webscout 8.2.9__py3-none-any.whl → 2026.1.19__py3-none-any.whl

webscout/Provider/Openai.py

@@ -1,496 +1,337 @@
-import json
-from ..AIutel import Optimizers
-from ..AIutel import Conversation
-from ..AIutel import AwesomePrompts, sanitize_stream
-from ..AIbase import  Provider, AsyncProvider
-
-from webscout import exceptions
-from typing import Any, AsyncGenerator, Dict, List, Optional, Union
-import requests
-import httpx
-#----------------------------------------------------------OpenAI-----------------------------------
-class OPENAI(Provider):
-    def __init__(
-        self,
-        api_key: str,
-        is_conversation: bool = True,
-        max_tokens: int = 600,
-        temperature: float = 1,
-        presence_penalty: int = 0,
-        frequency_penalty: int = 0,
-        top_p: float = 1,
-        model: str = "gpt-3.5-turbo",
-        timeout: int = 30,
-        intro: str = None,
-        filepath: str = None,
-        update_file: bool = True,
-        proxies: dict = {},
-        history_offset: int = 10250,
-        act: str = None,
-        base_url: str = "https://api.openai.com/v1/chat/completions",
-    ):
-        """Instantiates OPENAI
-
-        Args:
-            api_key (key): OpenAI's API key.
-            is_conversation (bool, optional): Flag for chatting conversationally. Defaults to True.
-            max_tokens (int, optional): Maximum number of tokens to be generated upon completion. Defaults to 600.
-            temperature (float, optional): Charge of the generated text's randomness. Defaults to 1.
-            presence_penalty (int, optional): Chances of topic being repeated. Defaults to 0.
-            frequency_penalty (int, optional): Chances of word being repeated. Defaults to 0.
-            top_p (float, optional): Sampling threshold during inference time. Defaults to 0.999.
-            model (str, optional): LLM model name. Defaults to "gpt-3.5-turbo".
-            timeout (int, optional): Http request timeout. Defaults to 30.
-            intro (str, optional): Conversation introductory prompt. Defaults to None.
-            filepath (str, optional): Path to file containing conversation history. Defaults to None.
-            update_file (bool, optional): Add new prompts and responses to the file. Defaults to True.
-            proxies (dict, optional): Http request proxies. Defaults to {}.
-            history_offset (int, optional): Limit conversation history to this number of last texts. Defaults to 10250.
-            act (str|int, optional): Awesome prompt key or index. (Used as intro). Defaults to None.
-        """
-        self.is_conversation = is_conversation
-        self.max_tokens_to_sample = max_tokens
-        self.api_key = api_key
-        self.model = model
-        self.temperature = temperature
-        self.presence_penalty = presence_penalty
-        self.frequency_penalty = frequency_penalty
-        self.top_p = top_p
-        self.chat_endpoint = base_url
-        self.stream_chunk_size = 64
-        self.timeout = timeout
-        self.last_response = {}
-        self.headers = {
-            "Content-Type": "application/json",
-            "Authorization": f"Bearer {self.api_key}",
-        }
-        self.session = requests.session()
-        self.__available_optimizers = (
-            method
-            for method in dir(Optimizers)
-            if callable(getattr(Optimizers, method)) and not method.startswith("__")
-        )
-        self.session.headers.update(self.headers)
-        Conversation.intro = (
-            AwesomePrompts().get_act(
-                act, raise_not_found=True, default=None, case_insensitive=True
-            )
-            if act
-            else intro or Conversation.intro
-        )
-        self.conversation = Conversation(
-            is_conversation, self.max_tokens_to_sample, filepath, update_file
-        )
-        self.conversation.history_offset = history_offset
-        self.session.proxies = proxies
-
-    def ask(
-        self,
-        prompt: str,
-        stream: bool = False,
-        raw: bool = False,
-        optimizer: str = None,
-        conversationally: bool = False,
-    ) -> dict:
-        """Chat with AI
-
-        Args:
-            prompt (str): Prompt to be send.
-            stream (bool, optional): Flag for streaming response. Defaults to False.
