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

webscout/Provider/PI.py

@@ -5,7 +5,7 @@ import json
 import re
 import threading
 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 typing import Dict, Union, Any, Optional
@@ -122,6 +122,13 @@ class PiAI(Provider):
         if self.is_conversation:
             self.start_conversation()
 
+    @staticmethod
+    def _pi_extractor(chunk: Union[str, Dict[str, Any]]) -> Optional[str]:
+        """Extracts text content from PiAI stream JSON objects."""
+        if isinstance(chunk, dict) and 'text' in chunk and chunk['text'] is not None:
+            return chunk.get("text")
+        return None
+
     def start_conversation(self) -> str:
         """
         Initializes a new conversation and returns the conversation ID.
@@ -245,17 +252,22 @@ class PiAI(Provider):
                     if line_bytes:
                         line = line_bytes.decode('utf-8')
                         full_raw_data_for_sids += line + "\n" # Accumulate for SID extraction
+                        
                         if line.startswith("data: "):
+                            json_line_str = line[6:] # Get the JSON part as string
                             try:
-                                parsed_data = json.loads(line[6:])
-                                if 'text' in parsed_data and parsed_data['text'] is not None:
-                                    chunk_text = parsed_data['text']
+                                # Process this single JSON line string with sanitize_stream
+                                processed_gen = sanitize_stream(
+                                    data=json_line_str,
+                                    to_json=True,
+                                    content_extractor=self._pi_extractor
+                                )
+                                chunk_text = next(processed_gen, None) # Get the single extracted text item
+                                if chunk_text and isinstance(chunk_text, str):
                                     streaming_text += chunk_text
-                                    # Yield raw JSON object or dict with aggregated text
-                                    yield parsed_data if raw else dict(text=streaming_text)
-                            except (json.JSONDecodeError, UnicodeDecodeError):
-                                continue
-
+                                    yield {"text": streaming_text} # Always yield dict with aggregated text
+                            except (StopIteration, json.JSONDecodeError, UnicodeDecodeError):
+                                continue # Skip if sanitize_stream fails or yields nothing
                 # Extract SIDs after processing the stream
                 sids = re.findall(r'"sid":"(.*?)"', full_raw_data_for_sids)
                 second_sid = sids[1] if len(sids) >= 2 else None
@@ -284,13 +296,10 @@ class PiAI(Provider):
         else:
             # For non-stream, collect all responses and return the final one
             final_text = ""
-            # Ensure raw=False so process_stream yields dicts
+            # process_stream always yields dicts now
             for res in process_stream():
                  if isinstance(res, dict) and "text" in res:
                      final_text = res["text"] # Keep updating with the latest aggregated text
-                 # Handle raw JSON object case if raw=True was passed
-                 elif raw and isinstance(res, dict) and 'text' in res and res['text'] is not None:
-                     final_text += res['text'] # Append chunks if raw
 
             # last_response and history are updated within process_stream
             # Return the final aggregated response dict or raw text

webscout/Provider/StandardInput.py

@@ -1,13 +1,10 @@
-from os import system
-import requests
-import json
+from curl_cffi.requests import Session 
 import uuid
 import re
-from datetime import datetime
-from typing import Any, Dict, Optional, Union, Generator
+from typing import Any, Dict, Optional, 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
@@ -98,9 +95,9 @@ class StandardInputAI(Provider):
             "ph_phc_f3wUUyCfmKlKtkc2pfT7OsdcW2mBEVGN2A87yEYbG3c_posthog": '''%7B%22distinct_id%22%3A%220195c7cc-ac8f-79ff-b901-e14a78fc2a67%22%2C%22%24sesid%22%3A%5B1744688627860%2C%220196377f-9f12-77e6-a9ea-0e9669423803%22%2C1744687832850%5D%2C%22%24initial_person_info%22%3A%7B%22r%22%3A%22%24direct%22%2C%22u%22%3A%22https%3A%2F%2Fstandard-input.com%2F%22%7D%7D'''
         }
 
