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

webscout/Provider/Cloudflare.py

@@ -1,13 +1,15 @@
 import json
 from uuid import uuid4
-import webscout
+
+import re # Import re
+from curl_cffi import CurlError
 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, AsyncProvider
 from webscout import exceptions
-from typing import Union, Any, AsyncGenerator, Dict
-import cloudscraper
+from typing import Optional, Union, Any, AsyncGenerator, Dict
+from curl_cffi.requests import Session
 from webscout.litagent import LitAgent
 
 class Cloudflare(Provider):
@@ -96,7 +98,7 @@ class Cloudflare(Provider):
         if model not in self.AVAILABLE_MODELS:
             raise ValueError(f"Invalid model: {model}. Choose from: {self.AVAILABLE_MODELS}")
 
-        self.scraper = cloudscraper.create_scraper()
+        self.session = Session() # Use curl_cffi Session
         self.is_conversation = is_conversation
         self.max_tokens_to_sample = max_tokens
         self.chat_endpoint = "https://playground.ai.cloudflare.com/api/inference"
@@ -136,7 +138,7 @@ class Cloudflare(Provider):
         )
 
         # Initialize session and apply proxies
-        self.session = cloudscraper.create_scraper()
+        # self.session = cloudscraper.create_scraper() # Replaced above
         self.session.headers.update(self.headers)
         self.session.proxies = proxies
 
@@ -156,6 +158,19 @@ class Cloudflare(Provider):
         # if self.logger:
         #     self.logger.info("Cloudflare initialized successfully")
 
+    @staticmethod
+    def _cloudflare_extractor(chunk: Union[str, Dict[str, Any]]) -> Optional[str]:
+        """Extracts content from Cloudflare stream JSON objects."""
+        # Updated for the 0:"..." format
+        if isinstance(chunk, str):
+            # Use re.search to find the pattern 0:"<content>"
+            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,
@@ -201,37 +216,69 @@ class Cloudflare(Provider):
 
         def for_stream():
             # if self.logger:
-            #     self.logger.debug("Sending streaming request to Cloudflare API...")
-            response = self.scraper.post(
-                self.chat_endpoint,
-                headers=self.headers,
-                cookies=self.cookies,
-                data=json.dumps(payload),
-                stream=True,
-                timeout=self.timeout
-            )
-            if not response.ok:
-                # if self.logger:
-                #     self.logger.error(f"Request failed: ({response.status_code}, {response.reason})")
-                raise exceptions.FailedToGenerateResponseError(
-                    f"Failed to generate response - ({response.status_code}, {response.reason})"
+            #     self.logger.debug("Sending streaming request to Cloudflare API...") 
+            streaming_text = "" # Initialize outside try block
+            try:
+                response = self.session.post(
+                    self.chat_endpoint,
+                    headers=self.headers,
+                    cookies=self.cookies,
+                    data=json.dumps(payload),
+                    stream=True,
+                    timeout=self.timeout,
+                    impersonate="chrome120" # Add impersonate
+                )
+                response.raise_for_status()
+
+                # Use sanitize_stream
+                processed_stream = sanitize_stream(
+                    data=response.iter_content(chunk_size=None), # Pass byte iterator
+                    intro_value=None,
+                    to_json=False, 
+                    skip_markers=None,
+                    content_extractor=self._cloudflare_extractor, # Use the specific extractor
+                    yield_raw_on_error=False # Skip non-JSON lines or lines where extractor fails
                 )
-            streaming_response = ""
-            for line in response.iter_lines(decode_unicode=True):
-                if line.startswith('data: ') and line != 'data: [DONE]':
-                    data = json.loads(line[6:])
-                    content = data.get('response', '')
-                    streaming_response += content
-                    yield content if raw else dict(text=content)
-            self.last_response.update(dict(text=streaming_response))
+
+                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)
+
+            except CurlError as e:
+                raise exceptions.FailedToGenerateResponseError(f"Request failed (CurlError): {e}") from e
+            except Exception as e:
+                raise exceptions.FailedToGenerateResponseError(f"An unexpected error occurred ({type(e).__name__}): {e}") from e
+            finally:
+                # Update history after stream finishes or fails
+                self.last_response.update(dict(text=streaming_text))
+                self.conversation.update_chat_history(prompt, streaming_text)
+
+        def for_non_stream():
+            # Aggregate the stream using the updated for_stream logic
+            full_text = ""
+            last_response_dict = {}
             self.conversation.update_chat_history(prompt, self.get_message(self.last_response))
             # if self.logger:
             #     self.logger.info("Streaming response completed successfully")
+            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"]
+                        last_response_dict = {"text": full_text} # Keep track of last dict structure
+                    # Handle raw string case if raw=True was passed
+                    elif raw and isinstance(chunk_data, str):
+                         full_text += chunk_data
+                         last_response_dict = {"text": full_text} # Update dict even for raw
+            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
 