-            raw (bool, optional): Stream back raw response as received. Defaults to False.
-            optimizer (str, optional): Prompt optimizer name - `[code, shell_command]`. Defaults to None.
-            conversationally (bool, optional): Chat conversationally when using optimizer. Defaults to False.
-        Returns:
-           dict : {}
-        ```json
-        {
-            "id": "chatcmpl-TaREJpBZsRVQFRFic1wIA7Q7XfnaD",
-            "object": "chat.completion",
-            "created": 1704623244,
-            "model": "gpt-3.5-turbo",
-            "usage": {
-                "prompt_tokens": 0,
-                "completion_tokens": 0,
-                "total_tokens": 0
-                },
-            "choices": [
-                {
-                    "message": {
-                        "role": "assistant",
-                        "content": "Hello! How can I assist you today?"
-                },
-                "finish_reason": "stop",
-                "index": 0
-                }
-            ]
-        }
-        ```
-        """
-        conversation_prompt = self.conversation.gen_complete_prompt(prompt)
-        if optimizer:
-            if optimizer in self.__available_optimizers:
-                conversation_prompt = getattr(Optimizers, optimizer)(
-                    conversation_prompt if conversationally else prompt
-                )
-            else:
-                raise exceptions.FailedToGenerateResponseError(
-                    f"Optimizer is not one of {self.__available_optimizers}"
-                )
-        self.session.headers.update(self.headers)
-        payload = {
-            "frequency_penalty": self.frequency_penalty,
-            "messages": [{"content": conversation_prompt, "role": "user"}],
-            "model": self.model,
-            "presence_penalty": self.presence_penalty,
-            "stream": stream,
-            "temperature": self.temperature,
-            "top_p": self.top_p,
-        }
-
-        def for_stream():
-            response = self.session.post(
-                self.chat_endpoint, json=payload, stream=True, timeout=self.timeout
-            )
-            if not response.ok:
-                raise exceptions.FailedToGenerateResponseError(
-                    f"Failed to generate response - ({response.status_code}, {response.reason}) - {response.text}"
-                )
-
-            message_load = ""
-            for value in response.iter_lines(
-                decode_unicode=True,
-                delimiter="" if raw else "data:",
-                chunk_size=self.stream_chunk_size,
-            ):
-                try:
-                    resp = json.loads(value)
-                    incomplete_message = self.get_message(resp)
-                    if incomplete_message:
-                        message_load += incomplete_message
-                        resp["choices"][0]["delta"]["content"] = message_load
-                        self.last_response.update(resp)
-                        yield value if raw else resp
-                    elif raw:
-                        yield value
-                except json.decoder.JSONDecodeError:
-                    pass
-            self.conversation.update_chat_history(
-                prompt, self.get_message(self.last_response)
-            )
-
-        def for_non_stream():
-            response = self.session.post(
-                self.chat_endpoint, json=payload, stream=False, timeout=self.timeout
-            )
-            if (
-                not response.ok
-                or not response.headers.get("Content-Type", "") == "application/json"
-            ):
-                raise exceptions.FailedToGenerateResponseError(
-                    f"Failed to generate response - ({response.status_code}, {response.reason}) - {response.text}"
-                )
-            resp = response.json()
-            self.last_response.update(resp)
-            self.conversation.update_chat_history(
-                prompt, self.get_message(self.last_response)
-            )
-            return resp
-
-        return for_stream() if stream else for_non_stream()
-
-    def chat(
-        self,
-        prompt: str,
-        stream: bool = False,
-        optimizer: str = None,
-        conversationally: bool = False,
-    ) -> str:
-        """Generate response `str`
-        Args:
-            prompt (str): Prompt to be send.
-            stream (bool, optional): Flag for streaming response. Defaults to False.
-            optimizer (str, optional): Prompt optimizer name - `[code, shell_command]`. Defaults to None.
-            conversationally (bool, optional): Chat conversationally when using optimizer. Defaults to False.
