webscout 8.2.4__py3-none-any.whl → 8.2.5__py3-none-any.whl

webscout/Provider/toolbaz.py

@@ -13,7 +13,7 @@ from typing import Any, Dict, Optional, Generator, Union, List
 from webscout import 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
 
 class Toolbaz(Provider):
@@ -26,6 +26,7 @@ class Toolbaz(Provider):
         "gemini-2.0-flash-thinking",
         "gemini-2.0-flash",
         "gemini-1.5-flash",
+        "o3-mini",
         "gpt-4o-latest",
         "gpt-4o",
         "deepseek-r1",
@@ -111,6 +112,13 @@ class Toolbaz(Provider):
         )
         self.conversation.history_offset = history_offset
 
+    @staticmethod
+    def _toolbaz_extractor(chunk: Union[str, Dict[str, Any]]) -> Optional[str]:
+        """Removes [model:...] tags from a string chunk."""
+        if isinstance(chunk, str):
+            return re.sub(r"\[model:.*?\]", "", chunk)
+        return None
+
     def random_string(self, length):
         return ''.join(random.choices(string.ascii_letters + string.digits, k=length))
 
@@ -207,53 +215,23 @@ class Toolbaz(Provider):
                 )
                 resp.raise_for_status()
 
-                buffer = ""
-                tag_start = "[model:"
                 streaming_text = ""
 
-                # Iterate over bytes and decode manually
-                for chunk_bytes in resp.iter_content(chunk_size=1024): # Read in larger chunks
-                    if chunk_bytes:
-                        text = chunk_bytes.decode(errors="ignore")
-                        buffer += text
-                        
-                        processed_buffer = ""
-                        last_processed_index = 0
-                        # Find all complete tags and process text between them
-                        for match in re.finditer(r"\[model:.*?\]", buffer):
-                            # Add text before the tag
-                            segment = buffer[last_processed_index:match.start()]
-                            if segment:
-                                processed_buffer += segment
-                            last_processed_index = match.end()
-                        
-                        # Add remaining text after the last complete tag
-                        processed_buffer += buffer[last_processed_index:]
-                        
-                        # Now, check for incomplete tag at the end
-                        last_tag_start_index = processed_buffer.rfind(tag_start)
-                        
-                        if last_tag_start_index != -1:
-                            # Text before the potential incomplete tag
-                            text_to_yield = processed_buffer[:last_tag_start_index]
-                            # Keep the potential incomplete tag start for the next iteration
-                            buffer = processed_buffer[last_tag_start_index:] 
-                        else:
-                            # No potential incomplete tag found, yield everything processed
-                            text_to_yield = processed_buffer
-                            buffer = "" # Clear buffer as everything is processed
-
-                        if text_to_yield:
-                            streaming_text += text_to_yield
-                            # Yield dict or raw string
-                            yield {"text": text_to_yield} if not raw else text_to_yield
-
-                # Process any remaining text in the buffer after the loop finishes
-                # Remove any potential tags (complete or incomplete)
-                final_text = re.sub(r"\[model:.*?\]", "", buffer)
-                if final_text:
-                    streaming_text += final_text
-                    yield {"text": final_text} if not raw else final_text
+                # Use sanitize_stream with the custom extractor
+                # It will decode bytes and yield processed string chunks
+                processed_stream = sanitize_stream(
+                    data=resp.iter_content(chunk_size=None), # Pass byte iterator
+                    intro_value=None, # No simple prefix
+                    to_json=False,    # Content is text
+                    content_extractor=self._toolbaz_extractor, # Use the tag remover
+                    yield_raw_on_error=True # Yield even if extractor somehow fails (though unlikely for regex)
+                )
+
+                for content_chunk in processed_stream:
+                    # content_chunk is the string with tags removed
+                    if content_chunk and isinstance(content_chunk, str):
+                        streaming_text += content_chunk
+                        yield {"text": content_chunk} if not raw else content_chunk
 
                 self.last_response = {"text": streaming_text}
                 self.conversation.update_chat_history(prompt, streaming_text)

webscout/Provider/turboseek.py

@@ -4,10 +4,10 @@ import json
 
 from webscout.AIutel import Optimizers
 from webscout.AIutel import Conversation
-from webscout.AIutel import AwesomePrompts, sanitize_stream
+from webscout.AIutel import AwesomePrompts, sanitize_stream # Import sanitize_stream
 from webscout.AIbase import Provider
 from webscout import exceptions
-from typing import Union, Any, AsyncGenerator, Dict
+from typing import Optional, Union, Any, AsyncGenerator, Dict
 from webscout.litagent import LitAgent
 
 class TurboSeek(Provider):
@@ -88,6 +88,13 @@ class TurboSeek(Provider):
         )
         self.conversation.history_offset = history_offset
 
