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

webscout/Provider/granite.py

@@ -1,8 +1,9 @@
-import requests
+from curl_cffi.requests import Session
+from curl_cffi import CurlError
 import json
-from typing import Union, Any, Dict, Generator
+from typing import Optional, Union, Any, Dict, Generator
 
-from webscout.AIutel import Optimizers, Conversation, AwesomePrompts
+from webscout.AIutel import Optimizers, Conversation, AwesomePrompts, sanitize_stream # Import sanitize_stream
 from webscout.AIbase import Provider
 from webscout import exceptions
 from webscout.litagent import LitAgent as Lit
@@ -19,7 +20,7 @@ class IBMGranite(Provider):
         self,
         api_key: str,
         is_conversation: bool = True,
-        max_tokens: int = 600,
+        max_tokens: int = 600, # Note: max_tokens is not used by this API
         timeout: int = 30,
         intro: str = None,
         filepath: str = None,
@@ -35,7 +36,8 @@ class IBMGranite(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.api_endpoint = "https://d18n68ssusgr7r.cloudfront.net/v1/chat/completions"
@@ -46,18 +48,19 @@ class IBMGranite(Provider):
         self.system_prompt = system_prompt
         self.thinking = thinking
 
-        # Use Lit agent to generate a random User-Agent
+        # Use Lit agent (keep if needed for other headers or logic)
         self.headers = {
-            "authority": "d18n68ssusgr7r.cloudfront.net",
-            "accept": "application/json,application/jsonl",
+            "authority": "d18n68ssusgr7r.cloudfront.net", # Keep authority
+            "accept": "application/json,application/jsonl", # Keep accept
             "content-type": "application/json",
-            "origin": "https://www.ibm.com",
-            "referer": "https://www.ibm.com/",
-            "user-agent": Lit().random(),
+            "origin": "https://www.ibm.com", # Keep origin
+            "referer": "https://www.ibm.com/", # Keep referer
         }
         self.headers["Authorization"] = f"Bearer {api_key}"
+        
+        # Update curl_cffi session headers and proxies
         self.session.headers.update(self.headers)
-        self.session.proxies = proxies
+        self.session.proxies = proxies # Assign proxies directly
 
         self.__available_optimizers = (
             method for method in dir(Optimizers)
@@ -74,10 +77,17 @@ class IBMGranite(Provider):
         self.conversation = Conversation(is_conversation, self.max_tokens_to_sample, filepath, update_file)
         self.conversation.history_offset = history_offset
 
+    @staticmethod
+    def _granite_extractor(chunk: Union[str, Dict[str, Any]]) -> Optional[str]:
+        """Extracts content from IBM Granite stream JSON lists [3, "text"]."""
+        if isinstance(chunk, list) and len(chunk) == 2 and chunk[0] == 3 and isinstance(chunk[1], str):
+            return chunk[1]
+        return None
+
     def ask(
         self,
         prompt: str,
-        stream: bool = False,
+        stream: bool = False, # API supports streaming
         raw: bool = False,
         optimizer: str = None,
         conversationally: bool = False,
@@ -107,48 +117,77 @@ class IBMGranite(Provider):
                 {"role": "system", "content": self.system_prompt},
                 {"role": "user", "content": conversation_prompt},
             ],
-            "stream": stream,
-            "thinking": self.thinking,
+            "stream": True # API seems to require stream=True based on response format
         }
 