-        self.session = requests.Session()
+        self.session = Session() # Use curl_cffi Session
         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
@@ -153,6 +150,17 @@ class StandardInputAI(Provider):
 
         return self.fingerprint
 
+    @staticmethod
+    def _standardinput_extractor(chunk: Union[str, Dict[str, Any]]) -> Optional[str]:
+        """Extracts content from the StandardInput 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,
@@ -183,45 +191,48 @@ class StandardInputAI(Provider):
         }
 
         try:
-            response = self.session.post(self.url, cookies=self.cookies, json=payload, stream=True, timeout=self.timeout)
+            # Use curl_cffi post with impersonate
+            response = self.session.post(
+                self.url,
+                cookies=self.cookies,
+                json=payload,
+                stream=True,
+                timeout=self.timeout,
+                impersonate="chrome120" # Add impersonate
+            )
+
             if response.status_code != 200:
-                # Try to get response content for better error messages
                 try:
                     error_content = response.text
                 except:
                     error_content = "<could not read response content>"
 
                 if response.status_code in [403, 429]:
-                    print(f"Received status code {response.status_code}, refreshing identity...")
                     self.refresh_identity()
-                    response = self.session.post(self.url, cookies=self.cookies, json=payload, stream=True, timeout=self.timeout)
+                    response = self.session.post(
+                        self.url, cookies=self.cookies, json=payload, stream=True,
+                        timeout=self.timeout, impersonate="chrome120"
+                    )
                     if not response.ok:
                         raise exceptions.FailedToGenerateResponseError(
                             f"Failed to generate response after identity refresh - ({response.status_code}, {response.reason}) - {error_content}"
                         )
-                    print("Identity refreshed successfully.")
                 else:
                     raise exceptions.FailedToGenerateResponseError(
                         f"Request failed with status code {response.status_code}. Response: {error_content}"
                     )
 
             full_response = ""
-            debug_lines = []
-
-            # Process the streaming response
-            for i, line in enumerate(response.iter_lines(decode_unicode=True)):
-                if line:
-                    try:
-                        line_str = line
-                        debug_lines.append(line_str)
-
-                        # Extract content from the response
-                        match = re.search(r'0:"(.*?)"', line_str)
-                        if match:
-                            content = match.group(1)
-                            full_response += content
-                            continue
-                    except: pass
+            # Use sanitize_stream
+            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._standardinput_extractor # Use the specific extractor
+            )
+            for content_chunk in processed_stream:
+                if content_chunk and isinstance(content_chunk, str):
+                    full_response += content_chunk
 
             self.last_response = {"text": full_response}
             self.conversation.update_chat_history(prompt, full_response)
@@ -243,7 +254,8 @@ class StandardInputAI(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')
+        # Extractor handles formatting
+        return response.get("text", "").replace('\\n', '\n').replace('\\n\\n', '\n\n')
 
 if __name__ == "__main__":
     print("-" * 100)

webscout/Provider/TeachAnything.py

@@ -4,7 +4,7 @@ from typing import Union, Any, Dict
 from webscout.AIbase import Provider # Import Provider base class
 from webscout import exceptions # Import custom exceptions
 from webscout.conversation import Conversation
-from webscout.optimizers import Optimizers
+from webscout.AIutel import Optimizers, sanitize_stream # Import sanitize_stream
 from webscout.prompt_manager import AwesomePrompts
 from webscout.litagent import LitAgent
 
@@ -131,13 +131,22 @@ class TeachAnything(Provider):
             )
             response.raise_for_status() # Check for HTTP errors
 
-            # Use response.text which is already decoded
-            resp_text = response.text
-            # The response is plain text, wrap it in the expected dict format
-            self.last_response = {"text": resp_text}
-            self.conversation.update_chat_history(
-                prompt, resp_text
+            resp_text_raw = response.text # Get raw response text
+
+            # Process the text using sanitize_stream (even though it's not streaming)
+            # This keeps the pattern consistent, though it won't do much here
+            processed_stream = sanitize_stream(
+                data=resp_text_raw,
+                intro_value=None, # No prefix
+                to_json=False     # It's plain text
             )
+
+            # Extract the single result from the generator
+            resp_text = "".join(list(processed_stream)) # Aggregate potential chunks (should be one)
+
+            self.last_response = {"text": resp_text}
+            self.conversation.update_chat_history(prompt, resp_text)
+
             # Return dict or raw string based on raw flag
             return resp_text if raw else self.last_response
 

webscout/Provider/TextPollinationsAI.py

@@ -2,7 +2,9 @@ from curl_cffi.requests import Session
 from curl_cffi import CurlError
 import json
 from typing import Union, Any, Dict, Generator, Optional, List
-from webscout.AIutel import Optimizers, Conversation, AwesomePrompts
+
+import requests
+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
@@ -13,32 +15,30 @@ class TextPollinationsAI(Provider):
     """
 
