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

webscout/Provider/sonus.py

@@ -1,9 +1,10 @@
-import requests
+from curl_cffi.requests import Session
+from curl_cffi import CurlError
 import json
 from typing import Any, Dict, Optional, Generator, Union
 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
@@ -21,7 +22,7 @@ class SonusAI(Provider):
     def __init__(
         self,
         is_conversation: bool = True,
-        max_tokens: int = 2049,
+        max_tokens: int = 2049, # Note: max_tokens is not directly used by this API
         timeout: int = 30,
         intro: str = None,
         filepath: str = None,
@@ -44,11 +45,14 @@ class SonusAI(Provider):
             'Origin': 'https://chat.sonus.ai',
             'Referer': 'https://chat.sonus.ai/',
             'User-Agent': LitAgent().random()
+            # Add sec-ch-ua headers if needed for impersonation consistency
         }
         
-        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)
+        self.session.proxies = proxies # Assign proxies directly
 
         self.is_conversation = is_conversation
         self.max_tokens_to_sample = max_tokens
@@ -74,6 +78,13 @@ class SonusAI(Provider):
         )
         self.conversation.history_offset = history_offset
 
+    @staticmethod
+    def _sonus_extractor(chunk: Union[str, Dict[str, Any]]) -> Optional[str]:
+        """Extracts content from Sonus stream JSON objects."""
+        if isinstance(chunk, dict) and "content" in chunk:
+            return chunk.get("content")
+        return None
+
     def ask(
         self,
         prompt: str,
@@ -92,72 +103,101 @@ class SonusAI(Provider):
             else:
                 raise Exception(f"Optimizer is not one of {self.__available_optimizers}")
 
-        # Prepare the multipart form data
-        files = {
-            'message': (None, conversation_prompt),
-            'history': (None),
-            'reasoning': (None, str(reasoning).lower()),
-            'model': (None, self.model)
+        # Prepare the multipart form data (curl_cffi handles tuples for files/data)
+        # No need for explicit (None, ...) for simple fields when using `data=`
+        form_data = {
+            'message': conversation_prompt,
+            'history': "", # Explicitly empty string if needed, or omit if None is acceptable
+            'reasoning': str(reasoning).lower(),
+            'model': self.model
         }
+        # Note: curl_cffi's `files` parameter is for actual file uploads.
+        # For simple key-value pairs like this, `data` is usually sufficient for multipart/form-data.
+        # If the server strictly requires `files`, keep the original structure but it might not work as expected with curl_cffi without actual file objects.
 
         def for_stream():
             try:
-                with requests.post(self.url, files=files, headers=self.headers, stream=True, timeout=self.timeout) as response:
-                    if response.status_code != 200:
-                        raise exceptions.FailedToGenerateResponseError(
-                            f"Request failed with status code {response.status_code}"
-                        )
-                    
-                    streaming_text = ""
-                    for line in response.iter_lines():
-                        if line:
-                            try:
-                                # Decode the line and remove 'data: ' prefix if present
-                                line = line.decode('utf-8')
-                                if line.startswith('data: '):
-                                    line = line[6:]
-                                
-                                data = json.loads(line)
-                                if "content" in data:
-                                    content = data["content"]
-                                    streaming_text += content
-                                    resp = dict(text=content)
-                                    yield resp if raw else resp
-                            except (json.JSONDecodeError, UnicodeDecodeError):
-                                continue
-                    
-                    self.last_response = {"text": streaming_text}
-                    self.conversation.update_chat_history(prompt, streaming_text)
+                # Use curl_cffi session post with impersonate
+                # Use `data` instead of `files` for simple key-value multipart
+                response = self.session.post(
+                    self.url, 
+                    # headers are set on the session
+                    data=form_data, # Use data for multipart form fields
+                    stream=True, 
+                    timeout=self.timeout,
+                    impersonate="chrome110" # Use a common impersonation profile
+                )
+                if response.status_code != 200:
+                    raise exceptions.FailedToGenerateResponseError(
+                        f"Request failed with status code {response.status_code} - {response.text}"
+                    )
+
+                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
+                    content_extractor=self._sonus_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 _sonus_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 history and last response after stream finishes
+                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): {str(e)}") from e
+            except Exception as e: # Catch other potential exceptions
+                 raise exceptions.FailedToGenerateResponseError(f"An unexpected error occurred ({type(e).__name__}): {e}") from e
 