-        def for_non_stream():
-            for _ in for_stream():
-                pass
-            return self.last_response
+            # last_response and history are updated within for_stream's finally block
+            # Return the final aggregated response dict or raw text
+            return full_text if raw else last_response_dict
 
         return for_stream() if stream else for_non_stream()
 

webscout/Provider/Deepinfra.py

@@ -1,11 +1,12 @@
-import requests
+from curl_cffi.requests import Session
+from curl_cffi import CurlError
 import json
 import os
-from typing import Any, Dict, Optional, Generator, Union
+from typing import Any, Dict, Optional, Generator, Union, List
 
 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, AsyncProvider
 from webscout import exceptions
 from webscout.litagent import LitAgent
@@ -65,14 +66,24 @@ class DeepInfra(Provider):
         # "Qwen/Qwen2.5-7B-Instruct",  # >>>> NOT WORKING
         "Qwen/Qwen2.5-72B-Instruct",
         "Qwen/Qwen2.5-Coder-32B-Instruct",
+        "Qwen/Qwen3-14B",
+        "Qwen/Qwen3-30B-A3B",
+        "Qwen/Qwen3-32B",
         # "Sao10K/L3.1-70B-Euryale-v2.2",  # >>>> NOT WORKING
         # "Sao10K/L3.3-70B-Euryale-v2.3",  # >>>> NOT WORKING
     ]
 
