Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

feat(llms): support OpenAI v1 for Azure OpenAI #755

Merged
merged 2 commits into from
Nov 15, 2023
Merged

feat(llms): support OpenAI v1 for Azure OpenAI #755

Show file tree
Hide file tree
Changes from 1 commit
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
0ed89c6
feat(llms): support OpenAI v1 for Azure OpenAI
mspronesti Nov 15, 2023
5d54ecb
'Refactored by Sourcery'
Nov 15, 2023
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
135 changes: 90 additions & 45 deletions pandasai/llm/azure_openai.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -12,13 +12,13 @@
"""

import os
from typing import Any, Dict, Optional
from typing import Any, Dict, Optional, Union

import openai
from ..helpers import load_dotenv

from ..exceptions import APIKeyNotFoundError, MissingModelError
from ..prompts.base import AbstractPrompt
from ..helpers.openai import is_openai_v1
from .base import BaseOpenAI

load_dotenv()
Expand All @@ -29,73 +29,115 @@ class AzureOpenAI(BaseOpenAI):
This class uses `BaseOpenAI` class to support Azure OpenAI features.
"""

azure_endpoint: Union[str, None] = None
"""Your Azure Active Directory token.
Automatically inferred from env var `AZURE_OPENAI_AD_TOKEN` if not provided.
For more:
https://www.microsoft.com/en-us/security/business/identity-access/microsoft-entra-id.
"""
azure_ad_token: Union[str, None] = None
"""A function that returns an Azure Active Directory token.
Will be invoked on every request.
"""
azure_ad_token_provider: Union[str, None] = None
deployment_name: str
api_version: str = ""
"""Legacy, for openai<1.0.0 support."""
api_base: str
"""Legacy, for openai<1.0.0 support."""
api_type: str = "azure"
Comment on lines +32 to 48
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

The class attributes are well-documented. However, the azure_ad_token_provider should be a callable that returns a token, not a string. This needs to be corrected to reflect the proper type.

- azure_ad_token_provider: Union[str, None] = None
+ azure_ad_token_provider: Optional[Callable[[], str]] = None
Commitable suggestion

[!IMPORTANT]
Carefully review the code before committing. Ensure that it accurately replaces the highlighted code, contains no missing lines, and has no issues with indentation.

Suggested change
azure_endpoint: Union[str, None] = None
"""Your Azure Active Directory token.
Automatically inferred from env var `AZURE_OPENAI_AD_TOKEN` if not provided.
For more:
https://www.microsoft.com/en-us/security/business/identity-access/microsoft-entra-id.
"""
azure_ad_token: Union[str, None] = None
"""A function that returns an Azure Active Directory token.
Will be invoked on every request.
"""
azure_ad_token_provider: Union[str, None] = None
deployment_name: str
api_version: str = ""
"""Legacy, for openai<1.0.0 support."""
api_base: str
"""Legacy, for openai<1.0.0 support."""
api_type: str = "azure"
azure_endpoint: Union[str, None] = None
"""Your Azure Active Directory token.
Automatically inferred from env var `AZURE_OPENAI_AD_TOKEN` if not provided.
For more:
https://www.microsoft.com/en-us/security/business/identity-access/microsoft-entra-id.
"""
azure_ad_token: Union[str, None] = None
"""A function that returns an Azure Active Directory token.
Will be invoked on every request.
"""
azure_ad_token_provider: Optional[Callable[[], str]] = None
deployment_name: str
api_version: str = ""
"""Legacy, for openai<1.0.0 support."""
api_base: str
"""Legacy, for openai<1.0.0 support."""
api_type: str = "azure"

api_version: str
engine: str

def __init__(
self,
api_token: Optional[str] = None,
api_base: Optional[str] = None,
api_version: Optional[str] = None,
deployment_name: str = None,
is_chat_model: bool = True,
**kwargs,
self,
api_token: Optional[str] = None,
azure_endpoint: Union[str, None] = None,
azure_ad_token: Union[str, None] = None,
azure_ad_token_provider: Union[str, None] = None,
api_base: Optional[str] = None,
api_version: Optional[str] = None,
deployment_name: str = None,
is_chat_model: bool = True,
**kwargs,
):
"""
__init__ method of AzureOpenAI Class.

