webscout 8.2.3__py3-none-any.whl → 8.2.5__py3-none-any.whl

webscout/Provider/typegpt.py

@@ -1,11 +1,11 @@
-import requests
+from curl_cffi.requests import Session
+from curl_cffi import CurlError
 import json
 from typing import Union, Any, Dict, Generator
-import requests.exceptions
 
 from webscout.AIutel import Optimizers
 from webscout.AIutel import Conversation
-from webscout.AIutel import AwesomePrompts
+from webscout.AIutel import AwesomePrompts, sanitize_stream # Import sanitize_stream
 from webscout.AIbase import Provider
 from webscout import exceptions
 from webscout.litagent import LitAgent
@@ -35,7 +35,7 @@ class TypeGPT(Provider):
         proxies: dict = {},
         history_offset: int = 10250,
         act: str = None,
-        model: str = "gpt-4o",
+        model: str = "gpt-4o-mini-2024-07-18",
         system_prompt: str = "You are a helpful assistant.",
         temperature: float = 0.5,
         presence_penalty: int = 0,
@@ -46,7 +46,8 @@ class TypeGPT(Provider):
         if model not in self.AVAILABLE_MODELS:
             raise ValueError(f"Invalid model: {model}. Choose from: {', '.join(self.AVAILABLE_MODELS)}")
 
-        self.session = requests.Session()
+        # Initialize curl_cffi Session
+        self.session = Session()
         self.is_conversation = is_conversation
         self.max_tokens_to_sample = max_tokens
         self.api_endpoint = "https://chat.typegpt.net/api/openai/v1/chat/completions"
@@ -82,6 +83,8 @@ class TypeGPT(Provider):
         )
         self.conversation = Conversation(is_conversation, self.max_tokens_to_sample, filepath, update_file)
         self.conversation.history_offset = history_offset
+        # Update curl_cffi session headers and proxies
+        self.session.headers.update(self.headers)
         self.session.proxies = proxies
 
     def ask(
@@ -120,57 +123,87 @@ class TypeGPT(Provider):
 
         def for_stream():
             try:
+                # Use curl_cffi session post with impersonate
                 response = self.session.post(
-                    self.api_endpoint, headers=self.headers, json=payload, stream=True, timeout=self.timeout
+                    self.api_endpoint, 
+                    headers=self.headers, 
+                    json=payload, 
+                    stream=True, 
+                    timeout=self.timeout,
+                    impersonate="chrome120"
                 )
-            except requests.exceptions.ConnectionError as ce:
+            except CurlError as ce:
                 raise exceptions.FailedToGenerateResponseError(
-                    f"Network connection failed. Check your firewall or antivirus settings. Original error: {ce}"
+                    f"Network connection failed (CurlError). Check your firewall or antivirus settings. Original error: {ce}"
                 ) from ce
 
-            if not response.ok:
-                raise exceptions.FailedToGenerateResponseError(
-                    f"Failed to generate response - ({response.status_code}, {response.reason}) - {response.text}"
-                )
-            message_load = ""
-            for line in response.iter_lines():
-                if line:
-                    line = line.decode("utf-8")
-                    if line.startswith("data: "):
-                        line = line[6:]  # Remove "data: " prefix
-                        # Skip [DONE] message
-                        if line.strip() == "[DONE]":
-                            break
-                        try:
-                            data = json.loads(line)
-                            # Extract and yield only new content
-                            if 'choices' in data and len(data['choices']) > 0:
-                                delta = data['choices'][0].get('delta', {})
-                                if 'content' in delta:
-                                    new_content = delta['content']
-                                    message_load += new_content
-                                    # Yield only the new content
-                                    yield dict(text=new_content) if not raw else new_content
-                                    self.last_response = dict(text=message_load)
-                        except json.JSONDecodeError:
-                            continue
-            self.conversation.update_chat_history(prompt, self.get_message(self.last_response))
+            response.raise_for_status() # Check for HTTP errors first
+
+            streaming_text = ""
+            # Use sanitize_stream
+            processed_stream = sanitize_stream(
+                data=response.iter_content(chunk_size=None), # Pass byte iterator
+                intro_value="data:",
+                to_json=True,     # Stream sends JSON
+                skip_markers=["[DONE]"],
+                content_extractor=lambda chunk: chunk.get('choices', [{}])[0].get('delta', {}).get('content') if isinstance(chunk, dict) else None,
+                yield_raw_on_error=False # Skip non-JSON or lines where extractor fails
+            )
+
+            for content_chunk in processed_stream:
+                # content_chunk is the string extracted by the content_extractor
+                if content_chunk and isinstance(content_chunk, str):
+                    streaming_text += content_chunk
+                    yield dict(text=content_chunk) if not raw else content_chunk
+                    # Update last_response incrementally
+                    self.last_response = dict(text=streaming_text)
+
+            # Update conversation history after stream finishes
+            if streaming_text: # Only update if something was received
+                 self.conversation.update_chat_history(prompt, streaming_text)
+
 