         def for_non_stream():
             try:
-                response = requests.post(self.url, files=files, headers=self.headers, timeout=self.timeout)
+                 # Use curl_cffi session post with impersonate
+                response = self.session.post(
+                    self.url, 
+                    # headers are set on the session
+                    data=form_data, # Use data for multipart form fields
+                    timeout=self.timeout,
+                    impersonate="chrome110" # Use a common impersonation profile
+                )
                 if response.status_code != 200:
                     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_raw = response.text # Get raw text
+
+                # Use sanitize_stream to process the non-streaming text
+                processed_stream = sanitize_stream(
+                    data=response_text_raw.splitlines(), # Split into lines
+                    intro_value="data:",
+                    to_json=True,
+                    content_extractor=self._sonus_extractor,
+                    yield_raw_on_error=False
+                )
+
+                # Aggregate the results
                 full_response = ""
-                for line in response.iter_lines():
-                    if line:
-                        try:
-                            line = line.decode('utf-8')
-                            if line.startswith('data: '):
-                                line = line[6:]
-                            data = json.loads(line)
-                            if "content" in data:
-                                full_response += data["content"]
-                        except (json.JSONDecodeError, UnicodeDecodeError):
-                            continue
+                for content in processed_stream:
+                    if content and isinstance(content, str):
+                        full_response += content
 
                 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}")
+                # Return dict or raw string
+                return full_response if raw else {"text": full_response}
+                
+            except CurlError as e: # Catch CurlError
+                raise exceptions.FailedToGenerateResponseError(f"Request failed (CurlError): {str(e)}") from e
+            except Exception as e: # Catch other potential exceptions
+                raise exceptions.FailedToGenerateResponseError(f"An unexpected error occurred ({type(e).__name__}): {e}") from e
 
         return for_stream() if stream else for_non_stream()
 
@@ -169,20 +209,30 @@ class SonusAI(Provider):
         conversationally: bool = False,
         reasoning: bool = False,
     ) -> Union[str, Generator[str, None, None]]:
-        def for_stream():
-            for response in self.ask(prompt, True, optimizer=optimizer, conversationally=conversationally, reasoning=reasoning):
-                yield self.get_message(response)
-        def for_non_stream():
-            return self.get_message(
-                self.ask(prompt, False, optimizer=optimizer, conversationally=conversationally, reasoning=reasoning)
+        def for_stream_chat():
+            # ask() yields dicts when raw=False
+            for response_dict in self.ask(
+                prompt, stream=True, raw=False, # Ensure ask yields dicts
+                optimizer=optimizer, conversationally=conversationally, reasoning=reasoning
+            ):
+                yield self.get_message(response_dict)
+        
+        def for_non_stream_chat():
+             # ask() returns dict or str when raw=False/True
+            response_data = self.ask(
+                prompt, stream=False, raw=False, # Ensure ask returns dict
+                optimizer=optimizer, conversationally=conversationally, reasoning=reasoning
             )
-        return for_stream() if stream else for_non_stream()
+            return self.get_message(response_data) # get_message expects dict
+
+        return for_stream_chat() if stream else for_non_stream_chat()
 
     def get_message(self, response: dict) -> str:
         assert isinstance(response, dict), "Response should be of dict data-type only"
         return response["text"]
 
 if __name__ == "__main__":
+    # Ensure curl_cffi is installed
     print("-" * 80)
     print(f"{'Model':<50} {'Status':<10} {'Response'}")
     print("-" * 80)
@@ -205,4 +255,4 @@ if __name__ == "__main__":
                 display_text = "Empty or invalid 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)}")

webscout/Provider/toolbaz.py

@@ -1,5 +1,6 @@
 import re
-import requests
+from curl_cffi.requests import Session
+from curl_cffi import CurlError
 import uuid
 import base64
 import json
@@ -9,11 +10,11 @@ import time
 from datetime import datetime
 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.AIbase import Provider, AsyncProvider
-from webscout import exceptions
+from webscout.AIutel import AwesomePrompts, sanitize_stream # Import sanitize_stream
+from webscout.AIbase import Provider
 
 class Toolbaz(Provider):
     """
@@ -25,6 +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",
@@ -47,7 +49,7 @@ class Toolbaz(Provider):
     def __init__(
         self,
         is_conversation: bool = True,
-        max_tokens: int = 600,
+        max_tokens: int = 600, # Note: max_tokens is not directly used by the API
         timeout: int = 30,
         intro: str = None,
         filepath: str = None,
@@ -56,7 +58,7 @@ class Toolbaz(Provider):
         history_offset: int = 10250,
         act: str = None,
         model: str = "gemini-2.0-flash",
-        system_prompt: str = "You are a helpful AI assistant."
+        system_prompt: str = "You are a helpful AI assistant." # Note: system_prompt is not directly used by the API
     ):
         """
         Initializes the Toolbaz API with given parameters.
@@ -64,28 +66,31 @@ class Toolbaz(Provider):
         if model not in self.AVAILABLE_MODELS:
             raise ValueError(f"Invalid model: {model}. Choose from: {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.timeout = timeout
         self.last_response = {}
         self.system_prompt = system_prompt
         self.model = model
-        self.proxies = proxies
+        self.proxies = proxies # Store proxies for later use in requests
 