         def for_stream():
+            streaming_text = "" # Initialize outside try block
             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 are set on the session
+                    json=payload, 
+                    stream=True, 
+                    timeout=self.timeout,
+                    impersonate="chrome110" # Use a common impersonation profile
+                )
+                response.raise_for_status() # Check for HTTP errors
+
+                # Use sanitize_stream
+                processed_stream = sanitize_stream(
+                    data=response.iter_content(chunk_size=None), # Pass byte iterator
+                    intro_value=None, # No prefix
+                    to_json=True,     # Stream sends JSON lines (which are lists)
+                    content_extractor=self._granite_extractor, # Use the specific extractor
+                    yield_raw_on_error=False # Skip non-JSON lines or lines where extractor fails
                 )
-                if not response.ok:
-                    msg = f"Request failed with status code {response.status_code}: {response.text}"
-                    raise exceptions.FailedToGenerateResponseError(msg)
-
-                streaming_text = ""
-                for line in response.iter_lines(decode_unicode=True):
-                    if line:
-                        try:
-                            data = json.loads(line)
-                            if len(data) == 2 and data[0] == 3 and isinstance(data[1], str):
-                                content = data[1]
-                                streaming_text += content
-                                yield content if raw else dict(text=content)
-                            else:
-                                # Skip unrecognized lines
-                                pass
-                        except json.JSONDecodeError:
-                            continue
-                self.last_response.update(dict(text=streaming_text))
-                self.conversation.update_chat_history(prompt, self.get_message(self.last_response))
-            except requests.exceptions.RequestException as e:
-                raise exceptions.ProviderConnectionError(f"Request failed: {e}")
-            except json.JSONDecodeError as e:
-                raise exceptions.InvalidResponseError(f"Failed to decode JSON response: {e}")
-            except Exception as e:
-                raise exceptions.FailedToGenerateResponseError(f"An unexpected error occurred: {e}")
+
+                for content_chunk in processed_stream:
+                    # content_chunk is the string extracted by _granite_extractor
+                    if content_chunk and isinstance(content_chunk, str):
+                        streaming_text += content_chunk
+                        resp = dict(text=content_chunk)
+                        yield resp if not raw else content_chunk
+                
+                # Update history after stream finishes
+                self.last_response = dict(text=streaming_text)
+                self.conversation.update_chat_history(prompt, streaming_text)
+                
+            except CurlError as e: # Catch CurlError
+                raise exceptions.ProviderConnectionError(f"Request failed (CurlError): {e}") from e
+            except json.JSONDecodeError as e: # Keep specific JSON error handling
+                raise exceptions.InvalidResponseError(f"Failed to decode JSON response: {e}") from e
+            except Exception as e: # Catch other potential exceptions (like HTTPError)
+                err_text = getattr(e, 'response', None) and getattr(e.response, 'text', '')
+                # Use specific exception type if available, otherwise generic
+                ex_type = exceptions.FailedToGenerateResponseError if not isinstance(e, exceptions.ProviderConnectionError) else type(e)
+                raise ex_type(f"An unexpected error occurred ({type(e).__name__}): {e} - {err_text}") from e
+
 
         def for_non_stream():
-            # Run the generator to completion
-            for _ in for_stream():
-                pass
-            return self.last_response
+            # Aggregate the stream using the updated for_stream logic
+            full_text = ""
+            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"]
+                    # Handle raw string case if raw=True was passed
+                    elif raw and isinstance(chunk_data, str):
+                         full_text += chunk_data
+            except Exception as e:
+                 # If aggregation fails but some text was received, use it. Otherwise, re-raise.
+                 if not full_text:
+                     raise exceptions.FailedToGenerateResponseError(f"Failed to get non-stream response: {str(e)}") from e
 
+            # last_response and history are updated within for_stream
+            # Return the final aggregated response dict or raw string
+            return full_text if raw else self.last_response
+
+
+        # Since the API endpoint suggests streaming, always call the stream generator.
+        # The non-stream wrapper will handle aggregation if stream=False.
         return for_stream() if stream else for_non_stream()
 