     AVAILABLE_MODELS = [
-        "openai",              # OpenAI GPT-4.1-nano (Azure) - vision capable
-        "openai-large",        # OpenAI GPT-4.1 mini (Azure) - vision capable
-        "openai-reasoning",    # OpenAI o4-mini (Azure) - vision capable, reasoning
-        "qwen-coder",          # Qwen 2.5 Coder 32B (Scaleway)
-        "llama",               # Llama 3.3 70B (Cloudflare)
-        "llamascout",          # Llama 4 Scout 17B (Cloudflare)
-        "mistral",             # Mistral Small 3 (Scaleway) - vision capable
-        "unity",               # Unity Mistral Large (Scaleway) - vision capable, uncensored
-        "midijourney",         # Midijourney (Azure)
-        "rtist",               # Rtist (Azure)
-        "searchgpt",           # SearchGPT (Azure) - vision capable
-        "evil",                # Evil (Scaleway) - vision capable, uncensored
-        "deepseek-reasoning",  # DeepSeek-R1 Distill Qwen 32B (Cloudflare) - reasoning
-        "deepseek-reasoning-large", # DeepSeek R1 - Llama 70B (Scaleway) - reasoning
-        "phi",                 # Phi-4 Instruct (Cloudflare) - vision and audio capable
-        "llama-vision",        # Llama 3.2 11B Vision (Cloudflare) - vision capable
-        "gemini",              # gemini-2.5-flash-preview-04-17 (Azure) - vision and audio capable
-        "hormoz",              # Hormoz 8b (Modal)
-        "hypnosis-tracy",      # Hypnosis Tracy 7B (Azure) - audio capable
-        "deepseek",            # DeepSeek-V3 (DeepSeek)
-        "sur",                 # Sur AI Assistant (Mistral) (Scaleway) - vision capable
-        "openai-audio",        # OpenAI GPT-4o-audio-preview (Azure) - vision and audio capable
+        "openai",
+        "openai-large",
+        "qwen-coder",
+        "llama",
+        "llamascout",
+        "mistral",
+        "unity",
+        "midijourney",
+        "rtist",
+        "searchgpt",
+        "evil",
+        "deepseek-reasoning",
+        "deepseek-reasoning-large",
+        "phi",
+        "llama-vision",
+        "hormoz",
+        "hypnosis-tracy",
+        "deepseek",
+        "sur",
+        "openai-audio",
     ]
+    _models_url = "https://text.pollinations.ai/models"
 
-    def __init__(
-        self,
+    def __init__(self,
         is_conversation: bool = True,
         max_tokens: int = 8096, # Note: max_tokens is not directly used by this API endpoint
         timeout: int = 30,
@@ -52,10 +52,6 @@ class TextPollinationsAI(Provider):
         system_prompt: str = "You are a helpful AI assistant.",
     ):
         """Initializes the TextPollinationsAI API client."""
-        if model not in self.AVAILABLE_MODELS:
-            raise ValueError(f"Invalid model: {model}. Choose from: {self.AVAILABLE_MODELS}")
-
-        # Initialize curl_cffi Session
         self.session = Session()
         self.is_conversation = is_conversation
         self.max_tokens_to_sample = max_tokens
@@ -66,6 +62,10 @@ class TextPollinationsAI(Provider):
         self.model = model
         self.system_prompt = system_prompt
 
+        # Validate against the hardcoded list
+        if model not in self.AVAILABLE_MODELS:
+            raise ValueError(f"Invalid model: {model}. Choose from: {self.AVAILABLE_MODELS}")
+
         self.headers = {
             'Accept': '*/*',
             'Accept-Language': 'en-US,en;q=0.9',
@@ -96,6 +96,7 @@ class TextPollinationsAI(Provider):
         )
         self.conversation.history_offset = history_offset
 
+
     def ask(
         self,
         prompt: str,
@@ -148,40 +149,34 @@ class TextPollinationsAI(Provider):
                         f"Failed to generate response - ({response.status_code}, {response.reason}) - {response.text}"
                     )
 
-                full_response = ""
-                # Iterate over bytes and decode manually
-                for line_bytes in response.iter_lines():
-                    if line_bytes:
-                        line = line_bytes.decode('utf-8').strip()
-                        if line == "data: [DONE]":
-                            break
-                        if line.startswith('data: '):
-                            try:
-                                json_data = json.loads(line[6:])
-                                if 'choices' in json_data and len(json_data['choices']) > 0:
-                                    choice = json_data['choices'][0]
-                                    if 'delta' in choice:
-                                        if 'content' in choice['delta'] and choice['delta']['content'] is not None:
-                                            content = choice['delta']['content']
-                                            full_response += content
-                                            # Yield dict or raw string
-                                            yield content if raw else dict(text=content)
-                                        elif 'tool_calls' in choice['delta']:
-                                            # Handle tool calls in streaming response
-                                            tool_calls = choice['delta']['tool_calls']
-                                            # Yield dict or raw list
-                                            yield tool_calls if raw else dict(tool_calls=tool_calls)
-                            except json.JSONDecodeError:
-                                continue
-                            except UnicodeDecodeError:
-                                continue
+                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]"],
+                    # Extractor handles both content and tool_calls
+                    content_extractor=lambda chunk: chunk.get('choices', [{}])[0].get('delta') if isinstance(chunk, dict) else None,
+                    yield_raw_on_error=False # Skip non-JSON or lines where extractor fails
+                )
+
+                for delta in processed_stream:
+                    # delta is the extracted 'delta' object or None
+                    if delta and isinstance(delta, dict):
+                        if 'content' in delta and delta['content'] is not None:
+                            content = delta['content']
+                            streaming_text += content
+                            yield content if raw else dict(text=content)
+                        elif 'tool_calls' in delta:
+                            tool_calls = delta['tool_calls']
+                            yield tool_calls if raw else dict(tool_calls=tool_calls)
 
                 # Update history and last response after stream finishes
-                # Note: last_response might only contain text, not tool calls if they occurred
-                self.last_response.update(dict(text=full_response))
-                if full_response: # Only update history if text was received
+                self.last_response.update(dict(text=streaming_text)) # Store aggregated text
+                if streaming_text: # Only update history if text was received
                     self.conversation.update_chat_history(
-                        prompt, full_response # Use the fully aggregated text
+                        prompt, streaming_text # Use the fully aggregated text
                     )
             except CurlError as e: # Catch CurlError
                 raise exceptions.FailedToGenerateResponseError(f"Request failed (CurlError): {e}") from e
@@ -193,22 +188,27 @@ class TextPollinationsAI(Provider):
             # Aggregate the stream using the updated for_stream logic
             final_content = ""
             tool_calls_aggregated = None # To store potential tool calls
-            for chunk_data in for_stream():
-                if isinstance(chunk_data, dict):
-                    if "text" in chunk_data:
-                        final_content += chunk_data["text"]
-                    elif "tool_calls" in chunk_data:
-                        # Aggregate tool calls (simple aggregation, might need refinement)
-                        if tool_calls_aggregated is None:
-                            tool_calls_aggregated = []
-                        tool_calls_aggregated.extend(chunk_data["tool_calls"])
-                elif isinstance(chunk_data, str): # Handle raw stream case
-                    final_content += chunk_data
-                # Handle raw tool calls list if raw=True
-                elif isinstance(chunk_data, list) and raw:
-                     if tool_calls_aggregated is None:
-                         tool_calls_aggregated = []
-                     tool_calls_aggregated.extend(chunk_data)
+            try: # Add try block for potential errors during aggregation
+                for chunk_data in for_stream():
+                    if isinstance(chunk_data, dict):
+                        if "text" in chunk_data:
+                            final_content += chunk_data["text"]
+                        elif "tool_calls" in chunk_data:
+                            # Aggregate tool calls (simple aggregation, might need refinement)
+                            if tool_calls_aggregated is None:
+                                tool_calls_aggregated = []
+                            tool_calls_aggregated.extend(chunk_data["tool_calls"])
+                    elif isinstance(chunk_data, str): # Handle raw stream case
+                        final_content += chunk_data
+                    # Handle raw tool calls list if raw=True
+                    elif isinstance(chunk_data, list) and raw:
+                         if tool_calls_aggregated is None:
+                             tool_calls_aggregated = []
+                         tool_calls_aggregated.extend(chunk_data)
+            except Exception as e:
+                 # If aggregation fails but some text was received, use it. Otherwise, re-raise.
+                 if not final_content and not tool_calls_aggregated:
+                     raise exceptions.FailedToGenerateResponseError(f"Failed to get non-stream response: {str(e)}") from e
 
 
             # last_response and history are updated within for_stream (for text)
@@ -263,6 +263,7 @@ class TextPollinationsAI(Provider):
         elif "tool_calls" in response:
             # For tool calls, return a string representation
             return json.dumps(response["tool_calls"])
+        return "" # Return empty string if neither text nor tool_calls found
 
 if __name__ == "__main__":
     # Ensure curl_cffi is installed
@@ -274,6 +275,7 @@ if __name__ == "__main__":
     working = 0
     total = len(TextPollinationsAI.AVAILABLE_MODELS)
 
+
     for model in TextPollinationsAI.AVAILABLE_MODELS:
         try:
             test_ai = TextPollinationsAI(model=model, timeout=60)