webscout 8.2.8__py3-none-any.whl → 8.3__py3-none-any.whl

webscout/Provider/AISEARCH/Perplexity.py

@@ -1,359 +1,333 @@
-import json
-import random
-from uuid import uuid4
-from typing import Dict, Optional, Generator, Union, Any
-from curl_cffi import requests
-
-from webscout.AIbase import AISearch
-from webscout import exceptions
-from webscout.litagent import LitAgent
-
-
-class Response:
-    """A wrapper class for Perplexity API responses.
-    
-    This class automatically converts response objects to their text representation
-    when printed or converted to string.
-    
-    Attributes:
-        text (str): The text content of the response
-        
-    Example:
-        >>> response = Response("Hello, world!")
-        >>> print(response)
-        Hello, world!
-        >>> str(response)
-        'Hello, world!'
-    """
-    def __init__(self, text: str):
-        self.text = text
-    
-    def __str__(self):
-        return self.text
-    
-    def __repr__(self):
-        return self.text
-
-
-class Perplexity(AISearch):
-    """A class to interact with the Perplexity AI search API.
-    
-    Perplexity provides a powerful search interface that returns AI-generated responses
-    based on web content. It supports both streaming and non-streaming responses,
-    multiple search modes, and model selection.
-    
-    Basic Usage:
-        >>> from webscout import Perplexity
-        >>> ai = Perplexity()
-        >>> # Non-streaming example
-        >>> response = ai.search("What is Python?")
-        >>> print(response)
-        Python is a high-level programming language...
-        
-        >>> # Streaming example
-        >>> for chunk in ai.search("Tell me about AI", stream=True):
-        ...     print(chunk, end="", flush=True)
-        Artificial Intelligence is...
-        
-        >>> # Pro search with specific model (requires authentication via cookies)
-        >>> cookies = {"perplexity-user": "your_cookie_value"}
-        >>> ai_pro = Perplexity(cookies=cookies)
-        >>> response = ai_pro.search("Latest AI research", mode="pro", model="gpt-4o")
-        >>> print(response)
-        
-        >>> # Raw response format
-        >>> for chunk in ai.search("Hello", stream=True, raw=True):
-        ...     print(chunk)
-        {'text': 'Hello'}
-        {'text': ' there!'}
-    
-    Args:
-        cookies (dict, optional): Cookies to use for authentication. Defaults to None.
-        timeout (int, optional): Request timeout in seconds. Defaults to 60.
-        proxies (dict, optional): Proxy configuration for requests. Defaults to None.
-    """
-    
-    def __init__(
-        self,
-        cookies: Optional[Dict[str, str]] = None,
-        timeout: int = 60,
-        proxies: Optional[Dict[str, str]] = None
-    ):
-        """
-        Initialize the Perplexity client.
-        
-        Args:
-            cookies (dict, optional): Cookies to use for authentication. Defaults to None.
-            timeout (int, optional): Request timeout in seconds. Defaults to 60.
-            proxies (dict, optional): Proxy configuration for requests. Defaults to None.
-        """
-        self.timeout = timeout
-        self.agent = LitAgent()
-        self.session = requests.Session(headers={
-            'accept': 'text/html,application/xhtml+xml,application/xml;q=0.9,image/avif,image/webp,image/apng,*/*;q=0.8,application/signed-exchange;v=b3;q=0.7',
-            'accept-language': 'en-US,en;q=0.9',
-            'cache-control': 'max-age=0',
-            'dnt': '1',
-            'priority': 'u=0, i',
-            'sec-ch-ua': '"Not;A=Brand";v="24", "Chromium";v="128"',
-            'sec-ch-ua-arch': '"x86"',
-            'sec-ch-ua-bitness': '"64"',
-            'sec-ch-ua-full-version': '"128.0.6613.120"',
-            'sec-ch-ua-full-version-list': '"Not;A=Brand";v="24.0.0.0", "Chromium";v="128.0.6613.120"',
-            'sec-ch-ua-mobile': '?0',
-            'sec-ch-ua-model': '""',
-            'sec-ch-ua-platform': '"Windows"',
-            'sec-ch-ua-platform-version': '"19.0.0"',
-            'sec-fetch-dest': 'document',
-            'sec-fetch-mode': 'navigate',
-            'sec-fetch-site': 'same-origin',
-            'sec-fetch-user': '?1',
-            'upgrade-insecure-requests': '1',
-            'user-agent': self.agent.random(),
-        }, cookies=cookies or {}, impersonate='chrome')
-        
-        # Apply proxies if provided
-        if proxies:
-            self.session.proxies.update(proxies)
-        
-        # Initialize session with socket.io
-        self.timestamp = format(random.getrandbits(32), '08x')
-        
-        # Get socket.io session ID
-        response = self.session.get(f'https://www.perplexity.ai/socket.io/?EIO=4&transport=polling&t={self.timestamp}')
-        self.sid = json.loads(response.text[1:])['sid']
-        
-        # Initialize socket.io connection
-        assert (self.session.post(
-            f'https://www.perplexity.ai/socket.io/?EIO=4&transport=polling&t={self.timestamp}&sid={self.sid}',
-            data='40{"jwt":"anonymous-ask-user"}'
-        )).text == 'OK'
-        
-        # Get session info
-        self.session.get('https://www.perplexity.ai/api/auth/session')
-        
-        # Set default values
-        self.copilot = 0 if not cookies else float('inf')
-        self.file_upload = 0 if not cookies else float('inf')
-    
-    def _extract_answer(self, response):
-        """
-        Extract the answer from the response.
