# Copyright (c) Microsoft Corporation.
# Licensed under the MIT license.
import json
import logging
from typing import Any, MutableSequence, Optional
import httpx
from pyrit.common import net_utility
from pyrit.exceptions import (
EmptyResponseException,
PyritException,
handle_bad_request_exception,
pyrit_target_retry,
)
from pyrit.exceptions.exception_classes import RateLimitException
from pyrit.models import (
ChatMessageListDictContent,
DataTypeSerializer,
PromptRequestPiece,
PromptRequestResponse,
construct_response_from_request,
data_serializer_factory,
)
from pyrit.models.chat_message import ChatMessage
from pyrit.prompt_target import OpenAITarget, limit_requests_per_minute
logger = logging.getLogger(__name__)
# Supported image formats for Azure OpenAI GPT-4o,
# https://learn.microsoft.com/en-us/azure/ai-services/openai/concepts/use-your-image-data
AZURE_OPENAI_GPT4O_SUPPORTED_IMAGE_FORMATS = [".jpg", ".jpeg", ".png", ".gif", ".bmp", ".tiff", "tif"]
[docs]
class OpenAIChatTarget(OpenAITarget):
"""
This class facilitates multimodal (image and text) input and text output generation
This works with GPT3.5, GPT4, GPT4o, GPT-V, and other compatible models
"""
[docs]
def __init__(
self,
*,
max_completion_tokens: Optional[int] = None,
max_tokens: Optional[int] = None,
temperature: Optional[float] = None,
top_p: Optional[float] = None,
frequency_penalty: Optional[float] = None,
presence_penalty: Optional[float] = None,
seed: Optional[int] = None,
n: Optional[int] = None,
is_json_supported: bool = True,
extra_body_parameters: Optional[dict[str, Any]] = None,
**kwargs,
):
"""
Args:
model_name (str, Optional): The name of the model.
endpoint (str, Optional): The target URL for the OpenAI service.
api_key (str, Optional): The API key for accessing the Azure OpenAI service.
Defaults to the `OPENAI_CHAT_KEY` environment variable.
headers (str, Optional): Headers of the endpoint (JSON).
use_aad_auth (bool, Optional): When set to True, user authentication is used
instead of API Key. DefaultAzureCredential is taken for
https://cognitiveservices.azure.com/.default . Please run `az login` locally
to leverage user AuthN.
api_version (str, Optional): The version of the Azure OpenAI API. Defaults to
"2024-06-01".
max_requests_per_minute (int, Optional): Number of requests the target can handle per
minute before hitting a rate limit. The number of requests sent to the target
will be capped at the value provided.
max_completion_tokens (int, Optional): An upper bound for the number of tokens that
can be generated for a completion, including visible output tokens and
reasoning tokens.
NOTE: Specify this value when using an o1 series model.
max_tokens (int, Optional): The maximum number of tokens that can be
generated in the chat completion. This value can be used to control
costs for text generated via API.
This value is now deprecated in favor of `max_completion_tokens`, and IS NOT
COMPATIBLE with o1 series models.
temperature (float, Optional): The temperature parameter for controlling the
randomness of the response.
top_p (float, Optional): The top-p parameter for controlling the diversity of the
response.
frequency_penalty (float, Optional): The frequency penalty parameter for penalizing
frequently generated tokens.
presence_penalty (float, Optional): The presence penalty parameter for penalizing
tokens that are already present in the conversation history.
seed (int, Optional): If specified, openAI will make a best effort to sample deterministically,
such that repeated requests with the same seed and parameters should return the same result.
n (int, Optional): The number of completions to generate for each prompt.
is_json_supported (bool, Optional): If True, the target will supports formatting responses as JSON by
setting the response_format header. Official OpenAI models all support this, but if you are using
this target with different models, is_json_supported should be set correctly to avoid issues when
using adversarial infrastructure (e.g. Crescendo scorers will set this flag).
extra_body_parameters (dict, Optional): Additional parameters to be included in the request body.
httpx_client_kwargs (dict, Optional): Additional kwargs to be passed to the
`httpx.AsyncClient()` constructor.
For example, to specify a 3 minutes timeout: httpx_client_kwargs={"timeout": 180}
"""
super().__init__(**kwargs)
if max_completion_tokens and max_tokens:
raise ValueError("Cannot provide both max_tokens and max_completion_tokens.")
self._max_completion_tokens = max_completion_tokens
self._max_tokens = max_tokens
self._temperature = temperature
self._top_p = top_p
self._frequency_penalty = frequency_penalty
self._presence_penalty = presence_penalty
self._seed = seed
self._n = n
self._is_json_supported = is_json_supported
self._extra_body_parameters = extra_body_parameters
def _set_openai_env_configuration_vars(self) -> None:
self.model_name_environment_variable = "OPENAI_CHAT_MODEL"
self.endpoint_environment_variable = "OPENAI_CHAT_ENDPOINT"
self.api_key_environment_variable = "OPENAI_CHAT_KEY"
@limit_requests_per_minute
@pyrit_target_retry
async def send_prompt_async(self, *, prompt_request: PromptRequestResponse) -> PromptRequestResponse:
"""Asynchronously sends a prompt request and handles the response within a managed conversation context.
