering embeddings.
+ Default is ``0.82``.
+ :type score_threshold: float, optional
+
+ :param dimensions: The number of dimensions for the embeddings. If not given, it defaults to the model's default setting.
+ :type dimensions: int, optional
+
+ :param max_retries: The maximum number of retries for API calls in case of failures.
+ Default is ``3``.
+ :type max_retries: int, optional
"""
- name = deployment_name
if name is None:
- name = EncoderDefault.AZURE.value["embedding_model"]
+ name = deployment_name
+ if name is None:
+ name = EncoderDefault.AZURE.value["embedding_model"]
super().__init__(name=name, score_threshold=score_threshold)
- self.api_key = api_key
+
+ azure_endpoint = azure_endpoint or os.getenv("AZURE_OPENAI_ENDPOINT")
+ if not azure_endpoint:
+ raise ValueError("No Azure OpenAI endpoint provided.")
+
+ api_version = api_version or os.getenv("AZURE_OPENAI_API_VERSION")
+ if not api_version:
+ raise ValueError("No Azure OpenAI API version provided.")
+
+ if not (
+ azure_ad_token
+ or azure_ad_token_provider
+ or api_key
+ or os.getenv("AZURE_OPENAI_API_KEY")
+ ):
+ raise ValueError(
+ "No authentication method provided. Please provide either `azure_ad_token`, "
+ "`azure_ad_token_provider`, or `api_key`."
+ )
+
+ # Only check API Key if no AD token or provider is used
+ if not azure_ad_token and not azure_ad_token_provider:
+ api_key = api_key or os.getenv("AZURE_OPENAI_API_KEY")
+ if not api_key:
+ raise ValueError("No Azure OpenAI API key provided.")
+
self.deployment_name = deployment_name
- self.azure_endpoint = azure_endpoint
- self.api_version = api_version
- self.model = model
+
# set dimensions to support openai embed 3 dimensions param
self.dimensions = dimensions
- if self.api_key is None:
- self.api_key = os.getenv("AZURE_OPENAI_API_KEY")
- if self.api_key is None:
- raise ValueError("No Azure OpenAI API key provided.")
+
if max_retries is not None:
self.max_retries = max_retries
- if self.deployment_name is None:
- self.deployment_name = EncoderDefault.AZURE.value["deployment_name"]
- # deployment_name may still be None, but it is optional in the API
- if self.azure_endpoint is None:
- self.azure_endpoint = os.getenv("AZURE_OPENAI_ENDPOINT")
- if self.azure_endpoint is None:
- raise ValueError("No Azure OpenAI endpoint provided.")
- if self.api_version is None:
- self.api_version = os.getenv("AZURE_OPENAI_API_VERSION")
- if self.api_version is None:
- raise ValueError("No Azure OpenAI API version provided.")
- if self.model is None:
- self.model = os.getenv("AZURE_OPENAI_MODEL")
- if self.model is None:
- raise ValueError("No Azure OpenAI model provided.")
- assert (
- self.api_key is not None
- and self.azure_endpoint is not None
- and self.api_version is not None
- and self.model is not None
+
+ # Only create HTTP clients if options are provided
+ sync_http_client = (
+ httpx.Client(**http_client_options) if http_client_options else None
+ )
+ async_http_client = (
+ httpx.AsyncClient(**http_client_options) if http_client_options else None
)
+ assert azure_endpoint is not None and self.deployment_name is not None
+
try:
self.client = openai.AzureOpenAI(
- azure_deployment=(
- str(self.deployment_name) if self.deployment_name else None
- ),
- api_key=str(self.api_key),
- azure_endpoint=str(self.azure_endpoint),
- api_version=str(self.api_version),
+ azure_endpoint=azure_endpoint,
+ api_version=api_version,
+ api_key=api_key,
+ azure_ad_token=azure_ad_token,
+ azure_ad_token_provider=azure_ad_token_provider,
+ http_client=sync_http_client,
)
self.async_client = openai.AsyncAzureOpenAI(
- azure_deployment=(
- str(self.deployment_name) if self.deployment_name else None
- ),
- api_key=str(self.api_key),
- azure_endpoint=str(self.azure_endpoint),
- api_version=str(self.api_version),
+ azure_endpoint=azure_endpoint,
+ api_version=api_version,
+ api_key=api_key,
+ azure_ad_token=azure_ad_token,
+ azure_ad_token_provider=azure_ad_token_provider,
+ http_client=async_http_client,
)
+
except Exception as e:
+ logger.error("OpenAI API client failed to initialize. Error: %s", e)
raise ValueError(
f"OpenAI API client failed to initialize. Error: {e}"
) from e
@@ -139,7 +175,7 @@ class AzureOpenAIEncoder(DenseEncoder):
try:
embeds = self.client.embeddings.create(
input=docs,
- model=str(self.model),
+ model=str(self.deployment_name),
dimensions=self.dimensions,
)
if embeds.data:
@@ -149,12 +185,12 @@ class AzureOpenAIEncoder(DenseEncoder):
if self.max_retries != 0 and j < self.max_retries:
sleep(2**j)
logger.warning(
- f"Retrying in {2**j} seconds due to OpenAIError: {e}"
+ "Retrying in %d seconds due to OpenAIError: %s", 2**j, e
)
else:
raise
except Exception as e:
- logger.error(f"Azure OpenAI API call failed. Error: {e}")