+    @staticmethod
+    def _deepinfra_extractor(chunk: Union[str, Dict[str, Any]]) -> Optional[str]:
+        """Extracts content from DeepInfra stream JSON objects."""
+        if isinstance(chunk, dict):
+            return chunk.get("choices", [{}])[0].get("delta", {}).get("content")
+        return None
+
     def __init__(
         self,
         is_conversation: bool = True,
-        max_tokens: int = 2049,  # Set a reasonable default
+        max_tokens: int = 2049,
         timeout: int = 30,
         intro: str = None,
         filepath: str = None,
@@ -82,43 +93,39 @@ class DeepInfra(Provider):
         act: str = None,
         model: str = "meta-llama/Llama-3.3-70B-Instruct-Turbo",
         system_prompt: str = "You are a helpful assistant.",
-        browser: str = "chrome"
+        browser: str = "chrome" # Note: browser fingerprinting might be less effective with impersonate
     ):
         """Initializes the DeepInfra API client."""
         if model not in self.AVAILABLE_MODELS:
             raise ValueError(f"Invalid model: {model}. Choose from: {self.AVAILABLE_MODELS}")
-            
+
         self.url = "https://api.deepinfra.com/v1/openai/chat/completions"
-        
-        # Initialize LitAgent for user agent generation
+
+        # Initialize LitAgent (keep if needed for other headers or logic)
         self.agent = LitAgent()
-        # Use fingerprinting to create a consistent browser identity
+        # Fingerprint generation might be less relevant with impersonate
         self.fingerprint = self.agent.generate_fingerprint(browser)
-        
-        # Use the fingerprint for headers
+
+        # Use the fingerprint for headers (keep relevant ones)
         self.headers = {
-            "Accept": self.fingerprint["accept"],
-            "Accept-Encoding": "gzip, deflate, br, zstd",
-            "Accept-Language": self.fingerprint["accept_language"],
+            "Accept": self.fingerprint["accept"], # Keep Accept
+            "Accept-Language": self.fingerprint["accept_language"], # Keep Accept-Language
             "Content-Type": "application/json",
-            "Cache-Control": "no-cache",
-            "Connection": "keep-alive",
-            "Origin": "https://deepinfra.com",
-            "Pragma": "no-cache",
-            "Referer": "https://deepinfra.com/",
-            "Sec-Fetch-Dest": "empty",
+            "Cache-Control": "no-cache", # Keep Cache-Control
+            "Origin": "https://deepinfra.com", # Keep Origin
+            "Pragma": "no-cache", # Keep Pragma
+            "Referer": "https://deepinfra.com/", # Keep Referer
+            "Sec-Fetch-Dest": "empty", # Keep Sec-Fetch-*
             "Sec-Fetch-Mode": "cors",
             "Sec-Fetch-Site": "same-site",
-            "X-Deepinfra-Source": "web-embed",
-            "Sec-CH-UA": self.fingerprint["sec_ch_ua"] or '"Not)A;Brand";v="99", "Microsoft Edge";v="127", "Chromium";v="127"',
-            "Sec-CH-UA-Mobile": "?0",
-            "Sec-CH-UA-Platform": f'"{self.fingerprint["platform"]}"',
-            "User-Agent": self.fingerprint["user_agent"],
+            "X-Deepinfra-Source": "web-embed", # Keep custom headers
         }
-        
-        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.system_prompt = system_prompt
         self.is_conversation = is_conversation
         self.max_tokens_to_sample = max_tokens
@@ -147,32 +154,28 @@ class DeepInfra(Provider):
     def refresh_identity(self, browser: str = None):
         """
         Refreshes the browser identity fingerprint.
-        
+
         Args:
             browser: Specific browser to use for the new fingerprint
         """
         browser = browser or self.fingerprint.get("browser_type", "chrome")
         self.fingerprint = self.agent.generate_fingerprint(browser)
-        
-        # Update headers with new fingerprint
+
+        # Update headers with new fingerprint (only relevant ones)
         self.headers.update({
             "Accept": self.fingerprint["accept"],
             "Accept-Language": self.fingerprint["accept_language"],
-            "Sec-CH-UA": self.fingerprint["sec_ch_ua"] or self.headers["Sec-CH-UA"],
-            "Sec-CH-UA-Platform": f'"{self.fingerprint["platform"]}"',
-            "User-Agent": self.fingerprint["user_agent"],
         })
-        
+
         # Update session headers
-        for header, value in self.headers.items():
-            self.session.headers[header] = value
-        
+        self.session.headers.update(self.headers) # Update only relevant headers
+
         return self.fingerprint
 
     def ask(
         self,
         prompt: str,
-        stream: bool = False,
+        stream: bool = False, # API supports streaming
         raw: bool = False,
         optimizer: str = None,
         conversationally: bool = False,
@@ -193,61 +196,88 @@ class DeepInfra(Provider):
                 {"role": "system", "content": self.system_prompt},
                 {"role": "user", "content": conversation_prompt},
             ],
-            "stream": stream
+            "stream": stream # Pass stream argument to payload
         }
 