-        Returns:
-            str: Response generated
-        """
-
-        def for_stream():
-            for response in self.ask(
-                prompt, True, optimizer=optimizer, conversationally=conversationally
-            ):
-                yield self.get_message(response)
-
-        def for_non_stream():
-            return self.get_message(
-                self.ask(
-                    prompt,
-                    False,
-                    optimizer=optimizer,
-                    conversationally=conversationally,
-                )
-            )
-
-        return for_stream() if stream else for_non_stream()
-
-    def get_message(self, response: dict) -> str:
-        """Retrieves message only from response
-
-        Args:
-            response (dict): Response generated by `self.ask`
-
-        Returns:
-            str: Message extracted
-        """
-        assert isinstance(response, dict), "Response should be of dict data-type only"
-        try:
-            if response["choices"][0].get("delta"):
-                return response["choices"][0]["delta"]["content"]
-            return response["choices"][0]["message"]["content"]
-        except KeyError:
-            return ""
-class AsyncOPENAI(AsyncProvider):
-    def __init__(
-        self,
-        api_key: str,
-        is_conversation: bool = True,
-        max_tokens: int = 600,
-        temperature: float = 1,
-        presence_penalty: int = 0,
-        frequency_penalty: int = 0,
-        top_p: float = 1,
-        model: str = "gpt-3.5-turbo",
-        timeout: int = 30,
-        intro: str = None,
-        filepath: str = None,
-        update_file: bool = True,
-        proxies: dict = {},
-        history_offset: int = 10250,
-        act: str = None,
-    ):
-        """Instantiates OPENAI
-
-        Args:
-            api_key (key): OpenAI's API key.
-            is_conversation (bool, optional): Flag for chatting conversationally. Defaults to True.
-            max_tokens (int, optional): Maximum number of tokens to be generated upon completion. Defaults to 600.
-            temperature (float, optional): Charge of the generated text's randomness. Defaults to 1.
-            presence_penalty (int, optional): Chances of topic being repeated. Defaults to 0.
-            frequency_penalty (int, optional): Chances of word being repeated. Defaults to 0.
-            top_p (float, optional): Sampling threshold during inference time. Defaults to 0.999.
-            model (str, optional): LLM model name. Defaults to "gpt-3.5-turbo".
-            timeout (int, optional): Http request timeout. Defaults to 30.
-            intro (str, optional): Conversation introductory prompt. Defaults to None.
-            filepath (str, optional): Path to file containing conversation history. Defaults to None.
-            update_file (bool, optional): Add new prompts and responses to the file. Defaults to True.
-            proxies (dict, optional): Http request proxies. Defaults to {}.
-            history_offset (int, optional): Limit conversation history to this number of last texts. Defaults to 10250.
-            act (str|int, optional): Awesome prompt key or index. (Used as intro). Defaults to None.
-        """
-        self.is_conversation = is_conversation
-        self.max_tokens_to_sample = max_tokens
-        self.api_key = api_key
-        self.model = model
-        self.temperature = temperature
-        self.presence_penalty = presence_penalty
-        self.frequency_penalty = frequency_penalty
-        self.top_p = top_p
-        self.chat_endpoint = "https://api.openai.com/v1/chat/completions"
-        self.stream_chunk_size = 64
-        self.timeout = timeout
-        self.last_response = {}
-        self.headers = {
-            "Content-Type": "application/json",
-            "Authorization": f"Bearer {self.api_key}",
-        }
-
-        self.__available_optimizers = (
-            method
-            for method in dir(Optimizers)
-            if callable(getattr(Optimizers, method)) and not method.startswith("__")
-        )
-        Conversation.intro = (
-            AwesomePrompts().get_act(
-                act, raise_not_found=True, default=None, case_insensitive=True
-            )
-            if act
-            else intro or Conversation.intro
-        )
-        self.conversation = Conversation(
-            is_conversation, self.max_tokens_to_sample, filepath, update_file
-        )
-        self.conversation.history_offset = history_offset
-        self.session = httpx.AsyncClient(
-            headers=self.headers,
-            proxies=proxies,
-        )
-
-    async def ask(
-        self,
-        prompt: str,
-        stream: bool = False,
-        raw: bool = False,
-        optimizer: str = None,
-        conversationally: bool = False,
-        tools: Optional[List[Dict[str, Any]]] = None,
-    ) -> Union[dict, AsyncGenerator]:
-        """Chat with AI asynchronously.