Args:
api_token (str): Azure OpenAI API token.
api_base (str): Base url of the Azure endpoint.
azure_endpoint (str): Azure endpoint.
It should look like the following:
<https://YOUR_RESOURCE_NAME.openai.azure.com/>
azure_ad_token (str): Your Azure Active Directory token.
Automatically inferred from env var `AZURE_OPENAI_AD_TOKEN` if not provided.
For more: https://www.microsoft.com/en-us/security/business/identity-access/microsoft-entra-id.
azure_ad_token_provider (str): A function that returns an Azure Active Directory token.
Will be invoked on every request.
api_version (str): Version of the Azure OpenAI API.
Be aware the API version may change.
api_base (str): Legacy, kept for backward compatibility with openai < 1.0
deployment_name (str): Custom name of the deployed model
is_chat_model (bool): Whether ``deployment_name`` corresponds to a Chat
or a Completion model.
**kwargs: Inference Parameters.
"""

self.api_token = api_token or os.getenv("OPENAI_API_KEY") or None
self.api_base = api_base or os.getenv("OPENAI_API_BASE") or None
self.api_token = (
api_token
or os.getenv("OPENAI_API_KEY")
or os.getenv("AZURE_OPENAI_API_KEY")
)
self.azure_endpoint = azure_endpoint or os.getenv("AZURE_OPENAI_ENDPOINT")
self.api_base = api_base or os.getenv("OPENAI_API_BASE")
self.api_version = api_version or os.getenv("OPENAI_API_VERSION")
if self.api_token is None:
raise APIKeyNotFoundError(
"Azure OpenAI key is required. Please add an environment variable "
"`OPENAI_API_KEY` or pass `api_token` as a named parameter"
)
if self.api_base is None:
raise APIKeyNotFoundError(
"Azure OpenAI base is required. Please add an environment variable "
"`OPENAI_API_BASE` or pass `api_base` as a named parameter"
)
if is_openai_v1():
if self.azure_endpoint is None:
raise APIKeyNotFoundError(
"Azure endpoint is required. Please add an environment variable "
"`AZURE_OPENAI_API_ENDPOINT` or pass `azure_endpoint` as a named parameter"
)
else:
if self.api_base is None:
raise APIKeyNotFoundError(
"Azure OpenAI base is required. Please add an environment variable "
"`OPENAI_API_BASE` or pass `api_base` as a named parameter"
)
if self.api_version is None:
raise APIKeyNotFoundError(
"Azure OpenAI version is required. Please add an environment variable "
"`OPENAI_API_VERSION` or pass `api_version` as a named parameter"
)
openai.api_key = self.api_token
openai.api_base = self.api_base
openai.api_version = self.api_version
openai.api_type = self.api_type

if deployment_name is None:
raise MissingModelError(
"No deployment name provided.",
"Please include deployment name from Azure dashboard.",
)

self.is_chat_model = is_chat_model
self.engine = deployment_name
self.azure_ad_token = azure_ad_token or os.getenv("AZURE_OPENAI_AD_TOKEN")
self.azure_ad_token_provider = azure_ad_token_provider
self._is_chat_model = is_chat_model
self.deployment_name = deployment_name

self.openai_proxy = kwargs.get("openai_proxy") or os.getenv("OPENAI_PROXY")
if self.openai_proxy:
openai.proxy = {"http": self.openai_proxy, "https": self.openai_proxy}

self._set_params(**kwargs)
# set the client
if self._is_chat_model:
if is_openai_v1():
self.client = openai.AzureOpenAI(**self._client_params).chat.completions
else:
self.client = openai.ChatCompletion
else:
if is_openai_v1():
self.client = openai.AzureOpenAI(**self._client_params).completions
else:
self.client = openai.Completion

@property
def _default_params(self) -> Dict[str, Any]:
Expand All @@ -106,27 +148,30 @@ def _default_params(self) -> Dict[str, Any]:
dict: A dictionary containing Default Params.