         def for_stream():
+            streaming_text = "" # Initialize outside try block
             try:
-                with requests.post(self.url, headers=self.headers, data=json.dumps(payload), 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(decode_unicode=True):
-                        if line:
-                            line = line.strip()
-                            if line.startswith("data: "):
-                                json_str = line[6:]
-                                if json_str == "[DONE]":
-                                    break
-                                try:
-                                    json_data = json.loads(json_str)
-                                    if 'choices' in json_data:
-                                        choice = json_data['choices'][0]
-                                        if 'delta' in choice and 'content' in choice['delta']:
-                                            content = choice['delta']['content']
-                                            streaming_text += content
-                                            resp = dict(text=content)
-                                            yield resp if raw else resp
-                                except json.JSONDecodeError:
-                                    continue
-                    
+                # Use curl_cffi session post with impersonate
+                response = self.session.post(
+                    self.url,
+                    # headers are set on the session
+                    data=json.dumps(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="data:",
+                    to_json=True,     # Stream sends JSON
+                    skip_markers=["[DONE]"],
+                    content_extractor=self._deepinfra_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 _deepinfra_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
+
+            except CurlError as e:
+                raise exceptions.FailedToGenerateResponseError(f"Request failed (CurlError): {str(e)}") from e
+            except Exception as e:
+                raise exceptions.FailedToGenerateResponseError(f"Request failed ({type(e).__name__}): {str(e)}") from e
+            finally:
+                # Update history after stream finishes or fails
+                if streaming_text:
                     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)}")
+
 
         def for_non_stream():
             try:
-                response = requests.post(self.url, headers=self.headers, data=json.dumps(payload), timeout=self.timeout)
-                if response.status_code != 200:
-                    raise exceptions.FailedToGenerateResponseError(
-                        f"Request failed with status code {response.status_code}"
-                    )
-
-                response_data = response.json()
-                if 'choices' in response_data and len(response_data['choices']) > 0:
-                    content = response_data['choices'][0].get('message', {}).get('content', '')
-                    self.last_response = {"text": content}
-                    self.conversation.update_chat_history(prompt, content)
-                    return {"text": content}
-                else:
-                    raise exceptions.FailedToGenerateResponseError("No response content found")
-            except Exception as e:
-                raise exceptions.FailedToGenerateResponseError(f"Request failed: {e}")
+                # Use curl_cffi session post with impersonate for non-streaming
+                response = self.session.post(
+                    self.url,
+                    # headers are set on the session
+                    data=json.dumps(payload),
+                    timeout=self.timeout,
+                    impersonate="chrome110" # Use a common impersonation profile
+                )
+                response.raise_for_status() # Check for HTTP errors
+
+                response_text = response.text # Get raw text
+
+                # Use sanitize_stream to parse the 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 = next(processed_stream, None)
+                content = content if isinstance(content, str) else "" # Ensure it's a string
+
+                self.last_response = {"text": content}
+                self.conversation.update_chat_history(prompt, content)
+                return self.last_response if not raw else content # Return dict or raw string
+
+            except CurlError as e: # Catch CurlError
+                raise exceptions.FailedToGenerateResponseError(f"Request failed (CurlError): {e}") from e
+            except Exception as e: # Catch other potential exceptions (like HTTPError, JSONDecodeError)
+                err_text = getattr(e, 'response', None) and getattr(e.response, 'text', '')
+                raise exceptions.FailedToGenerateResponseError(f"Request failed ({type(e).__name__}): {e} - {err_text}") from e
+
 
         return for_stream() if stream else for_non_stream()
 
@@ -258,20 +288,31 @@ class DeepInfra(Provider):
         optimizer: str = None,
         conversationally: bool = False,
     ) -> Union[str, Generator[str, None, None]]:
-        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
             )
-        return for_stream() if stream else for_non_stream()
+            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_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)
@@ -283,7 +324,7 @@ if __name__ == "__main__":
             response_text = ""
             for chunk in response:
                 response_text += chunk
-            
+
             if response_text and len(response_text.strip()) > 0:
                 status = "✓"
                 # Clean and truncate response
@@ -294,4 +335,4 @@ 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)}")