-
-        Args:
-            prompt (str): Prompt to be send.
-            stream (bool, optional): Flag for streaming response. Defaults to False.
-            raw (bool, optional): Stream back raw response as received. Defaults to False.
-            optimizer (str, optional): Prompt optimizer name - `[code, shell_command]`. Defaults to None.
-            conversationally (bool, optional): Chat conversationally when using optimizer. Defaults to False.
-            tools (Optional[List[Dict[str, Any]]], optional): List of tools to be used. Defaults to None.
-        Returns:
-           dict|AsyncGenerator : ai content.
-        ```json
-        {
-            "id": "chatcmpl-TaREJpBZsRVQFRFic1wIA7Q7XfnaD",
-            "object": "chat.completion",
-            "created": 1704623244,
-            "model": "gpt-3.5-turbo",
-            "usage": {
-                "prompt_tokens": 0,
-                "completion_tokens": 0,
-                "total_tokens": 0
-                },
-            "choices": [
-                {
-                    "message": {
-                        "role": "assistant",
-                        "content": "Hello! How can I assist you today?"
-                },
-                "finish_reason": "stop",
-                "index": 0
-                }
-            ]
-        }
-        ```
-        """
-        conversation_prompt = self.conversation.gen_complete_prompt(prompt)
-        if optimizer:
-            if optimizer in self.__available_optimizers:
-                conversation_prompt = getattr(Optimizers, optimizer)(
-                    conversation_prompt if conversationally else prompt
-                )
-            else:
-                raise Exception(
-                    f"Optimizer is not one of {self.__available_optimizers}"
-                )
-        payload = {
-            "frequency_penalty": self.frequency_penalty,
-            "messages": [{"content": conversation_prompt, "role": "user"}],
-            "model": self.model,
-            "presence_penalty": self.presence_penalty,
-            "stream": stream,
-            "temperature": self.temperature,
-            "top_p": self.top_p,
-        }
-
-        async def for_stream():
-            async with self.session.stream(
-                "POST", self.chat_endpoint, json=payload, timeout=self.timeout
-            ) as response:
-                if not response.is_success:
-                    raise Exception(
-                        f"Failed to generate response - ({response.status_code}, {response.reason_phrase})"
-                    )
-
-                message_load = ""
-                async for value in response.aiter_lines():
-                    try:
-
-                        resp = sanitize_stream(value)
-                        incomplete_message = await self.get_message(resp)
-                        if incomplete_message:
-                            message_load += incomplete_message
-                            resp["choices"][0]["delta"]["content"] = message_load
-                            self.last_response.update(resp)
-                            yield value if raw else resp
-                        elif raw:
-                            yield value
-                    except json.decoder.JSONDecodeError:
-                        pass
-            self.conversation.update_chat_history(
-                prompt, await self.get_message(self.last_response)
-            )
-
-        async def for_non_stream():
-            response = httpx.post(
-                self.chat_endpoint,
-                json=payload,
-                timeout=self.timeout,
-                headers=self.headers,
-            )
-            if (
-                not response.is_success
-                or not response.headers.get("Content-Type", "") == "application/json"
-            ):
-                raise Exception(
-                    f"Failed to generate response - ({response.status_code}, {response.reason_phrase})"
-                )
-            resp = response.json()
-            self.last_response.update(resp)
-            self.conversation.update_chat_history(
-                prompt, await self.get_message(self.last_response)
-            )
-            return resp
-
-        return for_stream() if stream else await for_non_stream()
-
-    async def chat(
-        self,
-        prompt: str,
-        stream: bool = False,
-        optimizer: str = None,
-        conversationally: bool = False,
-        tools: Optional[List[Dict[str, Any]]] = None,
-    ) -> Union[str, AsyncGenerator]:
-        """Generate response `str` asynchronously.
-        Args:
-            prompt (str): Prompt to be send.