"""
return {**super()._default_params, "engine": self.engine}

def call(self, instruction: AbstractPrompt, suffix: str = "") -> str:
"""
Call the Azure OpenAI LLM.
return {**super()._default_params, "model" if is_openai_v1() else "engine": self.deployment_name}
Comment on lines 147 to +150
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

There is a syntax error in the _default_params method. The ternary operator is used incorrectly. It should be fixed to correctly return the dictionary with the appropriate key based on the API version.

- return {**super()._default_params, "model" if is_openai_v1() else "engine": self.deployment_name}
+ return {**super()._default_params, ("model" if is_openai_v1() else "engine"): self.deployment_name}
Commitable suggestion

[!IMPORTANT]
Carefully review the code before committing. Ensure that it accurately replaces the highlighted code, contains no missing lines, and has no issues with indentation.

Suggested change
dict: A dictionary containing Default Params.
"""
return {**super()._default_params, "engine": self.engine}
def call(self, instruction: AbstractPrompt, suffix: str = "") -> str:
"""
Call the Azure OpenAI LLM.
return {**super()._default_params, "model" if is_openai_v1() else "engine": self.deployment_name}
"""
return {**super()._default_params, ("model" if is_openai_v1() else "engine"): self.deployment_name}


Args:
instruction (AbstractPrompt): A prompt object with instruction for LLM.
suffix (str): Suffix to pass.

Returns:
str: LLM response.

"""
self.last_prompt = instruction.to_string() + suffix
@property
def _invocation_params(self) -> Dict[str, Any]:
"""Get the parameters used to invoke the model."""
if is_openai_v1():
return super()._invocation_params
else:
return {
**super()._invocation_params,
"api_type": self.api_type,
"api_version": self.api_version,
}

return (
self.chat_completion(self.last_prompt)
if self.is_chat_model
else self.completion(self.last_prompt)
)
@property
def _client_params(self) -> Dict[str, any]:
client_params = {
"api_version": self.api_version,
"azure_endpoint": self.azure_endpoint,
"azure_deployment": self.deployment_name,
"azure_ad_token": self.azure_ad_token,
"azure_ad_token_provider": self.azure_ad_token_provider,
}
return {**client_params, **super()._client_params}

@property
def type(self) -> str:
Comment on lines 147 to 176
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Note: This review was outside of the patch, so it was mapped to the patch with the greatest overlap. Original lines [142-174]

The properties _default_params, _invocation_params, and _client_params are well-implemented to handle different configurations based on the API version. However, there is a syntax error in the _default_params method. The ternary operator is used incorrectly. It should be fixed to correctly return the dictionary with the appropriate key based on the API version.

- return {**super()._default_params, "model" if is_openai_v1() else "engine": self.deployment_name}
+ return {**super()._default_params, ("model" if is_openai_v1() else "engine"): self.deployment_name}

Expand Down
59 changes: 27 additions & 32 deletions tests/llms/test_azure_openai.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -5,7 +5,11 @@
from pandasai.exceptions import APIKeyNotFoundError, MissingModelError
from pandasai.llm import AzureOpenAI

pytest.skip(allow_module_level=True)

class OpenAIObject:
def __init__(self, dictionary):
self.__dict__.update(dictionary)


class TestAzureOpenAILLM:
"""Unit tests for the Azure Openai LLM class"""
Expand All @@ -20,28 +24,28 @@ def test_type_without_endpoint(self):

def test_type_without_api_version(self):
with pytest.raises(APIKeyNotFoundError):
AzureOpenAI(api_token="test", api_base="test")
AzureOpenAI(api_token="test", azure_endpoint="test")

def test_type_without_deployment(self):
with pytest.raises(MissingModelError):
AzureOpenAI(api_token="test", api_base="test", api_version="test")
AzureOpenAI(api_token="test", azure_endpoint="test", api_version="test")

def test_type_with_token(self):
assert (
AzureOpenAI(
api_token="test",
api_base="test",
api_version="test",
deployment_name="test",
).type
== "azure-openai"
AzureOpenAI(
api_token="test",
azure_endpoint="test",
api_version="test",
deployment_name="test",
).type
== "azure-openai"
)
Comment on lines 25 to 42
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