         def for_non_stream():
             try:
-                response = self.session.post(self.api_endpoint, headers=self.headers, json=payload, timeout=self.timeout)
-            except requests.exceptions.ConnectionError as ce:
+                # Use curl_cffi session post with impersonate
+                response = self.session.post(
+                    self.api_endpoint, 
+                    headers=self.headers, 
+                    json=payload, 
+                    timeout=self.timeout,
+                    impersonate="chrome120"
+                )
+            except CurlError as ce:
                 raise exceptions.FailedToGenerateResponseError(
-                    f"Network connection failed. Check your firewall or antivirus settings. Original error: {ce}"
+                    f"Network connection failed (CurlError). Check your firewall or antivirus settings. Original error: {ce}"
                 ) from ce
 
-            if not response.ok:
-                raise exceptions.FailedToGenerateResponseError(
-                    f"Request failed - {response.status_code}: {response.text}"
+            response.raise_for_status() # Check for HTTP errors
+
+            try:
+                response_text = response.text # Get raw text
+
+                # Use sanitize_stream for non-streaming JSON response
+                processed_stream = sanitize_stream(
+                    data=response_text,
+                    to_json=True, # Parse the whole text as JSON
+                    intro_value=None,
+                    # Extractor for non-stream structure
+                    content_extractor=lambda chunk: chunk.get('choices', [{}])[0].get('message', {}).get('content') if isinstance(chunk, dict) else None,
+                    yield_raw_on_error=False
                 )
-            self.last_response = response.json()
-            self.conversation.update_chat_history(prompt, self.get_message(self.last_response))
-            return self.last_response
+
+                # Extract the single result
+                content = ""
+                for extracted_content in processed_stream:
+                    content = extracted_content if isinstance(extracted_content, str) else ""
+
+                self.last_response = {"text": content} # Store in expected format
+                self.conversation.update_chat_history(prompt, content)
+                return self.last_response
+            except (json.JSONDecodeError, Exception) as je: # Catch potential JSON errors or others
+                 raise exceptions.FailedToGenerateResponseError(f"Failed to decode JSON response: {je} - Response text: {response.text}")
+
 
         return for_stream() if stream else for_non_stream()
 
@@ -183,23 +216,36 @@ class TypeGPT(Provider):
     ) -> Union[str, Generator[str, None, None]]:
         """Generate response string or stream."""
         if stream:
+            # ask() yields dicts or strings when streaming
             gen = self.ask(
-                prompt, stream=True, optimizer=optimizer, conversationally=conversationally
+                prompt, stream=True, raw=False, # Ensure ask yields dicts
+                optimizer=optimizer, conversationally=conversationally
             )
-            for chunk in gen:
-                yield self.get_message(chunk)  # Extract text from streamed chunks
+            for chunk_dict in gen:
+                 # get_message expects a dict
+                yield self.get_message(chunk_dict) 
         else:
-            return self.get_message(self.ask(prompt, stream=False, optimizer=optimizer, conversationally=conversationally))
+            # ask() returns a dict when not streaming
+            response_dict = self.ask(
+                prompt, stream=False, 
+                optimizer=optimizer, conversationally=conversationally
+            )
+            return self.get_message(response_dict)
 
     def get_message(self, response: Dict[str, Any]) -> str:
         """Retrieves message from response."""
-        if isinstance(response, str):  # Handle raw responses
-            return response
-        elif isinstance(response, dict):
+        if isinstance(response, dict):
             assert isinstance(response, dict), "Response should be of dict data-type only"
-            return response.get("text", "")  # Extract text from dictionary response
+            # Handle potential unicode escapes in the final text
+            text = response.get("text", "")
+            try:
+                # Attempt to decode escapes, return original if fails
+                return text.encode('utf-8').decode('unicode_escape')
+            except UnicodeDecodeError:
+                return text 
         else:
-            raise TypeError("Invalid response type. Expected str or dict.")
+            # This case should ideally not be reached if ask() behaves as expected
+            raise TypeError(f"Invalid response type: {type(response)}. Expected dict.")
 