-            stream (bool, optional): Flag for streaming response. Defaults to False.
-            optimizer (str, optional): Prompt optimizer name - `[code, shell_command]`. Defaults to None.
-            conversationally (bool, optional): Chat conversationally when using optimizer. Defaults to False.
-            tools (Optional[List[Dict[str, Any]]], optional): List of tools to be used. Defaults to None.
-        Returns:
-            str|AsyncGenerator: Response generated
-        """
-
-        async def for_stream():
-            async_ask = await self.ask(
-                prompt, True, optimizer=optimizer, conversationally=conversationally
-            )
-            async for response in async_ask:
-                yield await self.get_message(response)
-
-        async def for_non_stream():
-            return await self.get_message(
-                await self.ask(
-                    prompt,
-                    False,
-                    optimizer=optimizer,
-                    conversationally=conversationally,
-                )
-            )
-
-        return for_stream() if stream else await for_non_stream()
-
-    async def get_message(self, response: dict) -> str:
-        """Retrieves message only from response asynchronously.
-
-        Args:
-            response (dict): Response generated by `self.ask`
-
-        Returns:
-            str: Message extracted
-        """
-        assert isinstance(response, dict), "Response should be of dict data-type only"
-        try:
-            if response["choices"][0].get("delta"):
-                return response["choices"][0]["delta"]["content"]
-            return response["choices"][0]["message"]["content"]
-        except KeyError:
-            return ""
+import json
+from typing import Any, Dict, Generator, Optional, Union, cast
+
+from curl_cffi import CurlError
+from curl_cffi.requests import Session
+
+from webscout import exceptions
+from webscout.AIbase import Provider, Response
+from webscout.AIutel import AwesomePrompts, Conversation, Optimizers, sanitize_stream
+from webscout.litagent import LitAgent
+
+
+class OpenAI(Provider):
+    """
+    A class to interact with the OpenAI API with LitAgent user-agent.
+    """
+
+    required_auth = True
+
+    @classmethod
+    def get_models(cls, api_key: Optional[str] = None):
+        """Fetch available models from OpenAI API.
+
+        Args:
+            api_key (str, optional): OpenAI API key
+
+        Returns:
+            list: List of available model IDs
+        """
+        if not api_key:
+            return []
+        try:
+            # Use a temporary curl_cffi session for this class method
+            temp_session = Session()
+            headers = {
+                "Authorization": f"Bearer {api_key}",
+            }
+
+            response = temp_session.get(
+                "https://api.openai.com/v1/models", headers=headers, impersonate="chrome110"
+            )
+
+            if response.status_code != 200:
+                raise Exception(
+                    f"API request failed with status {response.status_code}: {response.text}"
+                )
+
+            data = response.json()
+            if "data" in data and isinstance(data["data"], list):
+                return [model["id"] for model in data["data"] if "id" in model]
+            raise Exception("Invalid response format from API")
+
+        except (CurlError, Exception) as e:
+            raise Exception(f"Failed to fetch models: {str(e)}")
+
+    @staticmethod
+    def _openai_extractor(chunk: Union[str, Dict[str, Any]]) -> Optional[str]:
+        """Extracts content from OpenAI stream JSON objects."""
+        if isinstance(chunk, dict):
+            return chunk.get("choices", [{}])[0].get("delta", {}).get("content")
+        return None
+
+    def __init__(
+        self,
+        api_key: str,
+        is_conversation: bool = True,
+        max_tokens: int = 600,
+        temperature: float = 1,
+        presence_penalty: int = 0,
+        frequency_penalty: int = 0,
+        top_p: float = 1,
+        model: str = "gpt-3.5-turbo",
+        timeout: int = 30,
+        intro: Optional[str] = None,
+        filepath: Optional[str] = None,
+        update_file: bool = True,
+        proxies: dict = {},
+        history_offset: int = 10250,
+        act: Optional[str] = None,
+        base_url: str = "https://api.openai.com/v1/chat/completions",
+        system_prompt: str = "You are a helpful assistant.",
+        browser: str = "chrome",
+    ):
+        """Initializes the OpenAI API client."""