+    @staticmethod
+    def _turboseek_extractor(chunk: Union[str, Dict[str, Any]]) -> Optional[str]:
+        """Extracts content from TurboSeek stream JSON objects."""
+        if isinstance(chunk, dict) and "text" in chunk:
+            return chunk.get("text") # json.loads already handles unicode escapes
+        return None
+
     def ask(
         self,
         prompt: str,
@@ -142,24 +149,24 @@ class TurboSeek(Provider):
                     raise exceptions.FailedToGenerateResponseError(
                         f"Failed to generate response - ({response.status_code}, {response.reason}) - {response.text}"
                     )
+
                 streaming_text = ""
-                # Iterate over bytes and decode manually
-                for value_bytes in response.iter_lines():
-                    try:
-                        if value_bytes and value_bytes.startswith(b"data: "): # Check for bytes
-                            # Decode bytes to string
-                            line = value_bytes[6:].decode('utf-8') 
-                            data = json.loads(line) 
-                            if "text" in data:
-                                # Decode potential unicode escapes
-                                content = data["text"].encode().decode('unicode_escape') 
-                                streaming_text += content
-                                resp = dict(text=content)
-                                self.last_response.update(resp) # Update last_response incrementally
-                                # Yield raw bytes or dict based on flag
-                                yield value_bytes if raw else resp 
-                    except (json.decoder.JSONDecodeError, UnicodeDecodeError):
-                        pass # Ignore lines that are not valid JSON or cannot be decoded
+                # Use sanitize_stream with the custom extractor
+                processed_stream = sanitize_stream(
+                    data=response.iter_content(chunk_size=None), # Pass byte iterator
+                    intro_value="data:",
+                    to_json=True,     # Stream sends JSON
+                    content_extractor=self._turboseek_extractor, # Use the specific extractor
+                    yield_raw_on_error=False # Skip non-JSON lines or lines where extractor fails
+                )
+
+                for content_chunk in processed_stream:
+                    # content_chunk is the string extracted by _turboseek_extractor
+                    if content_chunk and isinstance(content_chunk, str):
+                        streaming_text += content_chunk
+                        self.last_response.update(dict(text=streaming_text)) # Update last_response incrementally
+                        yield dict(text=content_chunk) if not raw else content_chunk # Yield dict or raw string
+
                 # Update conversation history after stream finishes
                 if streaming_text: # Only update if content was received
                     self.conversation.update_chat_history(
@@ -174,21 +181,15 @@ class TurboSeek(Provider):
         def for_non_stream():
             # Aggregate the stream using the updated for_stream logic
             full_text = ""
-            for chunk_data in for_stream():
-                 # Ensure chunk_data is a dict (not raw) and has 'text'
-                if isinstance(chunk_data, dict) and "text" in chunk_data:
-                    full_text += chunk_data["text"]
-                # If raw=True, chunk_data is bytes, decode and process if needed (though raw non-stream is less common)
-                elif isinstance(chunk_data, bytes):
-                     try:
-                         if chunk_data.startswith(b"data: "):
-                             line = chunk_data[6:].decode('utf-8')
-                             data = json.loads(line)
-                             if "text" in data:
-                                 content = data["text"].encode().decode('unicode_escape')
-                                 full_text += content
-                     except (json.decoder.JSONDecodeError, UnicodeDecodeError):
-                         pass
+            try:
+                # Ensure raw=False so for_stream yields dicts
+                for chunk_data in for_stream():
+                    if isinstance(chunk_data, dict) and "text" in chunk_data:
+                        full_text += chunk_data["text"]
+                    elif isinstance(chunk_data, str): # Handle case where raw=True was passed
+                        full_text += chunk_data
+            except Exception as e:
+                raise exceptions.FailedToGenerateResponseError(f"Failed to aggregate non-stream response: {e}") from e
             # last_response and history are updated within for_stream
             # Ensure last_response reflects the complete aggregated text
             self.last_response = {"text": full_text} 
@@ -241,7 +242,7 @@ class TurboSeek(Provider):
             str: Message extracted
         """
         assert isinstance(response, dict), "Response should be of dict data-type only"
-        # Text is already decoded in ask method
+        # Unicode escapes are handled by json.loads within sanitize_stream
         return response.get("text", "") 
 