-        # Set up headers
+        # Set up headers for the curl_cffi session
         self.session.headers.update({
-            "user-agent": "Mozilla/5.0 (Linux; Android 10)",
+            "user-agent": "Mozilla/5.0 (Linux; Android 10)", # Keep specific user-agent
             "accept": "*/*",
             "accept-language": "en-US",
             "cache-control": "no-cache",
-            "connection": "keep-alive",
             "content-type": "application/x-www-form-urlencoded; charset=UTF-8",
             "origin": "https://toolbaz.com",
             "pragma": "no-cache",
             "referer": "https://toolbaz.com/",
             "sec-fetch-mode": "cors"
+            # Add sec-ch-ua headers if needed for impersonation consistency
         })
+        # Assign proxies directly to the session
+        self.session.proxies = proxies
 
         # Initialize conversation history
         self.__available_optimizers = (
@@ -107,6 +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))
 
@@ -139,20 +151,34 @@ class Toolbaz(Provider):
                 "session_id": session_id,
                 "token": token
             }
-            resp = self.session.post("https://data.toolbaz.com/token.php", data=data)
-            resp.raise_for_status()
+            # Use curl_cffi session post WITHOUT impersonate for token request
+            resp = self.session.post(
+                "https://data.toolbaz.com/token.php", 
+                data=data
+                # Removed impersonate="chrome110" for this specific request
+            )
+            resp.raise_for_status() # Check for HTTP errors
             result = resp.json()
             if result.get("success"):
                 return {"token": result["token"], "session_id": session_id}
-            return None
-        except Exception:
-            return None
+            # Raise error if success is not true
+            raise exceptions.FailedToGenerateResponseError(f"Authentication failed: API response indicates failure. Response: {result}")
+        except CurlError as e: # Catch CurlError specifically
+            # Raise a specific error indicating CurlError during auth
+            raise exceptions.FailedToGenerateResponseError(f"Authentication failed due to network error (CurlError): {e}") from e
+        except json.JSONDecodeError as e:
+             # Raise error for JSON decoding issues
+             raise exceptions.FailedToGenerateResponseError(f"Authentication failed: Could not decode JSON response. Error: {e}. Response text: {getattr(resp, 'text', 'N/A')}") from e
+        except Exception as e: # Catch other potential errors (like HTTPError from raise_for_status)
+            # Raise a specific error indicating a general failure during auth
+            err_text = getattr(e, 'response', None) and getattr(e.response, 'text', '')
+            raise exceptions.FailedToGenerateResponseError(f"Authentication failed due to an unexpected error ({type(e).__name__}): {e} - {err_text}") from e
 
     def ask(
         self,
         prompt: str,
         stream: bool = False,
-        raw: bool = False,  # Kept for compatibility with other providers
+        raw: bool = False,  # Kept for compatibility, but output is always dict/string
         optimizer: Optional[str] = None,
         conversationally: bool = False,
     ) -> Union[Dict[str, Any], Generator]:
@@ -166,9 +192,9 @@ class Toolbaz(Provider):
                 conversation_prompt if conversationally else prompt
             )
 
-        auth = self.get_auth()
-        if not auth:
-            raise exceptions.ProviderConnectionError("Failed to authenticate with Toolbaz API")
+        # get_auth now raises exceptions on failure
+        auth = self.get_auth() 
+        # No need to check if auth is None, as an exception would have been raised
 
         data = {
             "text": conversation_prompt,
@@ -179,67 +205,55 @@ class Toolbaz(Provider):
 
         def for_stream():
             try:
+                # Use curl_cffi session post with impersonate for the main request
                 resp = self.session.post(
                     "https://data.toolbaz.com/writing.php",
                     data=data,
                     stream=True,
-                    proxies=self.proxies,
-                    timeout=self.timeout
+                    timeout=self.timeout,
+                    impersonate="chrome110" # Keep impersonate here
                 )
                 resp.raise_for_status()
 
-                buffer = ""
-                tag_start = "[model:"
                 streaming_text = ""
 