+        self.url = base_url
+
+        # Initialize LitAgent
+        self.agent = LitAgent()
+        self.fingerprint = self.agent.generate_fingerprint(browser)
+        self.api_key = api_key
+        # Use the fingerprint for headers
+        self.headers = {
+            "Accept": self.fingerprint["accept"],
+            "Accept-Language": self.fingerprint["accept_language"],
+            "User-Agent": self.fingerprint.get("user_agent", ""),
+        }
+        if self.api_key:
+            self.headers["Authorization"] = f"Bearer {self.api_key}"
+
+        # Initialize curl_cffi Session
+        self.session = Session()
+        # Update curl_cffi session headers and proxies
+        self.session.headers.update(self.headers)
+        if proxies:
+            self.session.proxies.update(cast(Any, proxies))  # Assign proxies directly
+        self.system_prompt = system_prompt
+        self.is_conversation = is_conversation
+        self.max_tokens_to_sample = max_tokens
+        self.timeout = timeout
+        self.last_response = {}
+        self.model = model
+        self.temperature = temperature
+        self.presence_penalty = presence_penalty
+        self.frequency_penalty = frequency_penalty
+        self.top_p = top_p
+
+        # Fetch available models
+        try:
+            self.available_models = self.get_models(self.api_key)
+        except Exception:
+            self.available_models = []
+
+        if self.available_models and self.model not in self.available_models:
+            raise ValueError(f"Invalid model: {self.model}. Choose from: {self.available_models}")
+
+        self.__available_optimizers = (
+            method
+            for method in dir(Optimizers)
+            if callable(getattr(Optimizers, method)) and not method.startswith("__")
+        )
+        self.conversation = Conversation(
+            is_conversation, self.max_tokens_to_sample, filepath, update_file
+        )
+        self.conversation.history_offset = history_offset
+
+        if act:
+            self.conversation.intro = (
+                AwesomePrompts().get_act(
+                    cast(Union[str, int], act),
+                    default=self.conversation.intro,
+                    case_insensitive=True,
+                )
+                or self.conversation.intro
+            )
+        elif intro:
+            self.conversation.intro = intro
+
+    def refresh_identity(self, browser: Optional[str] = None):
+        """
+        Refreshes the browser identity fingerprint.
+
+        Args:
+            browser: Specific browser to use for the new fingerprint
+        """
+        browser = browser or self.fingerprint.get("browser_type", "chrome")
+        self.fingerprint = self.agent.generate_fingerprint(browser)
+
+        # Update headers with new fingerprint (only relevant ones)
+        self.headers.update(
+            {
+                "Accept": self.fingerprint["accept"],
+                "Accept-Language": self.fingerprint["accept_language"],
+            }
+        )
+
+        # Update session headers
+        self.session.headers.update(self.headers)
+
+        return self.fingerprint
+
+    def ask(
+        self,
+        prompt: str,
+        stream: bool = False,
+        raw: bool = False,
+        optimizer: Optional[str] = None,
+        conversationally: bool = False,
+        **kwargs: Any,
+    ) -> Response:
+        conversation_prompt = self.conversation.gen_complete_prompt(prompt)
+        if optimizer:
+            if optimizer in self.__available_optimizers:
+                conversation_prompt = getattr(Optimizers, optimizer)(
+                    conversation_prompt if conversationally else prompt
+                )
+            else:
+                raise exceptions.FailedToGenerateResponseError(
+                    f"Optimizer is not one of {self.__available_optimizers}"
+                )
+
+        # Payload construction
+        payload = {
+            "model": self.model,
+            "messages": [
+                {"role": "system", "content": self.system_prompt},
+                {"role": "user", "content": conversation_prompt},
+            ],
+            "stream": stream,
+            "temperature": self.temperature,
+            "top_p": self.top_p,
+            "presence_penalty": self.presence_penalty,
+            "frequency_penalty": self.frequency_penalty,
+        }
+
+        def for_stream():
+            streaming_text = ""
+            try:
+                # Use curl_cffi session post with impersonate
+                response = self.session.post(
+                    self.url,
+                    data=json.dumps(payload),
+                    stream=True,
+                    timeout=self.timeout,
+                    impersonate="chrome110",
+                )