 if __name__ == '__main__':
@@ -250,13 +251,9 @@ if __name__ == '__main__':
     try: # Add try-except block for testing
         ai = TurboSeek(timeout=60)
         print("[bold blue]Testing Stream:[/bold blue]")
-        response_stream = ai.chat("hello buddy", stream=True)
-        full_stream_response = ""
+        response_stream = ai.chat("yooooooooooo", stream=True)
         for chunk in response_stream:
             print(chunk, end="", flush=True)
-            full_stream_response += chunk
-        print("\n[bold green]Stream Test Complete.[/bold green]\n")
-
         # Optional: Test non-stream
         # print("[bold blue]Testing Non-Stream:[/bold blue]")
         # response_non_stream = ai.chat("What is the capital of France?", stream=False)
@@ -267,4 +264,3 @@ if __name__ == '__main__':
         print(f"\n[bold red]API Error:[/bold red] {e}")
     except Exception as e:
         print(f"\n[bold red]An unexpected error occurred:[/bold red] {e}")
-

webscout/Provider/typefully.py

@@ -1,10 +1,10 @@
-from typing import Union, Any, Dict
+from typing import Optional, Union, Any, Dict
 import re
 from uuid import uuid4
 
 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
@@ -114,6 +114,17 @@ class TypefullyAI(Provider):
         )
         self.conversation.history_offset = history_offset
 
+    @staticmethod
+    def _typefully_extractor(chunk: Union[str, Dict[str, Any]]) -> Optional[str]:
+        """Extracts content from the Typefully stream format '0:"..."'."""
+        if isinstance(chunk, str):
+            match = re.search(r'0:"(.*?)"(?=,|$)', chunk) # Look for 0:"...", possibly followed by comma or end of string
+            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 ask(
         self,
         prompt: str,
@@ -174,23 +185,22 @@ class TypefullyAI(Provider):
                     raise exceptions.FailedToGenerateResponseError(
                         f"Failed to generate response - ({response.status_code}, {response.reason}) - {response.text}"
                     )
-                streaming_response = ""
-                # Iterate over bytes and decode manually
-                for line_bytes in response.iter_lines():
-                    if line_bytes:
-                        line = line_bytes.decode('utf-8') # Decode bytes
-                        match = re.search(r'0:"(.*?)"', line)
-                        if match:
-                            # Decode potential unicode escapes
-                            content = match.group(1).encode().decode('unicode_escape') 
-                            streaming_response += content
-                            # Yield dict or raw string
-                            yield content if raw else dict(text=content)
-                        elif line.startswith('e:') or line.startswith('d:'):
-                            # End of response
-                            break
+                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._typefully_extractor, # Use the specific extractor
+                    end_marker="e:", # Stop processing if "e:" line is encountered (adjust if needed)
+                )
+
+                for content_chunk in processed_stream:
+                    if content_chunk and isinstance(content_chunk, str):
+                        streaming_text += content_chunk
+                        yield content_chunk if raw else dict(text=content_chunk)
                 # Update history and last response after stream finishes
-                self.last_response.update(dict(text=streaming_response))
+                self.last_response.update(dict(text=streaming_text))
                 self.conversation.update_chat_history(
                     prompt, self.get_message(self.last_response)
                 )
@@ -271,13 +281,12 @@ class TypefullyAI(Provider):
         """
         assert isinstance(response, dict), "Response should be of dict data-type only"
         # Handle potential unicode escapes in the final text
+        # Formatting is now handled by the extractor
         text = response.get("text", "")
         try:
-            # Attempt to decode escapes, return original if fails
-            # Already decoded in ask method, just handle formatting
             formatted_text = text.replace('\\n', '\n').replace('\\n\\n', '\n\n')
             return formatted_text
-        except Exception: # Catch potential errors during formatting
+        except Exception: # Catch potential errors during newline replacement
              return text # Return original text if formatting fails
 