-                for chunk in resp.iter_content(chunk_size=1):
-                    if chunk:
-                        text = chunk.decode(errors="ignore")
-                        buffer += text
-                        # Remove all complete [model: ...] tags in buffer
-                        while True:
-                            match = re.search(r"\[model:.*?\]", buffer)
-                            if not match:
-                                break
-                            buffer = buffer[:match.start()] + buffer[match.end():]
-                        # Only yield up to the last possible start of a tag
-                        last_tag = buffer.rfind(tag_start)
-                        if last_tag == -1 or last_tag + len(tag_start) > len(buffer):
-                            if buffer:
-                                streaming_text += buffer
-                                yield {"text": buffer}
-                                buffer = ""
-                        else:
-                            if buffer[:last_tag]:
-                                streaming_text += buffer[:last_tag]
-                                yield {"text": buffer[:last_tag]}
-                            buffer = buffer[last_tag:]
-
-                # Remove any remaining [model: ...] tag in the buffer
-                buffer = re.sub(r"\[model:.*?\]", "", buffer)
-                if buffer:
-                    streaming_text += buffer
-                    yield {"text": buffer}
+                # 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)
 
-            except requests.exceptions.RequestException as e:
-                raise exceptions.ProviderConnectionError(f"Network error: {str(e)}") from e
-            except Exception as e:
-                raise exceptions.ProviderConnectionError(f"Unexpected error: {str(e)}") from e
+            except CurlError as e: # Catch CurlError
+                raise exceptions.FailedToGenerateResponseError(f"Network error (CurlError): {str(e)}") from e
+            except Exception as e: # Catch other exceptions
+                raise exceptions.FailedToGenerateResponseError(f"Unexpected error during stream: {str(e)}") from e
 
         def for_non_stream():
             try:
+                # Use curl_cffi session post with impersonate for the main request
                 resp = self.session.post(
                     "https://data.toolbaz.com/writing.php",
                     data=data,
-                    proxies=self.proxies,
-                    timeout=self.timeout
+                    timeout=self.timeout,
+                    impersonate="chrome110" # Keep impersonate here
                 )
                 resp.raise_for_status()
 
-                text = resp.text
+                # Use response.text which is already decoded
+                text = resp.text 
                 # Remove [model: ...] tags
                 text = re.sub(r"\[model:.*?\]", "", text)
 
@@ -248,9 +262,9 @@ class Toolbaz(Provider):
 
                 return self.last_response
 
-            except requests.exceptions.RequestException as e:
-                raise exceptions.FailedToGenerateResponseError(f"Network error: {str(e)}") from e
-            except Exception as e:
+            except CurlError as e: # Catch CurlError
+                raise exceptions.FailedToGenerateResponseError(f"Network error (CurlError): {str(e)}") from e
+            except Exception as e: # Catch other exceptions
                 raise exceptions.FailedToGenerateResponseError(f"Unexpected error: {str(e)}") from e
 
         return for_stream() if stream else for_non_stream()
@@ -263,26 +277,28 @@ class Toolbaz(Provider):
         conversationally: bool = False,
     ) -> Union[str, Generator[str, None, None]]:
         """Generates a response from the Toolbaz API."""
-        def for_stream():
-            for response in self.ask(
+        def for_stream_chat():
+            # ask() yields dicts when raw=False
+            for response_dict in self.ask(
                 prompt,
                 stream=True,
+                raw=False, # Ensure ask yields dicts
                 optimizer=optimizer,
                 conversationally=conversationally
             ):
-                yield self.get_message(response)
+                yield self.get_message(response_dict)
 
-        def for_non_stream():
-            return self.get_message(
-                self.ask(
-                    prompt,
-                    stream=False,
-                    optimizer=optimizer,
-                    conversationally=conversationally,
-                )
+        def for_non_stream_chat():
+            # ask() returns a dict when stream=False
+            response_dict = self.ask(
+                prompt,
+                stream=False,
+                optimizer=optimizer,
+                conversationally=conversationally,
             )
+            return self.get_message(response_dict)
 
-        return for_stream() if stream else for_non_stream()
+        return for_stream_chat() if stream else for_non_stream_chat()
 
     def get_message(self, response: Dict[str, Any]) -> str:
         """Extract the message from the response.
@@ -298,23 +314,40 @@ class Toolbaz(Provider):
 
 # Example usage
 if __name__ == "__main__":
+    # Ensure curl_cffi is installed
+    from rich import print # Use rich print if available
+    print("-" * 80)
+    print(f"{'Model':<50} {'Status':<10} {'Response'}")
+    print("-" * 80)
     # Test the provider with different models
     for model in Toolbaz.AVAILABLE_MODELS:
         try:
             test_ai = Toolbaz(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 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()
+                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
+            # 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)}")
+            # Print full error for debugging
+            print(f"\r{model:<50} {'✗':<10} Error: {str(e)}")