The tests are checking for the correct exceptions when required parameters are not provided. This is good practice as it ensures that the class constructor correctly handles missing arguments. However, the exception APIKeyNotFoundError seems to be used for different missing parameters, which might be confusing. It might be more appropriate to have different exceptions for different types of configuration errors or to rename the exception to something more generic like ConfigurationError.


def test_proxy(self):
proxy = "http://proxy.mycompany.com:8080"
client = AzureOpenAI(
api_token="test",
api_base="test",
azure_endpoint="test",
api_version="test",
deployment_name="test",
openai_proxy=proxy,
Expand All @@ -53,7 +57,7 @@ def test_proxy(self):
def test_params_setting(self):
llm = AzureOpenAI(
api_token="test",
api_base="test",
azure_endpoint="test",
api_version="test",
deployment_name="Deployed-GPT-3",
is_chat_model=True,
Comment on lines 57 to 63
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

The test is setting up an AzureOpenAI instance with various parameters. However, the is_chat_model parameter is being set, which does not match the updated attribute name _is_chat_model as per the summary. This discrepancy should be corrected to ensure the test reflects the current state of the code.

- is_chat_model=True,
+ _is_chat_model=True,
Commitable suggestion

[!IMPORTANT]
Carefully review the code before committing. Ensure that it accurately replaces the highlighted code, contains no missing lines, and has no issues with indentation.

Suggested change
def test_params_setting(self):
llm = AzureOpenAI(
api_token="test",
api_base="test",
azure_endpoint="test",
api_version="test",
deployment_name="Deployed-GPT-3",
is_chat_model=True,
def test_params_setting(self):
llm = AzureOpenAI(
api_token="test",
azure_endpoint="test",
api_version="test",
deployment_name="Deployed-GPT-3",
_is_chat_model=True,

Expand All @@ -65,8 +69,8 @@ def test_params_setting(self):
stop=["\n"],
)

assert llm.engine == "Deployed-GPT-3"
assert llm.is_chat_model
assert llm.deployment_name == "Deployed-GPT-3"
assert llm._is_chat_model
assert llm.temperature == 0.5
assert llm.max_tokens == 50
assert llm.top_p == 1.0
Comment on lines 69 to 76
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

The assertions here are checking that the properties of the AzureOpenAI instance are set correctly. This is a good practice as it ensures that the instance variables are being assigned as expected. However, the test should also include assertions for the new _invocation_params and _client_params properties if they are critical to the operation of the class.




<!-- This is an auto-generated comment by CodeRabbit -->

Expand All @@ -75,9 +79,8 @@ def test_params_setting(self):
assert llm.stop == ["\n"]

def test_completion(self, mocker):
openai_mock = mocker.patch("openai.Completion.create")
expected_text = "This is the generated text."
openai_mock.return_value = OpenAIObject.construct_from(
expected_response = OpenAIObject(
{
"choices": [{"text": expected_text}],
"usage": {
Expand All @@ -91,34 +94,25 @@ def test_completion(self, mocker):

openai = AzureOpenAI(
api_token="test",
api_base="test",
azure_endpoint="test",
api_version="test",
deployment_name="test",
)
mocker.patch.object(openai, "completion", return_value=expected_response)
result = openai.completion("Some prompt.")

openai_mock.assert_called_once_with(
engine=openai.engine,
prompt="Some prompt.",
temperature=openai.temperature,
max_tokens=openai.max_tokens,
top_p=openai.top_p,
frequency_penalty=openai.frequency_penalty,
presence_penalty=openai.presence_penalty,
seed=openai.seed
)

assert result == expected_text
openai.completion.assert_called_once_with("Some prompt.")
assert result == expected_response

def test_chat_completion(self, mocker):
openai = AzureOpenAI(
api_token="test",
api_base="test",
azure_endpoint="test",
api_version="test",
deployment_name="test",
is_chat_model=True,
)
expected_response = OpenAIObject.construct_from(
expected_response = OpenAIObject(
{
"choices": [
{
Expand All @@ -135,4 +129,5 @@ def test_chat_completion(self, mocker):
mocker.patch.object(openai, "chat_completion", return_value=expected_response)

result = openai.chat_completion("Hi")
assert result == expected_response
openai.chat_completion.assert_called_once_with("Hi")
assert result == expected_response