-        
-        Args:
-            response (dict): The response from Perplexity AI.
-            
-        Returns:
-            str: The extracted answer text.
-        """
-        if not response:
-            return ""
-        
-        # Find the FINAL step in the text array
-        final_step = None
-        if 'text' in response and isinstance(response['text'], list):
-            for step in response['text']:
-                if step.get('step_type') == 'FINAL' and 'content' in step and 'answer' in step['content']:
-                    final_step = step
-                    break
-        
-        if not final_step:
-            return ""
-        
-        try:
-            # Parse the answer JSON string
-            answer_json = json.loads(final_step['content']['answer'])
-            return answer_json.get('answer', '')
-        except (json.JSONDecodeError, KeyError):
-            return ""
-    
-    def search(
-        self,
-        prompt: str,
-        mode: str = 'auto',
-        model: Optional[str] = None,
-        sources: Optional[list] = None,
-        stream: bool = False,
-        raw: bool = False,
-        language: str = 'en-US',
-        follow_up: Optional[Dict[str, Any]] = None,
-        incognito: bool = False
-    ) -> Union[Response, Generator[Union[Dict[str, str], Response], None, None]]:
-        """Search using the Perplexity API and get AI-generated responses.
-        
-        This method sends a search query to Perplexity and returns the AI-generated response.
-        It supports both streaming and non-streaming modes, as well as raw response format.
-        
-        Args:
-            prompt (str): The search query or prompt to send to the API.
-            mode (str, optional): Search mode. Options: 'auto', 'pro', 'reasoning', 'deep research'. 
-                                Defaults to 'auto'.
-            model (str, optional): Model to use. Available models depend on the mode. Defaults to None.
-            sources (list, optional): Sources to use. Options: 'web', 'scholar', 'social'. 
-                                    Defaults to ['web'].
-            stream (bool, optional): If True, yields response chunks as they arrive.
-                                   If False, returns complete response. Defaults to False.
-            raw (bool, optional): If True, returns raw response dictionaries.
-                                If False, returns Response objects that convert to text automatically.
-                                Defaults to False.
-            language (str, optional): Language to use. Defaults to 'en-US'.
-            follow_up (dict, optional): Follow-up information. Defaults to None.
-            incognito (bool, optional): Whether to use incognito mode. Defaults to False.