     def chat(
@@ -159,16 +198,24 @@ class IBMGranite(Provider):
         conversationally: bool = False,
     ) -> Union[str, Generator[str, None, None]]:
         """Generate response as a string using chat method"""
-        def for_stream():
-            for response in self.ask(prompt, True, optimizer=optimizer, conversationally=conversationally):
-                yield self.get_message(response)
-
-        def for_non_stream():
-            return self.get_message(
-                self.ask(prompt, False, optimizer=optimizer, conversationally=conversationally)
+        def for_stream_chat():
+            # ask() yields dicts or strings when streaming
+            gen = self.ask(
+                prompt, stream=True, raw=False, # Ensure ask yields dicts
+                optimizer=optimizer, conversationally=conversationally
+            )
+            for response_dict in gen:
+                yield self.get_message(response_dict) # get_message expects dict
+
+        def for_non_stream_chat():
+            # ask() returns dict or str when not streaming
+            response_data = self.ask(
+                prompt, stream=False, raw=False, # Ensure ask returns dict
+                optimizer=optimizer, conversationally=conversationally
             )
+            return self.get_message(response_data) # get_message expects 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:
         """Retrieves message only from response"""
@@ -176,6 +223,7 @@ class IBMGranite(Provider):
         return response["text"]
 
 if __name__ == "__main__":
+    # Ensure curl_cffi is installed
     from rich import print
     # Example usage: Initialize without logging.
     ai = IBMGranite(

webscout/Provider/hermes.py

@@ -1,9 +1,10 @@
-import requests
+from curl_cffi.requests import Session
+from curl_cffi import CurlError
 import json
 from typing import Union, Any, Dict, Generator, Optional
 
 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
@@ -38,7 +39,7 @@ class NousHermes(Provider):
                 f"Invalid model: {model}. Choose from: {self.AVAILABLE_MODELS}"
             )
 
-        self.session = requests.Session()
+        self.session = Session()
         self.is_conversation = is_conversation
         self.max_tokens_to_sample = max_tokens
         self.timeout = timeout
@@ -49,15 +50,14 @@ class NousHermes(Provider):
         self.temperature = temperature
         self.top_p = top_p
         self.cookies_path = cookies_path
-        self.cookies = self._load_cookies()
+        self.cookies_dict = self._load_cookies()
+        
         self.headers = {
             'accept': '*/*',
             'accept-language': 'en-US,en;q=0.9',
             'content-type': 'application/json',
             'origin': 'https://hermes.nousresearch.com',
             'referer': 'https://hermes.nousresearch.com/',
-            'user-agent': LitAgent().random(),
-            'cookie': self.cookies
         }
 
         self.__available_optimizers = (
@@ -77,20 +77,38 @@ class NousHermes(Provider):
             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.proxies = proxies
+        
+        # Apply cookies to curl_cffi session
+        if self.cookies_dict:
+            for name, value in self.cookies_dict.items():
+                self.session.cookies.set(name, value, domain="hermes.nousresearch.com")
 
-    def _load_cookies(self) -> Optional[str]:
-        """Load cookies from a JSON file and convert them to a string."""
+    def _load_cookies(self) -> Optional[Dict[str, str]]:
+        """Load cookies from a JSON file and return them as a dictionary."""
         try:
             with open(self.cookies_path, 'r') as f:
                 cookies_data = json.load(f)
-            return '; '.join([f"{cookie['name']}={cookie['value']}" for cookie in cookies_data])
+            # Convert list of cookie objects to a dictionary
+            return {cookie['name']: cookie['value'] for cookie in cookies_data if 'name' in cookie and 'value' in cookie}
         except FileNotFoundError:
-            print("Error: cookies.json file not found!")
+            print(f"Warning: Cookies file not found at {self.cookies_path}")
             return None
         except json.JSONDecodeError:
-            print("Error: Invalid JSON format in cookies.json!")
+            print(f"Warning: Invalid JSON format in cookies file at {self.cookies_path}")
             return None
+        except Exception as e:
+            print(f"Warning: Error loading cookies: {e}")
+            return None
+
+    @staticmethod
+    def _hermes_extractor(chunk: Union[str, Dict[str, Any]]) -> Optional[str]:
+        """Extracts content from Hermes stream JSON objects."""
+        if isinstance(chunk, dict) and chunk.get('type') == 'llm_response':
+            return chunk.get('content')
+        return None
+
 