+ logger.error("Azure OpenAI API call failed. Error: %s", e)
raise ValueError(f"Azure OpenAI API call failed. Error: {e}") from e
if (
@@ -183,23 +219,22 @@ class AzureOpenAIEncoder(DenseEncoder):
try:
embeds = await self.async_client.embeddings.create(
input=docs,
- model=str(self.model),
+ model=str(self.deployment_name),
dimensions=self.dimensions,
)
if embeds.data:
break
-
except OpenAIError as e:
logger.error("Exception occurred", exc_info=True)
if self.max_retries != 0 and j < self.max_retries:
await asleep(2**j)
logger.warning(
- f"Retrying in {2**j} seconds due to OpenAIError: {e}"
+ "Retrying in %d seconds due to OpenAIError: %s", 2**j, e
)
else:
raise
except Exception as e:
- logger.error(f"Azure OpenAI API call failed. Error: {e}")
+ logger.error("Azure OpenAI API call failed. Error: %s", e)
raise ValueError(f"Azure OpenAI API call failed. Error: {e}") from e
if (
diff --git a/tests/unit/encoders/test_azure.py b/tests/unit/encoders/test_azure.py
index 01a495a44ec1f4e58f9ee0a0c45f211d62226d76..15554805e5b2891d72a15fce39cecd0c8f2f9299 100644
--- a/tests/unit/encoders/test_azure.py
+++ b/tests/unit/encoders/test_azure.py
@@ -23,11 +23,12 @@ def mock_openai_async_client():
@pytest.fixture
def openai_encoder(mock_openai_client, mock_openai_async_client):
return AzureOpenAIEncoder(
+ azure_endpoint="https://test-endpoint.openai.azure.com",
+ api_version="test-version",
api_key="test_api_key",
+ http_client_options={"timeout": 10},
deployment_name="test-deployment",
- azure_endpoint="test_endpoint",
- api_version="test_version",
- model="test_model",
+ dimensions=1536,
max_retries=2,
)
@@ -84,7 +85,7 @@ class TestAzureOpenAIEncoder:
mocker.patch.object(
openai_encoder.client.embeddings, "create", side_effect=responses
)
- with patch("semantic_router.encoders.zure.sleep", return_value=None):
+ with patch("semantic_router.encoders.azure_openai.sleep", return_value=None):
embeddings = openai_encoder(["test document"])
assert embeddings == [[0.1, 0.2]]
@@ -96,7 +97,7 @@ class TestAzureOpenAIEncoder:
"create",
side_effect=Exception("Non-OpenAIError"),
)
- with patch("semantic_router.encoders.zure.sleep", return_value=None):
+ with patch("semantic_router.encoders.azure_openai.sleep", return_value=None):
with pytest.raises(ValueError) as e:
openai_encoder(["test document"])
@@ -124,7 +125,7 @@ class TestAzureOpenAIEncoder:
mocker.patch.object(
openai_encoder.client.embeddings, "create", side_effect=responses
)
- with patch("semantic_router.encoders.zure.sleep", return_value=None):
+ with patch("semantic_router.encoders.azure_openai.sleep", return_value=None):
embeddings = openai_encoder(["test document"])
assert embeddings == [[0.1, 0.2]]
@@ -150,7 +151,7 @@ class TestAzureOpenAIEncoder:
mocker.patch("time.sleep", return_value=None) # To speed up the test
# Patch the sleep function in the encoder module to avoid actual sleep
- with patch("semantic_router.encoders.zure.sleep", return_value=None):
+ with patch("semantic_router.encoders.azure_openai.sleep", return_value=None):
result = openai_encoder(["test document"])
assert result == [[0.1, 0.2, 0.3]]
@@ -176,7 +177,7 @@ class TestAzureOpenAIEncoder:
mocker.patch("time.sleep", return_value=None) # To speed up the test
# Patch the sleep function in the encoder module to avoid actual sleep
- with patch("semantic_router.encoders.zure.sleep", return_value=None):
+ with patch("semantic_router.encoders.azure_openai.sleep", return_value=None):
with pytest.raises(OpenAIError):
openai_encoder(["test document"])
@@ -207,7 +208,7 @@ class TestAzureOpenAIEncoder:
mocker.patch("asyncio.sleep", return_value=None) # To speed up the test
# Patch the asleep function in the encoder module to avoid actual sleep
- with patch("semantic_router.encoders.zure.asleep", return_value=None):
+ with patch("semantic_router.encoders.azure_openai.asleep", return_value=None):
result = await openai_encoder.acall(["test document"])
assert result == [[0.1, 0.2, 0.3]]
@@ -226,7 +227,7 @@ class TestAzureOpenAIEncoder:
mocker.patch("asyncio.sleep", return_value=None) # To speed up the test
# Patch the asleep function in the encoder module to avoid actual sleep
- with patch("semantic_router.encoders.zure.asleep", return_value=None):
+ with patch("semantic_router.encoders.azure_openai.asleep", return_value=None):
with pytest.raises(OpenAIError):
await openai_encoder.acall(["test document"])
diff --git a/uv.lock b/uv.lock
index f859bf7600b00bff08297b00f81ee13955672999..855adaa29460cd9fc2e582c47b21f447937db1dd 100644
--- a/uv.lock
+++ b/uv.lock
@@ -3326,7 +3326,7 @@ wheels = [
[[package]]
name = "semantic-router"
-version = "0.1.0"
+version = "0.1.1"
source = { editable = "." }
dependencies = [
{ name = "aiohttp" },