-            
-        Returns:
-            If stream=True: Generator yielding response chunks as they arrive
-            If stream=False: Complete response
-            
-        Raises:
-            ValueError: If invalid mode or model is provided
-            exceptions.APIConnectionError: If connection to API fails
-            exceptions.FailedToGenerateResponseError: If response generation fails
-        """
-        if sources is None:
-            sources = ['web']
-        
-        # Validate inputs
-        if mode not in ['auto', 'pro', 'reasoning', 'deep research']:
-            raise ValueError('Search modes -> ["auto", "pro", "reasoning", "deep research"]')
-        
-        if not all([source in ('web', 'scholar', 'social') for source in sources]):
-            raise ValueError('Sources -> ["web", "scholar", "social"]')
-        
-        # Check if model is valid for the selected mode
-        valid_models = {
-            'auto': [None],
-            'pro': [None, 'sonar', 'gpt-4.5', 'gpt-4o', 'claude 3.7 sonnet', 'gemini 2.0 flash', 'grok-2'],
-            'reasoning': [None, 'r1', 'o3-mini', 'claude 3.7 sonnet'],
-            'deep research': [None]
-        }
-        
-        if mode in valid_models and model not in valid_models[mode] and model is not None:
-            raise ValueError(f'Invalid model for {mode} mode. Valid models: {valid_models[mode]}')
-        
-        # Prepare request data
-        json_data = {
-            'query_str': prompt,
-            'params': {
-                'attachments': follow_up['attachments'] if follow_up else [],
-                'frontend_context_uuid': str(uuid4()),
-                'frontend_uuid': str(uuid4()),
-                'is_incognito': incognito,
-                'language': language,
-                'last_backend_uuid': follow_up['backend_uuid'] if follow_up else None,
-                'mode': 'concise' if mode == 'auto' else 'copilot',
-                'model_preference': {
-                    'auto': {
-                        None: 'turbo'
-                    },
-                    'pro': {
-                        None: 'pplx_pro',
-                        'sonar': 'experimental',
-                        'gpt-4.5': 'gpt45',
-                        'gpt-4o': 'gpt4o',
-                        'claude 3.7 sonnet': 'claude2',
-                        'gemini 2.0 flash': 'gemini2flash',
-                        'grok-2': 'grok'
-                    },
-                    'reasoning': {
-                        None: 'pplx_reasoning',
-                        'r1': 'r1',
-                        'o3-mini': 'o3mini',
-                        'claude 3.7 sonnet': 'claude37sonnetthinking'
-                    },
-                    'deep research': {
-                        None: 'pplx_alpha'
-                    }
-                }[mode][model],
-                'source': 'default',
-                'sources': sources,
-                'version': '2.18'
-            }
-        }
-        
-        try:
-            # Make the request
-            resp = self.session.post(
-                'https://www.perplexity.ai/rest/sse/perplexity_ask', 
-                json=json_data, 
-                stream=True,
-                timeout=self.timeout
-            )
-            
-            if resp.status_code != 200:
-                raise exceptions.APIConnectionError(f"API returned status code {resp.status_code}")
-            
-            # Define streaming response handler
-            def stream_response():
-                for chunk in resp.iter_lines(delimiter=b'\r\n\r\n'):
-                    content = chunk.decode('utf-8')
-                    if content.startswith('event: message\r\n'):
-                        content_json = json.loads(content[len('event: message\r\ndata: '):])
-                        if 'text' in content_json:
-                            try:
-                                # If text is a string, try to parse it as JSON
-                                if isinstance(content_json['text'], str):
-                                    content_json['text'] = json.loads(content_json['text'])
-                            except json.JSONDecodeError:
-                                pass
-                        
-                        if raw:
-                            yield content_json
-                        else:
-                            # For non-raw responses, extract text from each chunk
-                            if 'text' in content_json and isinstance(content_json['text'], list):
-                                for step in content_json['text']:
-                                    if step.get('type') == 'answer' and 'value' in step:
-                                        yield Response(step['value'])
-                                    elif step.get('type') == 'thinking' and 'value' in step:
-                                        yield Response(step['value'])
-                    elif content.startswith('event: end_of_stream\r\n'):
-                        return
-            
-            # Handle streaming or non-streaming response
-            if stream:
-                return stream_response()
-            else:
-                chunks = []
-                final_response = None
-                
-                for chunk in resp.iter_lines(delimiter=b'\r\n\r\n'):
-                    content = chunk.decode('utf-8')
-                    if content.startswith('event: message\r\n'):
-                        content_json = json.loads(content[len('event: message\r\ndata: '):])
-                        if 'text' in content_json:
-                            try:
-                                # If text is a string, try to parse it as JSON
-                                if isinstance(content_json['text'], str):
-                                    content_json['text'] = json.loads(content_json['text'])
-                            except json.JSONDecodeError:
-                                pass
-                        chunks.append(content_json)
-                        final_response = content_json
-                    elif content.startswith('event: end_of_stream\r\n'):
-                        # Process the final response to extract the answer
-                        if final_response:
-                            answer_text = self._extract_answer(final_response)
-                            return Response(answer_text) if not raw else final_response
-                        elif chunks:
-                            answer_text = self._extract_answer(chunks[-1])
-                            return Response(answer_text) if not raw else chunks[-1]
-                        else:
-                            return Response("") if not raw else {}
-                
-                # If we get here, something went wrong
-                raise exceptions.FailedToGenerateResponseError("Failed to get complete response")
-                
-        except requests.RequestsError as e:
-            raise exceptions.APIConnectionError(f"Connection error: {str(e)}")
-        except json.JSONDecodeError:
-            raise exceptions.FailedToGenerateResponseError("Failed to parse response JSON")
-        except Exception as e:
-            raise exceptions.FailedToGenerateResponseError(f"Error: {str(e)}")
-
-
-if __name__ == "__main__":
-    # Simple test
-    ai = Perplexity()
-    response = ai.search("What is Python?")