 if __name__ == "__main__":
     print("-" * 80)
@@ -213,20 +259,31 @@ if __name__ == "__main__":
     for model in TypeGPT.AVAILABLE_MODELS:
         try:
             test_ai = TypeGPT(model=model, timeout=60)
-            response = test_ai.chat("Say 'Hello' in one word", stream=True)
+            # Test stream first
+            response_stream = test_ai.chat("Say 'Hello' in one word", stream=True)
             response_text = ""
-            for chunk in response:
+            print(f"\r{model:<50} {'Streaming...':<10}", end="", flush=True)
+            for chunk in response_stream:
                 response_text += chunk
-                print(f"\r{model:<50} {'Testing...':<10}", end="", flush=True)
+                # Optional: print chunks as they arrive for visual feedback
+                # print(chunk, end="", flush=True) 
             
             if response_text and len(response_text.strip()) > 0:
                 status = "✓"
-                # Truncate response if too long
-                display_text = response_text.strip()[:50] + "..." if len(response_text.strip()) > 50 else response_text.strip()
+                # Clean and truncate response
+                clean_text = response_text.strip() # Already decoded in get_message
+                display_text = clean_text[:50] + "..." if len(clean_text) > 50 else clean_text
             else:
-                status = "✗"
-                display_text = "Empty or invalid response"
+                status = "✗ (Stream)"
+                display_text = "Empty or invalid stream response"
             print(f"\r{model:<50} {status:<10} {display_text}")
+            
+            # Optional: Add non-stream test if needed, but stream test covers basic functionality
+            # print(f"\r{model:<50} {'Non-Stream...':<10}", end="", flush=True)
+            # response_non_stream = test_ai.chat("Say 'Hi' again", stream=False)
+            # if not response_non_stream or len(response_non_stream.strip()) == 0:
+            #      print(f"\r{model:<50} {'✗ (Non-Stream)':<10} Empty non-stream response")
+
+
         except Exception as e:
             print(f"\r{model:<50} {'✗':<10} {str(e)}")
-    

webscout/Provider/uncovr.py

@@ -1,10 +1,11 @@
-import requests
+from curl_cffi.requests import Session
+from curl_cffi import CurlError
 import json
 import uuid
 import re
 from typing import Any, Dict, Optional, Generator, Union
 from webscout.AIutel import Optimizers
-from webscout.AIutel import Conversation
+from webscout.AIutel import Conversation, sanitize_stream # Import sanitize_stream
 from webscout.AIutel import AwesomePrompts
 from webscout.AIbase import Provider
 from webscout import exceptions
@@ -76,7 +77,9 @@ class UncovrAI(Provider):
             "Sec-Fetch-Site": "same-origin"
         }
         
-        self.session = requests.Session()
+        # Initialize curl_cffi Session
+        self.session = Session()
+        # Update curl_cffi session headers and proxies
         self.session.headers.update(self.headers)
         self.session.proxies.update(proxies)
 
@@ -106,6 +109,17 @@ class UncovrAI(Provider):
         )
         self.conversation.history_offset = history_offset
 
+    @staticmethod
+    def _uncovr_extractor(chunk: Union[str, Dict[str, Any]]) -> Optional[str]:
+        """Extracts content from the UncovrAI stream format '0:"..."'."""
+        if isinstance(chunk, str):
+            match = re.match(r'^0:\s*"?(.*?)"?$', chunk) # Match 0: maybe optional quotes
+            if match:
+                # Decode potential unicode escapes like \u00e9 and handle escaped quotes/backslashes
+                content = match.group(1).encode().decode('unicode_escape')
+                return content.replace('\\\\', '\\').replace('\\"', '"')
+        return None
+
     def refresh_identity(self, browser: str = None):
         """
         Refreshes the browser identity fingerprint.
@@ -169,87 +183,118 @@ class UncovrAI(Provider):
 