Args:
prompt_request (PromptRequestResponse): The prompt request response object.
Returns:
PromptRequestResponse: The updated conversation entry with the response from the prompt target.
"""
self._validate_request(prompt_request=prompt_request)
self.refresh_auth_headers()
request_piece: PromptRequestPiece = prompt_request.request_pieces[0]
is_json_response = self.is_response_format_json(request_piece)
conversation = self._memory.get_conversation(conversation_id=request_piece.conversation_id)
conversation.append(prompt_request)
logger.info(f"Sending the following prompt to the prompt target: {prompt_request}")
body = await self._construct_request_body(conversation=conversation, is_json_response=is_json_response)
params = {}
if self._api_version is not None:
params["api-version"] = self._api_version
try:
str_response: httpx.Response = await net_utility.make_request_and_raise_if_error_async(
endpoint_uri=self._endpoint,
method="POST",
headers=self._headers,
request_body=body,
params=params,
**self._httpx_client_kwargs,
)
except httpx.HTTPStatusError as StatusError:
if StatusError.response.status_code == 400:
# Handle Bad Request
return handle_bad_request_exception(
response_text=StatusError.response.text,
request=request_piece,
error_code=StatusError.response.status_code,
)
elif StatusError.response.status_code == 429:
raise RateLimitException()
else:
raise
logger.info(f'Received the following response from the prompt target "{str_response.text}"')
response: PromptRequestResponse = self._construct_prompt_response_from_openai_json(
open_ai_str_response=str_response.text, request_piece=request_piece
)
return response
async def _convert_local_image_to_data_url(self, image_path: str) -> str:
"""Converts a local image file to a data URL encoded in base64.
Args:
image_path (str): The file system path to the image file.
Raises:
FileNotFoundError: If no file is found at the specified `image_path`.
ValueError: If the image file's extension is not in the supported formats list.
Returns:
str: A string containing the MIME type and the base64-encoded data of the image, formatted as a data URL.
"""
ext = DataTypeSerializer.get_extension(image_path)
if ext.lower() not in AZURE_OPENAI_GPT4O_SUPPORTED_IMAGE_FORMATS:
raise ValueError(
f"Unsupported image format: {ext}. Supported formats are: {AZURE_OPENAI_GPT4O_SUPPORTED_IMAGE_FORMATS}"
)
mime_type = DataTypeSerializer.get_mime_type(image_path)
if not mime_type:
mime_type = "application/octet-stream"
image_serializer = data_serializer_factory(
category="prompt-memory-entries", value=image_path, data_type="image_path", extension=ext
)
base64_encoded_data = await image_serializer.read_data_base64()
# Azure OpenAI GPT-4o documentation doesn't specify the local image upload format for API.
# GPT-4o image upload format is determined using "view code" functionality in Azure OpenAI deployments
# The image upload format is same as GPT-4 Turbo.
# Construct the data URL, as per Azure OpenAI GPT-4 Turbo local image format
# https://learn.microsoft.com/en-us/azure/ai-services/openai/how-to/gpt-with-vision?tabs=rest%2Csystem-assigned%2Cresource#call-the-chat-completion-apis
return f"data:{mime_type};base64,{base64_encoded_data}"
async def _build_chat_messages_async(self, conversation: MutableSequence[PromptRequestResponse]) -> list[dict]:
"""Builds chat messages based on prompt request response entries.
Args:
conversation (list[PromptRequestResponse]): A list of PromptRequestResponse objects.
Returns:
list[dict]: The list of constructed chat messages.
"""
if self._is_text_message_format(conversation):
return self._build_chat_messages_for_text(conversation)
else:
return await self._build_chat_messages_for_multi_modal_async(conversation)
def _is_text_message_format(self, conversation: MutableSequence[PromptRequestResponse]) -> bool:
"""Checks if the request piece is in text message format.
Args:
conversation list[PromptRequestResponse]: The conversation
Returns:
bool: True if the request piece is in text message format, False otherwise.
"""
for turn in conversation:
if len(turn.request_pieces) != 1:
return False
if turn.request_pieces[0].converted_value_data_type != "text":
return False
return True
def _build_chat_messages_for_text(self, conversation: MutableSequence[PromptRequestResponse]) -> list[dict]:
"""
Builds chat messages based on prompt request response entries. This is needed because many
openai "compatible" models don't support ChatMessageListDictContent format (this is more univerally accepted)
Args:
conversation (list[PromptRequestResponse]): A list of PromptRequestResponse objects.
Returns:
list[dict]: The list of constructed chat messages.
"""
chat_messages: list[dict] = []
for prompt_req_resp_entry in conversation:
# validated to only have one text entry
if len(prompt_req_resp_entry.request_pieces) != 1:
raise ValueError("_build_chat_messages_for_text only supports a single prompt request piece.")
prompt_request_piece = prompt_req_resp_entry.request_pieces[0]
if prompt_request_piece.converted_value_data_type != "text":
raise ValueError("_build_chat_messages_for_text only supports text.")
message = ChatMessage(role=prompt_request_piece.role, content=prompt_request_piece.converted_value)
chat_messages.append(message.model_dump(exclude_none=True))
return chat_messages
async def _build_chat_messages_for_multi_modal_async(
self, conversation: MutableSequence[PromptRequestResponse]
) -> list[dict]:
"""
Builds chat messages based on prompt request response entries.
Args:
conversation (list[PromptRequestResponse]): A list of PromptRequestResponse objects.
Returns:
list[dict]: The list of constructed chat messages.
"""
chat_messages: list[dict] = []
for prompt_req_resp_entry in conversation:
prompt_request_pieces = prompt_req_resp_entry.request_pieces
content = []
role = None
for prompt_request_piece in prompt_request_pieces:
role = prompt_request_piece.role
if prompt_request_piece.converted_value_data_type == "text":
entry = {"type": "text", "text": prompt_request_piece.converted_value}
content.append(entry)
elif prompt_request_piece.converted_value_data_type == "image_path":
data_base64_encoded_url = await self._convert_local_image_to_data_url(
prompt_request_piece.converted_value
)
image_url_entry = {"url": data_base64_encoded_url}
entry = {"type": "image_url", "image_url": image_url_entry} # type: ignore
content.append(entry)
else:
raise ValueError(
f"Multimodal data type {prompt_request_piece.converted_value_data_type} is not yet supported."
)
if not role:
raise ValueError("No role could be determined from the prompt request pieces.")
chat_message = ChatMessageListDictContent(role=role, content=content) # type: ignore
chat_messages.append(chat_message.model_dump(exclude_none=True))
return chat_messages
async def _construct_request_body(
self, conversation: MutableSequence[PromptRequestResponse], is_json_response: bool
) -> dict:
messages = await self._build_chat_messages_async(conversation)
body_parameters = {
"model": self._model_name,
"max_completion_tokens": self._max_completion_tokens,
"max_tokens": self._max_tokens,
"temperature": self._temperature,
"top_p": self._top_p,
"frequency_penalty": self._frequency_penalty,
"presence_penalty": self._presence_penalty,
"stream": False,
"seed": self._seed,
"n": self._n,
"messages": messages,
"response_format": {"type": "json_object"} if is_json_response else None,
}
if self._extra_body_parameters:
for key, value in self._extra_body_parameters.items():
body_parameters[key] = value
# Filter out None values
return {k: v for k, v in body_parameters.items() if v is not None}
def _construct_prompt_response_from_openai_json(
self,
*,
open_ai_str_response: str,
request_piece: PromptRequestPiece,
) -> PromptRequestResponse:
try:
response = json.loads(open_ai_str_response)
except json.JSONDecodeError as e:
raise PyritException(message=f"Failed to parse JSON response. Please check your endpoint: {e}")
finish_reason = response["choices"][0]["finish_reason"]
extracted_response: str = ""
# finish_reason="stop" means API returned complete message and
# "length" means API returned incomplete message due to max_tokens limit.
if finish_reason in ["stop", "length"]:
extracted_response = response["choices"][0]["message"]["content"]
# Handle empty response
if not extracted_response:
logger.log(logging.ERROR, "The chat returned an empty response.")
raise EmptyResponseException(message="The chat returned an empty response.")
elif finish_reason == "content_filter":
return handle_bad_request_exception(
response_text=open_ai_str_response, request=request_piece, error_code=400, is_content_filter=True
)
else:
raise PyritException(message=f"Unknown finish_reason {finish_reason} from response: {response}")
return construct_response_from_request(request=request_piece, response_text_pieces=[extracted_response])
def _validate_request(self, *, prompt_request: PromptRequestResponse) -> None:
"""Validates the structure and content of a prompt request for compatibility of this target.
Args:
prompt_request (PromptRequestResponse): The prompt request response object.
Raises:
ValueError: If more than two request pieces are provided.
ValueError: If any of the request pieces have a data type other than 'text' or 'image_path'.
"""
converted_prompt_data_types = [
request_piece.converted_value_data_type for request_piece in prompt_request.request_pieces
]
# Some models may not support all of these
for prompt_data_type in converted_prompt_data_types:
if prompt_data_type not in ["text", "image_path"]:
raise ValueError(f"This target only supports text and image_path. Received: {prompt_data_type}.")
[docs]
def is_json_response_supported(self) -> bool:
"""Indicates that this target supports JSON response format."""
return self._is_json_supported