+                response.raise_for_status()
+
+                # Use sanitize_stream
+                processed_stream = sanitize_stream(
+                    data=response.iter_content(chunk_size=None),
+                    intro_value="data:",
+                    to_json=True,
+                    skip_markers=["[DONE]"],
+                    content_extractor=self._openai_extractor,
+                    yield_raw_on_error=False,
+                    raw=raw,
+                )
+
+                for content_chunk in processed_stream:
+                    if raw:
+                        yield content_chunk
+                    else:
+                        if content_chunk and isinstance(content_chunk, str):
+                            streaming_text += content_chunk
+                            resp = dict(text=content_chunk)
+                            yield resp if not raw else content_chunk
+
+            except CurlError as e:
+                raise exceptions.FailedToGenerateResponseError(
+                    f"Request failed (CurlError): {str(e)}"
+                ) from e
+            except Exception as e:
+                raise exceptions.FailedToGenerateResponseError(
+                    f"Request failed ({type(e).__name__}): {str(e)}"
+                ) from e
+            finally:
+                if streaming_text:
+                    self.last_response = {"text": streaming_text}
+                    self.conversation.update_chat_history(prompt, streaming_text)
+
+        def for_non_stream():
+            try:
+                # Use curl_cffi session post with impersonate for non-streaming
+                response = self.session.post(
+                    self.url,
+                    data=json.dumps(payload),
+                    timeout=self.timeout,
+                    impersonate="chrome110",
+                )
+                response.raise_for_status()
+
+                response_text = response.text
+
+                # Use sanitize_stream to parse the non-streaming JSON response
+                processed_stream = sanitize_stream(
+                    data=response_text,
+                    to_json=True,
+                    intro_value=None,
+                    content_extractor=lambda chunk: chunk.get("choices", [{}])[0]
+                    .get("message", {})
+                    .get("content")
+                    if isinstance(chunk, dict)
+                    else None,
+                    yield_raw_on_error=False,
+                    raw=raw,
+                )
+                # Extract the single result
+                content = next(processed_stream, None)
+                if raw:
+                    return content
+                content = content if isinstance(content, str) else ""
+
+                self.last_response = {"text": content}
+                self.conversation.update_chat_history(prompt, content)
+                return self.last_response if not raw else content
+
+            except CurlError as e:
+                raise exceptions.FailedToGenerateResponseError(
+                    f"Request failed (CurlError): {e}"
+                ) from e
+            except Exception as e:
+                err_text = ""
+                if hasattr(e, "response"):
+                    response_obj = getattr(e, "response")
+                    if hasattr(response_obj, "text"):
+                        err_text = getattr(response_obj, "text")
+                raise exceptions.FailedToGenerateResponseError(
+                    f"Request failed ({type(e).__name__}): {e} - {err_text}"
+                ) from e
+
+        return for_stream() if stream else for_non_stream()
+
+    def chat(
+        self,
+        prompt: str,
+        stream: bool = False,
+        optimizer: Optional[str] = None,
+        conversationally: bool = False,
+        **kwargs: Any,
+    ) -> Union[str, Generator[str, None, None]]:
+        def for_stream_chat():
+            gen = self.ask(
+                prompt,
+                stream=True,
+                raw=False,
+                optimizer=optimizer,
+                conversationally=conversationally,
+            )
+            for response_dict in gen:
+                yield self.get_message(response_dict)
+
+        def for_non_stream_chat():
+            response_data = self.ask(
+                prompt,
+                stream=False,
+                raw=False,
+                optimizer=optimizer,
+                conversationally=conversationally,
+            )
+            return self.get_message(response_data)
+
+        return for_stream_chat() if stream else for_non_stream_chat()
+
+    def get_message(self, response: Response) -> str:
+        if not isinstance(response, dict):
+            return str(response)
+        return cast(Dict[str, Any], response).get("text", "")