         def for_stream():
             try:
-                with self.session.post(self.url, json=payload, stream=True, timeout=self.timeout) as response:
-                    if response.status_code != 200:
-                        # If we get a non-200 response, try refreshing our identity once
-                        if response.status_code in [403, 429]:
-                            self.refresh_identity()
-                            # Retry with new identity
-                            with self.session.post(self.url, json=payload, stream=True, timeout=self.timeout) as retry_response:
-                                if not retry_response.ok:
-                                    raise exceptions.FailedToGenerateResponseError(
-                                        f"Failed to generate response after identity refresh - ({retry_response.status_code}, {retry_response.reason}) - {retry_response.text}"
-                                    )
-                                response = retry_response
-                        else:
+                # Use curl_cffi session post with impersonate
+                response = self.session.post(
+                    self.url, 
+                    json=payload, 
+                    stream=True, 
+                    timeout=self.timeout,
+                    impersonate=self.fingerprint.get("browser_type", "chrome110") # Use fingerprint browser type
+                )
+                
+                if response.status_code != 200:
+                    # If we get a non-200 response, try refreshing our identity once
+                    if response.status_code in [403, 429]:
+                        self.refresh_identity()
+                        # Retry with new identity using curl_cffi session
+                        retry_response = self.session.post(
+                            self.url, 
+                            json=payload, 
+                            stream=True, 
+                            timeout=self.timeout,
+                            impersonate=self.fingerprint.get("browser_type", "chrome110") # Use updated fingerprint
+                        )
+                        if not retry_response.ok:
                             raise exceptions.FailedToGenerateResponseError(
-                                f"Request failed with status code {response.status_code}"
+                                f"Failed to generate response after identity refresh - ({retry_response.status_code}, {retry_response.reason}) - {retry_response.text}"
                             )
+                        response = retry_response # Use the successful retry response
+                    else:
+                        raise exceptions.FailedToGenerateResponseError(
+                            f"Request failed with status code {response.status_code} - {response.text}"
+                        )
+
+                streaming_text = ""
+                # Use sanitize_stream with the custom extractor
+                processed_stream = sanitize_stream(
+                    data=response.iter_content(chunk_size=None), # Pass byte iterator
+                    intro_value=None, # No simple prefix
+                    to_json=False,    # Content is not JSON
+                    content_extractor=self._uncovr_extractor, # Use the specific extractor
+                    yield_raw_on_error=True # Keep yielding even if extractor fails, for potential error messages? (Adjust if needed)
+                )
+
+                for content_chunk in processed_stream:
+                    if content_chunk and isinstance(content_chunk, str):
+                        streaming_text += content_chunk
+                        yield dict(text=content_chunk) if not raw else content_chunk
+                
+                self.last_response = {"text": streaming_text}
+                self.conversation.update_chat_history(prompt, streaming_text)
                     
-                    streaming_text = ""
-                    for line in response.iter_lines():
-                        if line:
-                            try:
-                                line = line.decode('utf-8')
-                                # Use regex to match content messages
-                                content_match = re.match(r'^0:\s*"?(.*?)"?$', line)
-                                if content_match:  # Content message
-                                    content = content_match.group(1)
-                                    streaming_text += content
-                                    resp = dict(text=content)
-                                    yield resp if raw else resp
-                                # Check for error messages
-                                error_match = re.match(r'^2:\[{"type":"error","error":"(.*?)"}]$', line)
-                                if error_match:
-                                    error_msg = error_match.group(1)
-                                    raise exceptions.FailedToGenerateResponseError(f"API Error: {error_msg}")
-                            except (json.JSONDecodeError, UnicodeDecodeError):
-                                continue
-                    
-                    self.last_response = {"text": streaming_text}
-                    self.conversation.update_chat_history(prompt, streaming_text)
-                    
-            except requests.RequestException as e:
-                raise exceptions.FailedToGenerateResponseError(f"Request failed: {str(e)}")
+            except CurlError as e: # Catch CurlError
+                raise exceptions.FailedToGenerateResponseError(f"Request failed (CurlError): {e}")
+            except Exception as e: # Catch other potential exceptions
+                raise exceptions.FailedToGenerateResponseError(f"An unexpected error occurred ({type(e).__name__}): {e}")
+
 
         def for_non_stream():
             try:
-                response = self.session.post(self.url, json=payload, timeout=self.timeout)
+                # Use curl_cffi session post with impersonate
+                response = self.session.post(
+                    self.url, 
+                    json=payload, 
+                    timeout=self.timeout,
+                    impersonate=self.fingerprint.get("browser_type", "chrome110")
+                )
+                
                 if response.status_code != 200:
                     if response.status_code in [403, 429]:
                         self.refresh_identity()
-                        response = self.session.post(self.url, json=payload, timeout=self.timeout)
+                        # Retry with new identity using curl_cffi session
+                        response = self.session.post(
+                            self.url, 
+                            json=payload, 
+                            timeout=self.timeout,
+                            impersonate=self.fingerprint.get("browser_type", "chrome110")
+                        )
                         if not response.ok:
                             raise exceptions.FailedToGenerateResponseError(
                                 f"Failed to generate response after identity refresh - ({response.status_code}, {response.reason}) - {response.text}"
                             )
                     else:
                         raise exceptions.FailedToGenerateResponseError(
-                            f"Request failed with status code {response.status_code}"
+                            f"Request failed with status code {response.status_code} - {response.text}"
                         )
 