 
@@ -319,4 +328,3 @@ if __name__ == "__main__":
 
         except Exception as e:
             print(f"\r{model:<50} {'✗':<10} {str(e)}")
-

webscout/Provider/typegpt.py

@@ -5,7 +5,7 @@ from typing import Union, Any, Dict, Generator
 
 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
@@ -137,37 +137,30 @@ class TypeGPT(Provider):
                     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 = ""
-            # Iterate over bytes and decode manually
-            for line_bytes in response.iter_lines():
-                if line_bytes:
-                    line = line_bytes.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
-                                    # Update last_response incrementally for potential non-stream use later
-                                    self.last_response = dict(text=message_load) 
-                        except json.JSONDecodeError:
-                            continue
+            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 message_load: # Only update if something was received
-                 self.conversation.update_chat_history(prompt, message_load)
+            if streaming_text: # Only update if something was received
+                 self.conversation.update_chat_history(prompt, streaming_text)
 
 
         def for_non_stream():
@@ -185,26 +178,30 @@ class TypeGPT(Provider):
                     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:
-                # curl_cffi response.json() handles decoding
-                response_data = response.json() 
-                # Extract the message content for history and return value
-                if 'choices' in response_data and len(response_data['choices']) > 0:
-                    message = response_data['choices'][0].get('message', {})
-                    content = message.get('content', '')
-                    self.last_response = {"text": content} # Store in expected format
-                    self.conversation.update_chat_history(prompt, content)
-                    return self.last_response
-                else:
-                     # Handle cases where response structure is unexpected
-                     self.last_response = {"text": ""} 
-                     return self.last_response
-            except json.JSONDecodeError as je:
+                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
+                )
+
+                # 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}")
 
 
@@ -290,4 +287,3 @@ if __name__ == "__main__":
 
         except Exception as e:
             print(f"\r{model:<50} {'✗':<10} {str(e)}")
-

webscout/Provider/uncovr.py

@@ -5,7 +5,7 @@ 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
@@ -109,6 +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*"?(.*?)"?$', chunk) # Match 0: maybe optional quotes
+            if match:
+                # Decode potential unicode escapes like \u00e9 and handle escaped quotes/backslashes
+                content = match.group(1).encode().decode('unicode_escape')
+                return content.replace('\\\\', '\\').replace('\\"', '"')
+        return None
+
     def refresh_identity(self, browser: str = None):
         """
         Refreshes the browser identity fingerprint.
@@ -202,27 +213,21 @@ class UncovrAI(Provider):
                         raise exceptions.FailedToGenerateResponseError(
                             f"Request failed with status code {response.status_code} - {response.text}"
                         )
-                
+
                 streaming_text = ""
-                # Iterate over bytes and decode manually
-                for line_bytes in response.iter_lines():
-                    if line_bytes:
-                        try:
-                            line = line_bytes.decode('utf-8')
-                            # Use regex to match content messages
-                            content_match = re.match(r'^0:\s*"?(.*?)"?$', line)
-                            if content_match:  # Content message
-                                content = content_match.group(1).encode().decode('unicode_escape') # Decode escapes
-                                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
+                # 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)
@@ -262,25 +267,25 @@ class UncovrAI(Provider):
                             f"Request failed with status code {response.status_code} - {response.text}"
                         )
 
-                # Process the non-streamed response content (assuming it's similar line format)
+                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 = ""
-                # Use response.text which should contain the full body for non-streamed curl_cffi requests
-                for line in response.text.splitlines(): 
-                    if line:
-                        try:
-                            # line is already decoded string
-                            content_match = re.match(r'^0:\s*"?(.*?)"?$', line)
-                            if content_match:
-                                content = content_match.group(1).encode().decode('unicode_escape') # Decode escapes
-                                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 less likely here
-                            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}
@@ -323,7 +328,9 @@ 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
@@ -359,4 +366,3 @@ if __name__ == "__main__":
             print(f"\r{model:<50} {status:<10} {display_text}")
         except Exception as e:
             print(f"\r{model:<50} {'✗':<10} {str(e)}")
-