-    print(response)
+import json
+import random
+from uuid import uuid4
+from typing import Dict, Optional, Generator, Union, Any
+from curl_cffi import requests
+
+from webscout.AIbase import AISearch, SearchResponse
+from webscout import exceptions
+from webscout.litagent import LitAgent
+
+
+class Perplexity(AISearch):
+    """A class to interact with the Perplexity AI search API.
+    
+    Perplexity provides a powerful search interface that returns AI-generated responses
+    based on web content. It supports both streaming and non-streaming responses,
+    multiple search modes, and model selection.
+    
+    Basic Usage:
+        >>> from webscout import Perplexity
+        >>> ai = Perplexity()
+        >>> # Non-streaming example
+        >>> response = ai.search("What is Python?")
+        >>> print(response)
+        Python is a high-level programming language...
+        
+        >>> # Streaming example
+        >>> for chunk in ai.search("Tell me about AI", stream=True):
+        ...     print(chunk, end="", flush=True)
+        Artificial Intelligence is...
+        
+        >>> # Pro search with specific model (requires authentication via cookies)
+        >>> cookies = {"perplexity-user": "your_cookie_value"}
+        >>> ai_pro = Perplexity(cookies=cookies)
+        >>> response = ai_pro.search("Latest AI research", mode="pro", model="gpt-4o")
+        >>> print(response)
+        
+        >>> # Raw response format
+        >>> for chunk in ai.search("Hello", stream=True, raw=True):
+        ...     print(chunk)
+        {'text': 'Hello'}
+        {'text': ' there!'}
+    
+    Args:
+        cookies (dict, optional): Cookies to use for authentication. Defaults to None.
+        timeout (int, optional): Request timeout in seconds. Defaults to 60.
+        proxies (dict, optional): Proxy configuration for requests. Defaults to None.
+    """
+    
+    def __init__(
+        self,
+        cookies: Optional[Dict[str, str]] = None,
+        timeout: int = 60,
+        proxies: Optional[Dict[str, str]] = None
+    ):
+        """
+        Initialize the Perplexity client.
+        
+        Args:
+            cookies (dict, optional): Cookies to use for authentication. Defaults to None.
+            timeout (int, optional): Request timeout in seconds. Defaults to 60.
+            proxies (dict, optional): Proxy configuration for requests. Defaults to None.
+        """
+        self.timeout = timeout
+        self.agent = LitAgent()
+        self.session = requests.Session(headers={
+            'accept': 'text/html,application/xhtml+xml,application/xml;q=0.9,image/avif,image/webp,image/apng,*/*;q=0.8,application/signed-exchange;v=b3;q=0.7',
+            'accept-language': 'en-US,en;q=0.9',
+            'cache-control': 'max-age=0',
+            'dnt': '1',
+            'priority': 'u=0, i',
+            'sec-ch-ua': '"Not;A=Brand";v="24", "Chromium";v="128"',
+            'sec-ch-ua-arch': '"x86"',
+            'sec-ch-ua-bitness': '"64"',
+            'sec-ch-ua-full-version': '"128.0.6613.120"',
+            'sec-ch-ua-full-version-list': '"Not;A=Brand";v="24.0.0.0", "Chromium";v="128.0.6613.120"',
+            'sec-ch-ua-mobile': '?0',
+            'sec-ch-ua-model': '""',
+            'sec-ch-ua-platform': '"Windows"',
+            'sec-ch-ua-platform-version': '"19.0.0"',
+            'sec-fetch-dest': 'document',
+            'sec-fetch-mode': 'navigate',
+            'sec-fetch-site': 'same-origin',
+            'sec-fetch-user': '?1',
+            'upgrade-insecure-requests': '1',
+            'user-agent': self.agent.random(),
+        }, cookies=cookies or {}, impersonate='chrome')
+        
+        # Apply proxies if provided
+        if proxies:
+            self.session.proxies.update(proxies)
+        
+        # Initialize session with socket.io
+        self.timestamp = format(random.getrandbits(32), '08x')
+        
+        # Get socket.io session ID
+        response = self.session.get(f'https://www.perplexity.ai/socket.io/?EIO=4&transport=polling&t={self.timestamp}')
+        self.sid = json.loads(response.text[1:])['sid']
+        
+        # Initialize socket.io connection
+        assert (self.session.post(
+            f'https://www.perplexity.ai/socket.io/?EIO=4&transport=polling&t={self.timestamp}&sid={self.sid}',
+            data='40{"jwt":"anonymous-ask-user"}'
+        )).text == 'OK'
+        
+        # Get session info
+        self.session.get('https://www.perplexity.ai/api/auth/session')
+        
+        # Set default values
+        self.copilot = 0 if not cookies else float('inf')
+        self.file_upload = 0 if not cookies else float('inf')
+    
+    def _extract_answer(self, response):
+        """
+        Extract the answer from the response.
+        
+        Args:
+            response (dict): The response from Perplexity AI.
+            
+        Returns:
+            str: The extracted answer text.
+        """
+        if not response:
+            return ""
+        
+        # Find the FINAL step in the text array
+        final_step = None
+        if 'text' in response and isinstance(response['text'], list):
+            for step in response['text']:
+                if step.get('step_type') == 'FINAL' and 'content' in step and 'answer' in step['content']:
+                    final_step = step
+                    break
+        
+        if not final_step:
+            return ""
+        
+        try:
+            # Parse the answer JSON string
+            answer_json = json.loads(final_step['content']['answer'])
+            return answer_json.get('answer', '')
+        except (json.JSONDecodeError, KeyError):
+            return ""
+    
+    def search(
+        self,
+        prompt: str,
+        mode: str = 'auto',
+        model: Optional[str] = None,
+        sources: Optional[list] = None,
+        stream: bool = False,
+        raw: bool = False,
+        language: str = 'en-US',
+        follow_up: Optional[Dict[str, Any]] = None,
+        incognito: bool = False
+    ) -> Union[SearchResponse, Generator[Union[Dict[str, str], SearchResponse], None, None]]:
+        """Search using the Perplexity API and get AI-generated responses.
+        
+        This method sends a search query to Perplexity and returns the AI-generated response.
+        It supports both streaming and non-streaming modes, as well as raw response format.
+        
+        Args:
+            prompt (str): The search query or prompt to send to the API.
+            mode (str, optional): Search mode. Options: 'auto', 'pro', 'reasoning', 'deep research'. 
+                                Defaults to 'auto'.
+            model (str, optional): Model to use. Available models depend on the mode. Defaults to None.
+            sources (list, optional): Sources to use. Options: 'web', 'scholar', 'social'. 
+                                    Defaults to ['web'].
+            stream (bool, optional): If True, yields response chunks as they arrive.
+                                   If False, returns complete response. Defaults to False.
+            raw (bool, optional): If True, returns raw response dictionaries.
+                                If False, returns Response objects that convert to text automatically.
+                                Defaults to False.
+            language (str, optional): Language to use. Defaults to 'en-US'.
+            follow_up (dict, optional): Follow-up information. Defaults to None.
+            incognito (bool, optional): Whether to use incognito mode. Defaults to False.
+            
+        Returns:
+            If stream=True: Generator yielding response chunks as they arrive
+            If stream=False: Complete response
+            
+        Raises:
+            ValueError: If invalid mode or model is provided
+            exceptions.APIConnectionError: If connection to API fails
+            exceptions.FailedToGenerateResponseError: If response generation fails
+        """
+        if sources is None:
+            sources = ['web']
+        
+        # Validate inputs
+        if mode not in ['auto', 'pro', 'reasoning', 'deep research']:
+            raise ValueError('Search modes -> ["auto", "pro", "reasoning", "deep research"]')
+        
+        if not all([source in ('web', 'scholar', 'social') for source in sources]):
+            raise ValueError('Sources -> ["web", "scholar", "social"]')
+        
+        # Check if model is valid for the selected mode
+        valid_models = {
+            'auto': [None],
+            'pro': [None, 'sonar', 'gpt-4.5', 'gpt-4o', 'claude 3.7 sonnet', 'gemini 2.0 flash', 'grok-2'],
+            'reasoning': [None, 'r1', 'o3-mini', 'claude 3.7 sonnet'],
+            'deep research': [None]
+        }
+        
+        if mode in valid_models and model not in valid_models[mode] and model is not None:
+            raise ValueError(f'Invalid model for {mode} mode. Valid models: {valid_models[mode]}')
+        
+        # Prepare request data
+        json_data = {
+            'query_str': prompt,
+            'params': {
+                'attachments': follow_up['attachments'] if follow_up else [],
+                'frontend_context_uuid': str(uuid4()),
+                'frontend_uuid': str(uuid4()),
+                'is_incognito': incognito,
+                'language': language,
+                'last_backend_uuid': follow_up['backend_uuid'] if follow_up else None,
+                'mode': 'concise' if mode == 'auto' else 'copilot',
+                'model_preference': {
+                    'auto': {
+                        None: 'turbo'
+                    },
+                    'pro': {
+                        None: 'pplx_pro',
+                        'sonar': 'experimental',
+                        'gpt-4.5': 'gpt45',
+                        'gpt-4o': 'gpt4o',
+                        'claude 3.7 sonnet': 'claude2',
+                        'gemini 2.0 flash': 'gemini2flash',
+                        'grok-2': 'grok'
+                    },
+                    'reasoning': {
+                        None: 'pplx_reasoning',
+                        'r1': 'r1',
+                        'o3-mini': 'o3mini',
+                        'claude 3.7 sonnet': 'claude37sonnetthinking'
+                    },
+                    'deep research': {
+                        None: 'pplx_alpha'
+                    }
+                }[mode][model],
+                'source': 'default',
+                'sources': sources,
+                'version': '2.18'
+            }
+        }
+        
+        try:
+            # Make the request
+            resp = self.session.post(
+                'https://www.perplexity.ai/rest/sse/perplexity_ask', 
+                json=json_data, 
+                stream=True,
+                timeout=self.timeout
+            )
+            
+            if resp.status_code != 200:
+                raise exceptions.APIConnectionError(f"API returned status code {resp.status_code}")
+            
+            # Define streaming response handler
+            def stream_response():
+                for chunk in resp.iter_lines(delimiter=b'\r\n\r\n'):
+                    content = chunk.decode('utf-8')
+                    if content.startswith('event: message\r\n'):
+                        content_json = json.loads(content[len('event: message\r\ndata: '):])
+                        if 'text' in content_json:
+                            try:
+                                # If text is a string, try to parse it as JSON
+                                if isinstance(content_json['text'], str):
+                                    content_json['text'] = json.loads(content_json['text'])
+                            except json.JSONDecodeError:
+                                pass
+                        
+                        if raw:
+                            yield content_json
+                        else:
+                            # For non-raw responses, extract text from each chunk
+                            if 'text' in content_json and isinstance(content_json['text'], list):
+                                for step in content_json['text']:
+                                    if step.get('type') == 'answer' and 'value' in step:
+                                        yield SearchResponse(step['value'])
+                                    elif step.get('type') == 'thinking' and 'value' in step:
+                                        yield SearchResponse(step['value'])
+                    elif content.startswith('event: end_of_stream\r\n'):
+                        return
+            
+            # Handle streaming or non-streaming response
+            if stream:
+                return stream_response()
+            else:
+                chunks = []
+                final_response = None
+                
+                for chunk in resp.iter_lines(delimiter=b'\r\n\r\n'):
+                    content = chunk.decode('utf-8')
+                    if content.startswith('event: message\r\n'):
+                        content_json = json.loads(content[len('event: message\r\ndata: '):])
+                        if 'text' in content_json:
+                            try:
+                                # If text is a string, try to parse it as JSON
+                                if isinstance(content_json['text'], str):
+                                    content_json['text'] = json.loads(content_json['text'])
+                            except json.JSONDecodeError:
+                                pass
+                        chunks.append(content_json)
+                        final_response = content_json
+                    elif content.startswith('event: end_of_stream\r\n'):
+                        # Process the final response to extract the answer
+                        if final_response:
+                            answer_text = self._extract_answer(final_response)
+                            return SearchResponse(answer_text) if not raw else final_response
+                        elif chunks:
+                            answer_text = self._extract_answer(chunks[-1])
+                            return SearchResponse(answer_text) if not raw else chunks[-1]
+                        else:
+                            return SearchResponse("") if not raw else {}
+                
+                # If we get here, something went wrong
+                raise exceptions.FailedToGenerateResponseError("Failed to get complete response")
+                
+        except requests.RequestsError as e:
+            raise exceptions.APIConnectionError(f"Connection error: {str(e)}")
+        except json.JSONDecodeError:
+            raise exceptions.FailedToGenerateResponseError("Failed to parse response JSON")
+        except Exception as e:
+            raise exceptions.FailedToGenerateResponseError(f"Error: {str(e)}")
+
+
+if __name__ == "__main__":
+    # Simple test
+    ai = Perplexity()
+    response = ai.search("What is Python?")
+    print(response)