+                response_text = response.text # Get the full response text
+                
+                # Use sanitize_stream to process the non-streaming text
+                # It won't parse as JSON, but will apply the extractor line by line
+                processed_stream = sanitize_stream(
+                    data=response_text.splitlines(), # Split into lines first
+                    intro_value=None,
+                    to_json=False,
+                    content_extractor=self._uncovr_extractor,
+                    yield_raw_on_error=True 
+                )
+
+                # Aggregate the results from the generator
                 full_response = ""
-                for line in response.iter_lines():
-                    if line:
-                        try:
-                            line = line.decode('utf-8')
-                            content_match = re.match(r'^0:\s*"?(.*?)"?$', line)
-                            if content_match:
-                                content = content_match.group(1)
-                                full_response += content
-                            # Check for error messages
-                            error_match = re.match(r'^2:\[{"type":"error","error":"(.*?)"}]$', line)
-                            if error_match:
-                                error_msg = error_match.group(1)
-                                raise exceptions.FailedToGenerateResponseError(f"API Error: {error_msg}")
-                        except (json.JSONDecodeError, UnicodeDecodeError):
-                            continue
+                for content in processed_stream:
+                    if content and isinstance(content, str):
+                        full_response += content
 
+                # Check if aggregation resulted in empty response (might indicate error not caught by extractor)
                 self.last_response = {"text": full_response}
                 self.conversation.update_chat_history(prompt, full_response)
                 return {"text": full_response}
-            except Exception as e:
-                raise exceptions.FailedToGenerateResponseError(f"Request failed: {e}")
+                
+            except CurlError as e: # Catch CurlError
+                raise exceptions.FailedToGenerateResponseError(f"Request failed (CurlError): {e}")
+            except Exception as e: # Catch other potential exceptions
+                raise exceptions.FailedToGenerateResponseError(f"Request failed ({type(e).__name__}): {e}")
+
 
         return for_stream() if stream else for_non_stream()
 
@@ -283,9 +328,12 @@ class UncovrAI(Provider):
 
     def get_message(self, response: dict) -> str:
         assert isinstance(response, dict), "Response should be of dict data-type only"
-        return response["text"].replace('\\n', '\n').replace('\\n\\n', '\n\n')
+        # Formatting handled by extractor
+        text = response.get("text", "")
+        return text.replace('\\n', '\n').replace('\\n\\n', '\n\n') # Keep newline replacement
 
 if __name__ == "__main__":
+    # Ensure curl_cffi is installed
     print("-" * 80)
     print(f"{'Model':<50} {'Status':<10} {'Response'}")
     print("-" * 80)
@@ -293,20 +341,28 @@ if __name__ == "__main__":
     for model in UncovrAI.AVAILABLE_MODELS:
         try:
             test_ai = UncovrAI(model=model, timeout=60)
-            response = test_ai.chat("Say 'Hello' in one word", stream=True)
-            response_text = ""
-            for chunk in response:
-                response_text += chunk
+            # Test non-stream first as stream logic depends on it
+            response_non_stream = test_ai.chat("Say 'Hello' in one word", stream=False) 
             
-            if response_text and len(response_text.strip()) > 0:
-                status = "✓"
-                # Clean and truncate response
-                clean_text = response_text.strip().encode('utf-8', errors='ignore').decode('utf-8')
-                display_text = clean_text[:50] + "..." if len(clean_text) > 50 else clean_text
+            if response_non_stream and len(response_non_stream.strip()) > 0:
+                 # Now test stream
+                response_stream = test_ai.chat("Say 'Hi' in one word", stream=True)
+                response_text = ""
+                for chunk in response_stream:
+                    response_text += chunk
+                
+                if response_text and len(response_text.strip()) > 0:
+                    status = "✓"
+                    # Clean and truncate response
+                    clean_text = response_text.strip().encode('utf-8', errors='ignore').decode('utf-8')
+                    display_text = clean_text[:50] + "..." if len(clean_text) > 50 else clean_text
+                else:
+                    status = "✗ (Stream)"
+                    display_text = "Empty or invalid stream response"
             else:
-                status = "✗"
-                display_text = "Empty or invalid response"
+                status = "✗ (Non-Stream)"
+                display_text = "Empty or invalid non-stream response"
+                
             print(f"\r{model:<50} {status:<10} {display_text}")
         except Exception as e:
-            print(f"\r{model:<50} {'✗':<10} {str(e)}") 
-
+            print(f"\r{model:<50} {'✗':<10} {str(e)}")