     def ask(
         self,
@@ -134,32 +152,59 @@ class NousHermes(Provider):
             "top_p": self.top_p,
         }
         def for_stream():
-            response = self.session.post(self.api_endpoint, headers=self.headers, json=payload, stream=True, timeout=self.timeout)
-            if not response.ok:
-                raise exceptions.FailedToGenerateResponseError(
-                    f"Failed to generate response - ({response.status_code}, {response.reason}) - {response.text}"
+            streaming_text = "" # Initialize outside try block
+            try:
+                response = self.session.post(
+                    self.api_endpoint, 
+                    json=payload, 
+                    stream=True, 
+                    timeout=self.timeout,
+                    impersonate="chrome110" # Keep impersonate
+                )
+                response.raise_for_status()
+
+                # 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._hermes_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 _hermes_extractor
+                    if content_chunk and isinstance(content_chunk, str):
+                        streaming_text += content_chunk
+                        resp = dict(text=content_chunk)
+                        yield resp if not raw else content_chunk
+
+                self.last_response = dict(text=streaming_text) # Use streaming_text
+                self.conversation.update_chat_history(
+                    prompt, streaming_text # Use streaming_text
                 )
-            full_response = ""
-            for line in response.iter_lines():
-                if line:
-                    decoded_line = line.decode('utf-8').replace('data: ', '')
-                    try:
-                        data = json.loads(decoded_line)
-                        if data['type'] == 'llm_response':
-                            content = data['content']
-                            full_response += content
-                            yield content if raw else dict(text=content)
-                    except json.JSONDecodeError:
-                        continue
-            self.last_response.update(dict(text=full_response))
-            self.conversation.update_chat_history(
-                prompt, self.get_message(self.last_response)
-            )
+
+            except CurlError as e:
+                raise exceptions.FailedToGenerateResponseError(f"Request failed (CurlError): {e}") from e
+            except Exception as e:
+                err_text = getattr(e, 'response', None) and getattr(e.response, 'text', '')
+                raise exceptions.FailedToGenerateResponseError(f"Failed to generate response ({type(e).__name__}): {e} - {err_text}") from e
+
 
         def for_non_stream():
-            for _ in for_stream():
-                pass
-            return self.last_response
+            collected_text = ""
+            try:
+                for chunk_data in for_stream():
+                    if isinstance(chunk_data, dict) and "text" in chunk_data:
+                        collected_text += chunk_data["text"]
+                    elif raw and isinstance(chunk_data, str):
+                         collected_text += chunk_data
+            except Exception as e:
+                 if not collected_text:
+                     raise exceptions.FailedToGenerateResponseError(f"Failed to get non-stream response: {str(e)}") from e
+
+            return collected_text if raw else self.last_response
+
 
         return for_stream() if stream else for_non_stream()
 
@@ -180,23 +225,25 @@ class NousHermes(Provider):
             str: Response generated
         """
 
-        def for_stream():
-            for response in self.ask(
-                prompt, True, optimizer=optimizer, conversationally=conversationally
-            ):
-                yield self.get_message(response)
+        def for_stream_chat():
+            gen = self.ask(
+                prompt, stream=True, raw=False,
+                optimizer=optimizer, conversationally=conversationally
+            )
+            for response_dict in gen:
+                yield self.get_message(response_dict)
 
-        def for_non_stream():
-            return self.get_message(
-                self.ask(
-                    prompt,
-                    False,
-                    optimizer=optimizer,
-                    conversationally=conversationally,
-                )
+        def for_non_stream_chat():
+            response_data = self.ask(
+                prompt,
+                stream=False,
+                raw=False,
+                optimizer=optimizer,
+                conversationally=conversationally,
             )
+            return self.get_message(response_data)
 
-        return for_stream() if stream else for_non_stream()
+        return for_stream_chat() if stream else for_non_stream_chat()
 
     def get_message(self, response: dict) -> str:
         """Retrieves message only from response

webscout/Provider/koala.py

@@ -28,7 +28,7 @@ class KOALA(Provider):
         proxies: dict = {},
         history_offset: int = 10250,
         act: str = None,
-        model: str = "gpt-4o",
+        model: str = "gpt-4.1",
         web_search: bool = True,
 
     ) -> None: