webscout 8.3.2__py3-none-any.whl → 8.3.3__py3-none-any.whl

webscout/Provider/llmchatco.py

@@ -129,8 +129,8 @@ class LLMChatCo(Provider):
         optimizer: str = None,
         conversationally: bool = False,
         web_search: bool = False,
-    ) -> Union[Dict[str, Any], Generator[Any, None, None]]:
-        """Chat with LLMChat.co with streaming capabilities"""
+    ) -> Union[Dict[str, Any], Generator[Any, None, None], str]:
+        """Chat with LLMChat.co with streaming capabilities and raw output support using sanitize_stream."""
 
         conversation_prompt = self.conversation.gen_complete_prompt(prompt)
         if optimizer:
@@ -143,7 +143,6 @@ class LLMChatCo(Provider):
                     f"Optimizer is not one of {self.__available_optimizers}"
                 )
 
-
         # Generate a unique ID for this message
         thread_item_id = ''.join(str(uuid.uuid4()).split('-'))[:20]
         messages = [
@@ -164,79 +163,59 @@ class LLMChatCo(Provider):
         }
 
         def for_stream():
-            full_response = "" # Initialize outside try block
+            full_response = ""
             try:
-                # Use curl_cffi session post with impersonate
                 response = self.session.post(
                     self.api_endpoint,
                     json=payload,
-                    # headers are set on the session
                     stream=True,
                     timeout=self.timeout,
-                    # proxies are set on the session
-                    impersonate="chrome110" # Use a common impersonation profile
+                    impersonate="chrome110"
                 )
-                response.raise_for_status() # Check for HTTP errors
+                response.raise_for_status()
 
-                # Use sanitize_stream
-                # Note: This won't handle SSE 'event:' lines, only 'data:' lines.
-                # The original code checked for event == 'answer'. We assume relevant data is JSON after 'data:'.
                 processed_stream = sanitize_stream(
-                    data=response.iter_content(chunk_size=None), # Pass byte iterator
+                    data=response.iter_content(chunk_size=None),
                     intro_value="data:",
-                    to_json=True,     # Stream sends JSON
-                    content_extractor=self._llmchatco_extractor, # Use the specific extractor
-                    yield_raw_on_error=False # Skip non-JSON lines or lines where extractor fails
+                    to_json=True,
+                    content_extractor=self._llmchatco_extractor,
+                    yield_raw_on_error=False,
+                    raw=raw
                 )
 
                 last_yielded_text = ""
                 for current_full_text in processed_stream:
-                    # current_full_text is the full text extracted by _llmchatco_extractor
                     if current_full_text and isinstance(current_full_text, str):
-                        # Calculate the new part of the text
                         new_text = current_full_text[len(last_yielded_text):]
                         if new_text:
-                            full_response = current_full_text # Keep track of the latest full text
-                            last_yielded_text = current_full_text # Update tracker
-                            resp = dict(text=new_text)
-                            # Yield dict or raw string chunk
-                            yield resp if not raw else new_text
-
-                # Update history after stream finishes
+                            full_response = current_full_text
+                            last_yielded_text = current_full_text
+                            if raw:
+                                yield new_text
+                            else:
+                                yield dict(text=new_text)
                 self.last_response = dict(text=full_response)
                 self.last_assistant_response = full_response
                 self.conversation.update_chat_history(
                     prompt, full_response
                 )
-
-            except CurlError as e: # Catch CurlError
+            except CurlError as e:
                 raise exceptions.FailedToGenerateResponseError(f"Request failed (CurlError): {e}") from e
-            except Exception as e: # Catch other potential exceptions (like HTTPError)
+            except Exception as e:
                 err_text = getattr(e, 'response', None) and getattr(e.response, 'text', '')
                 raise exceptions.FailedToGenerateResponseError(f"Unexpected error ({type(e).__name__}): {str(e)} - {err_text}") from e
-
         def for_non_stream():
-            # Aggregate the stream using the updated for_stream logic
             full_response_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:
+                    if raw and isinstance(chunk_data, str):
+                        full_response_text += chunk_data
+                    elif isinstance(chunk_data, dict) and "text" in chunk_data:
                         full_response_text += chunk_data["text"]
-                    # Handle raw string case if raw=True was passed
-                    elif raw and isinstance(chunk_data, str):
-                         full_response_text += chunk_data
-
             except Exception as e:
-                # If aggregation fails but some text was received, use it. Otherwise, re-raise.
                 if not full_response_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_response_text if raw else self.last_response
-
-
         return for_stream() if stream else for_non_stream()
 
     def chat(
@@ -246,31 +225,33 @@ class LLMChatCo(Provider):
         optimizer: str = None,
         conversationally: bool = False,
         web_search: bool = False,
+        raw: bool = False
     ) -> Union[str, Generator[str, None, None]]:
-        """Generate response with streaming capabilities"""
-
+        """Generate response with streaming capabilities and raw output support"""
         def for_stream_chat():
-            # ask() yields dicts or strings when streaming
             gen = self.ask(
-                prompt, stream=True, raw=False, # Ensure ask yields dicts
+                prompt, stream=True, raw=raw,
                 optimizer=optimizer, conversationally=conversationally,
                 web_search=web_search
             )
-            for response_dict in gen:
-                yield self.get_message(response_dict) # get_message expects dict
-
+            for response in gen:
+                if raw:
+                    yield response
+                else:
+                    yield self.get_message(response)
         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
+                raw=raw,
                 optimizer=optimizer,
                 conversationally=conversationally,
                 web_search=web_search
             )
-            return self.get_message(response_data) # get_message expects dict
-
+            if raw:
+                return response_data if isinstance(response_data, str) else self.get_message(response_data)
+            else:
+                return self.get_message(response_data)
         return for_stream_chat() if stream else for_non_stream_chat()
 
     def get_message(self, response: Dict[str, Any]) -> str:
@@ -279,28 +260,32 @@ class LLMChatCo(Provider):
         return response["text"]
 
 if __name__ == "__main__":
-    # Ensure curl_cffi is installed
-    print("-" * 80)
-    print(f"{'Model':<50} {'Status':<10} {'Response'}")
-    print("-" * 80)
-
-    # Test all available models
-    working = 0
-    total = len(LLMChatCo.AVAILABLE_MODELS)
-
-    for model in LLMChatCo.AVAILABLE_MODELS:
-        try:
-            test_ai = LLMChatCo(model=model, timeout=60)
-            response = test_ai.chat("Say 'Hello' in one word")
-            response_text = response
-
-            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()
-            else:
-                status = "✗"
-                display_text = "Empty or invalid response"
-            print(f"{model:<50} {status:<10} {display_text}")
-        except Exception as e:
-            print(f"{model:<50} {'✗':<10} {str(e)}")
+    # # Ensure curl_cffi is installed
+    # print("-" * 80)
+    # print(f"{'Model':<50} {'Status':<10} {'Response'}")
+    # print("-" * 80)
+
+    # # Test all available models
+    # working = 0
+    # total = len(LLMChatCo.AVAILABLE_MODELS)
+
+    # for model in LLMChatCo.AVAILABLE_MODELS:
+    #     try:
+    #         test_ai = LLMChatCo(model=model, timeout=60)
+    #         response = test_ai.chat("Say 'Hello' in one word")
+    #         response_text = response
+
+    #         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()
+    #         else:
+    #             status = "✗"
+    #             display_text = "Empty or invalid response"
+    #         print(f"{model:<50} {status:<10} {display_text}")
+    #     except Exception as e:
+    #         print(f"{model:<50} {'✗':<10} {str(e)}")
+    ai = LLMChatCo()
+    response = ai.chat("yooo", stream=True, raw=False)
+    for chunk in response:
+        print(chunk, end="", flush=True)

webscout/Provider/multichat.py

@@ -2,7 +2,7 @@ from curl_cffi.requests import Session
 from curl_cffi import CurlError
 import json
 import uuid
-from typing import Any, Dict, Union
+from typing import Any, Dict, Union, Generator
 from datetime import datetime
 from webscout.AIutel import Optimizers, Conversation, AwesomePrompts, sanitize_stream # Import sanitize_stream
 from webscout.AIbase import Provider
@@ -257,10 +257,9 @@ class MultiChatAI(Provider):
         raw: bool = False, # Keep raw param for interface consistency
         optimizer: str = None,
         conversationally: bool = False,
-        # Add stream parameter for consistency, though API doesn't stream
         stream: bool = False 
-    ) -> Dict[str, Any]:
-        """Sends a prompt to the MultiChatAI API and returns the response."""
+    ) -> Union[Dict[str, Any], str, Generator[str, None, None]]:
+        """Sends a prompt to the MultiChatAI API and returns the response. Supports raw output and direct text streaming."""
         conversation_prompt = self.conversation.gen_complete_prompt(prompt)
         if optimizer:
             if optimizer in self.__available_optimizers:
@@ -275,26 +274,32 @@ class MultiChatAI(Provider):
             "customModelId": "",
         }
 
-        # API does not stream, implement non-stream logic directly
         response = self._make_request(payload)
         try:
-            # Use response.text which is already decoded
-            response_text_raw = response.text # Get raw text
-
-            # Process the text using sanitize_stream (even though it's not streaming)
-            processed_stream = sanitize_stream(
-                data=response_text_raw,
-                intro_value=None, # No prefix
-                to_json=False     # It's plain text
-            )
-            # Aggregate the single result
-            full_response = "".join(list(processed_stream)).strip()
-
-            self.last_response = {"text": full_response} # Store processed text
-            self.conversation.update_chat_history(prompt, full_response)
-            # Return dict or raw string based on raw flag
-            return full_response if raw else self.last_response
-        except Exception as e: # Catch potential errors during text processing
+            response_text_raw = response.text
+            if stream:
+                chunk_size = 64
+                text = response_text_raw
+                for i in range(0, len(text), chunk_size):
+                    chunk = text[i:i+chunk_size]
+                    if raw:
+                        yield chunk
+                    else:
+                        yield {"text": chunk}
+                self.last_response = {"text": text}
+                self.conversation.update_chat_history(prompt, text)
+            else:
+                processed_stream = sanitize_stream(
+                    data=response_text_raw,
+                    intro_value=None,
+                    to_json=False,
+                    raw=raw
+                )
+                full_response = "".join(list(processed_stream)).strip()
+                self.last_response = {"text": full_response}
+                self.conversation.update_chat_history(prompt, full_response)
+                return full_response if raw else self.last_response
+        except Exception as e:
             raise exceptions.FailedToGenerateResponseError(f"Failed to process response: {e}") from e
 
     def chat(
@@ -302,26 +307,32 @@ class MultiChatAI(Provider):
         prompt: str,
         optimizer: str = None,
         conversationally: bool = False,
-        # Add stream parameter for consistency
-        stream: bool = False
-    ) -> str:
-        """Generate response."""
-        # Since ask() now handles both stream=True/False by returning the full response dict/str:
-        response_data = self.ask(
-            prompt, 
-            stream=False, # Call ask in non-stream mode internally
-            raw=False, # Ensure ask returns dict
-            optimizer=optimizer, 
-            conversationally=conversationally
-        )
-        # If stream=True was requested, simulate streaming by yielding the full message at once
+        stream: bool = False,
+        raw: bool = False
+    ) -> Union[str, Generator[str, None, None]]:
+        """Generate response. Supports raw output and streaming."""
         if stream:
-            def stream_wrapper():
-                yield self.get_message(response_data)
-            return stream_wrapper()
+            # Streaming mode: yield chunks from ask
+            return self.ask(
+                prompt,
+                raw=raw,
+                optimizer=optimizer,
+                conversationally=conversationally,
+                stream=True
+            )
         else:
-            # If stream=False, return the full message directly
-            return self.get_message(response_data)
+            # Non-streaming mode: return full message
+            response_data = self.ask(
+                prompt,
+                raw=raw,
+                optimizer=optimizer,
+                conversationally=conversationally,
+                stream=False
+            )
+            if raw:
+                return response_data if isinstance(response_data, str) else self.get_message(response_data)
+            else:
+                return self.get_message(response_data)
 
     def get_message(self, response: Union[Dict[str, Any], str]) -> str:
         """

webscout/Provider/oivscode.py

@@ -303,7 +303,7 @@ class oivscode(Provider):
 if __name__ == "__main__":
     from rich import print
     chatbot = oivscode()
-    print(chatbot.fetch_available_models())
+    chatbot.fetch_available_models()
     response = chatbot.chat(input(">>> "), stream=True)
     for chunk in response:
         print(chunk, end="", flush=True)

webscout/Provider/scira_chat.py

@@ -154,14 +154,20 @@ class SciraAI(Provider):
         return self.fingerprint
 
     @staticmethod
-    def _scira_extractor(chunk: Union[str, Dict[str, Any]]) -> Optional[str]:
-        """Extracts content from the Scira stream format '0:"..."'."""
+    def _scira_extractor(chunk: Union[str, Dict[str, Any]]) -> Optional[dict]:
+        """Extracts g and 0 chunks from the Scira stream format.
+        Returns a dict: {"g": [g1, g2, ...], "0": zero} if present.
+        """
         if isinstance(chunk, str):
-            match = re.search(r'0:"(.*?)"(?=,|$)', chunk) # Look for 0:"...", possibly followed by comma or end of string
-            if match:
-                # Decode potential unicode escapes like \u00e9 and handle escaped quotes/backslashes
-                content = match.group(1).encode().decode('unicode_escape')
-                return content.replace('\\\\', '\\').replace('\\"', '"')
+            g_matches = re.findall(r'g:"(.*?)"', chunk)
+            zero_match = re.search(r'0:"(.*?)"(?=,|$)', chunk)
+            result = {}
+            if g_matches:
+                result["g"] = [g.encode().decode('unicode_escape').replace('\\', '\\').replace('\\"', '"') for g in g_matches]
+            if zero_match:
+                result["0"] = zero_match.group(1).encode().decode('unicode_escape').replace('\\', '\\').replace('\\"', '"')
+            if result:
+                return result
         return None
 
     def ask(
@@ -169,7 +175,9 @@ class SciraAI(Provider):
         prompt: str,
         optimizer: str = None,
         conversationally: bool = False,
-    ) -> Dict[str, Any]: # Note: Stream parameter removed as API doesn't seem to support it
+        stream: bool = True,  # Default to True, always stream
+        raw: bool = False,    # Added raw parameter
+    ) -> Union[Dict[str, Any], Any]:
         conversation_prompt = self.conversation.gen_complete_prompt(prompt)
         if optimizer:
             if optimizer in self.__available_optimizers:
@@ -194,107 +202,162 @@ class SciraAI(Provider):
             "timezone": "Asia/Calcutta"
         }
 
-        try:
-            # Use curl_cffi post with impersonate
-            response = self.session.post(
-                self.url,
-                json=payload,
-                timeout=self.timeout,
-                impersonate="chrome120" # Add impersonate
-            )
-            if response.status_code != 200:
-                # Try to get response content for better error messages
-                try: # Use try-except for reading response content
-                    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, json=payload, timeout=self.timeout,
-                        impersonate="chrome120" # Add impersonate to retry
-                    )
-                    if not response.ok:
+        def for_stream():
+            try:
+                response = self.session.post(
+                    self.url,
+                    json=payload,
+                    timeout=self.timeout,
+                    impersonate="chrome120",
+                    stream=True
+                )
+                if response.status_code != 200:
+                    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, json=payload, timeout=self.timeout,
+                            impersonate="chrome120", stream=True
+                        )
+                        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"Failed to generate response after identity refresh - ({response.status_code}, {response.reason}) - {error_content}"
+                            f"Request failed with status code {response.status_code}. Response: {error_content}"
                         )
-                    print("Identity refreshed successfully.")
-                else:
-                    raise exceptions.FailedToGenerateResponseError(
-                        f"Request failed with status code {response.status_code}. Response: {error_content}"
-                    )
-
-            response_text_raw = response.text # Get raw response text
-
-            # Process the text using sanitize_stream line by line
-            processed_stream = sanitize_stream(
-                data=response_text_raw.splitlines(), # Split into lines
-                intro_value=None, # No simple prefix
-                to_json=False,    # Content is not JSON
-                content_extractor=self._scira_extractor # Use the specific extractor
-            )
 
-            # Aggregate the results from the generator
+                processed_stream = sanitize_stream(
+                    data=response.iter_content(chunk_size=None),
+                    intro_value=None,
+                    to_json=False,
+                    content_extractor=self._scira_extractor,
+                    raw=raw
+                )
+
+                streaming_response = ""
+                in_think = False
+                for content in processed_stream:
+                    if content is None:
+                        continue
+                    if isinstance(content, dict):
+                        # Handle g chunks
+                        g_chunks = content.get("g", [])
+                        zero_chunk = content.get("0")
+                        if g_chunks:
+                            if not in_think:
+                                if raw:
+                                    yield "<think>\n\n"
+                                else:
+                                    yield "<think>\n\n"
+                                in_think = True
+                            for g in g_chunks:
+                                if raw:
+                                    yield g
+                                else:
+                                    yield dict(text=g)
+                        if zero_chunk is not None:
+                            if in_think:
+                                if raw:
+                                    yield "</think>\n\n"
+                                else:
+                                    yield "</think>\n\n"
+                                in_think = False
+                            if raw:
+                                yield zero_chunk
+                            else:
+                                streaming_response += zero_chunk
+                                yield dict(text=zero_chunk)
+                    else:
+                        # fallback for old string/list logic
+                        if raw:
+                            yield content
+                        else:
+                            if content and isinstance(content, str):
+                                streaming_response += content
+                                yield dict(text=content)
+                if not raw:
+                    self.last_response = {"text": streaming_response}
+                    self.conversation.update_chat_history(prompt, streaming_response)
+            except CurlError as e:
+                raise exceptions.FailedToGenerateResponseError(f"Request failed (CurlError): {e}") from e
+            except Exception as e:
+                raise exceptions.FailedToGenerateResponseError(f"Request failed: {e}")
+
+        def for_non_stream():
+            # Always use streaming logic, but aggregate the result
             full_response = ""
-            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 CurlError as e: # Catch CurlError
-            raise exceptions.FailedToGenerateResponseError(f"Request failed (CurlError): {e}") from e
-        except Exception as e:
-            raise exceptions.FailedToGenerateResponseError(f"Request failed: {e}")
+            for chunk in for_stream():
+                if raw:
+                    if isinstance(chunk, str):
+                        full_response += chunk
+                else:
+                    if isinstance(chunk, dict) and "text" in chunk:
+                        full_response += chunk["text"]
+            if not raw:
+                self.last_response = {"text": full_response}
+                self.conversation.update_chat_history(prompt, full_response)
+                return {"text": full_response}
+            else:
+                return full_response
+
+        return for_stream() if stream else for_non_stream()
 
     def chat(
         self,
         prompt: str,
         optimizer: str = None,
         conversationally: bool = False,
-    ) -> str:
-        return self.get_message(
-            self.ask(
-                prompt, optimizer=optimizer, conversationally=conversationally
+        stream: bool = True,  # Default to True, always stream
+        raw: bool = False,    # Added raw parameter
+    ) -> Any:
+        def for_stream():
+            for response in self.ask(
+                prompt, optimizer=optimizer, conversationally=conversationally, stream=True, raw=raw
+            ):
+                if raw:
+                    yield response
+                else:
+                    if isinstance(response, dict):
+                        yield self.get_message(response)
+                    else:
+                        # For <think> and </think> tags (strings), yield as is
+                        yield response
+        def for_non_stream():
+            result = self.ask(
+                prompt, optimizer=optimizer, conversationally=conversationally, stream=False, raw=raw
             )
-        )
+            if raw:
+                return result
+            else:
+                if isinstance(result, dict):
+                    return self.get_message(result)
+                else:
+                    return result
+        return for_stream() if stream else for_non_stream()
 
     def get_message(self, response: dict) -> str:
-        assert isinstance(response, dict), "Response should be of dict data-type only"
-        # Extractor handles formatting
-        return response.get("text", "").replace('\\n', '\n').replace('\\n\\n', '\n\n')
-
-if __name__ == "__main__":
-    print("-" * 100)
-    print(f"{'Model':<50} {'Status':<10} {'Response'}")
-    print("-" * 100)
-
-    test_prompt = "Say 'Hello' in one word"
+        """
+        Retrieves message only from response
 
-    # Test each model
-    for model in SciraAI.AVAILABLE_MODELS:
-        print(f"\rTesting {model}...", end="")
+        Args:
+            response (dict): Response generated by `self.ask`
 
-        try:
-            test_ai = SciraAI(model=model, timeout=120)  # Increased timeout
-            response = test_ai.chat(test_prompt)
+        Returns:
+            str: Message extracted
+        """
+        assert isinstance(response, dict), "Response should be of dict data-type only"
+        return response.get("text", "")
 
-            if response and len(response.strip()) > 0:
-                status = "✓"
-                # Clean and truncate response
-                clean_text = response.strip().encode('utf-8', errors='ignore').decode('utf-8')
-                display_text = clean_text[:50] + "..." if len(clean_text) > 50 else clean_text
-            else:
-                status = "✗"
-                display_text = "Empty or invalid response"
-
-            print(f"\r{model:<50} {status:<10} {display_text}")
-        except Exception as e:
-            error_msg = str(e)
-            # Truncate very long error messages
-            if len(error_msg) > 100:
-                error_msg = error_msg[:97] + "..."
-            print(f"\r{model:<50} {'✗':<10} Error: {error_msg}")
+if __name__ == "__main__":
+    ai = SciraAI()
+    resp = ai.chat("What is the capital of France?", stream=True, raw=False)
+    for chunk in resp:
+        print(chunk, end="", flush=True)