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

webscout/Provider/chatglm.py

@@ -1,11 +1,12 @@
-import requests
+from curl_cffi import CurlError
+from curl_cffi.requests import Session
 import json
 from typing import Any, Dict, Optional, Generator, List, Union
 import uuid
 
 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
@@ -29,7 +30,7 @@ class ChatGLM(Provider):
         plus_model: bool = True,
     ):
         """Initializes the ChatGLM API client."""
-        self.session = requests.Session()
+        self.session = Session() # Use curl_cffi Session
         self.is_conversation = is_conversation
         self.max_tokens_to_sample = max_tokens
         self.api_endpoint = "https://chatglm.cn/chatglm/mainchat-api/guest/stream"
@@ -55,7 +56,7 @@ class ChatGLM(Provider):
             if callable(getattr(Optimizers, method)) and not method.startswith("__")
         )
         self.session.headers.update(self.headers)
-        Conversation.intro = (
+        Conversation.intro = ( # type: ignore
             AwesomePrompts().get_act(
                 act, raise_not_found=True, default=None, case_insensitive=True
             )
@@ -66,7 +67,16 @@ class ChatGLM(Provider):
             is_conversation, self.max_tokens_to_sample, filepath, update_file
         )
         self.conversation.history_offset = history_offset
-        self.session.proxies = proxies
+        self.session.proxies = proxies # Assign proxies directly
+
+    @staticmethod
+    def _chatglm_extractor(chunk: Union[str, Dict[str, Any]]) -> Optional[str]:
+        """Extracts content from ChatGLM stream JSON objects."""
+        if isinstance(chunk, dict):
+            parts = chunk.get('parts', [])
+            if parts and isinstance(parts[0].get('content'), list) and parts[0]['content']:
+                return parts[0]['content'][0].get('text')
+        return None
 
     def ask(
         self,
@@ -119,45 +129,45 @@ class ChatGLM(Provider):
         }
 
         def for_stream():
+            streaming_text = "" # Initialize outside try block
+            last_processed_content = "" # Track the last processed content
             try:
-                with self.session.post(
-                    self.api_endpoint, json=payload, stream=True, timeout=self.timeout
-                ) as response:
-                    response.raise_for_status()
-                    
-                    streaming_text = ""
-                    last_processed_content = "" # Track the last processed content
-                    for chunk in response.iter_lines():
-                        if chunk:
-                            decoded_chunk = chunk.decode('utf-8')
-                            if decoded_chunk.startswith('data: '):
-                                try:
-                                    json_data = json.loads(decoded_chunk[6:])
-                                    parts = json_data.get('parts', [])
-                                    if parts:
-                                        content = parts[0].get('content', [])
-                                        if content:
-                                            text = content[0].get('text', '')
-                                            new_text = text[len(last_processed_content):]
-                                            if new_text:  # Check for new content
-                                                streaming_text += new_text
-                                                last_processed_content = text
-                                                yield new_text if raw else dict(text=new_text)
-                                except json.JSONDecodeError:
-                                    continue
+                response = self.session.post(
+                    self.api_endpoint, json=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="data:",
+                    to_json=True,     # Stream sends JSON
+                    content_extractor=self._chatglm_extractor, # Use the specific extractor
+                    yield_raw_on_error=False # Skip non-JSON lines or lines where extractor fails
+                )
 
+                for current_full_text in processed_stream:
+                    # current_full_text is the full text extracted by _chatglm_extractor
+                    if current_full_text and isinstance(current_full_text, str):
+                        new_text = current_full_text[len(last_processed_content):]
+                        if new_text: # Check for new content
+                            streaming_text += new_text
+                            last_processed_content = current_full_text # Update tracker
+                            yield new_text if raw else dict(text=new_text)
+
+            except CurlError as e:
+                raise exceptions.ProviderConnectionError(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
+                if streaming_text:
                     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: {e}")
-            except Exception as e:
-                raise exceptions.FailedToGenerateResponseError(f"An unexpected error occurred: {e}")
-
         def for_non_stream():
             for _ in for_stream():
                 pass

webscout/Provider/elmo.py

@@ -1,13 +1,14 @@
 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 import exceptions
 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.litagent import LitAgent
+import re # Import re for the extractor
 
 
 class Elmo(Provider):
@@ -84,6 +85,17 @@ class Elmo(Provider):
         )
         self.conversation.history_offset = history_offset
 
+    @staticmethod
+    def _elmo_extractor(chunk: Union[str, Dict[str, Any]]) -> Optional[str]:
+        """Extracts content from the Elmo 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,
@@ -144,7 +156,7 @@ class Elmo(Provider):
         }
 
         def for_stream():
-            full_response = ""  # Initialize outside try block
+            streaming_text = "" # Initialize outside try block
             try:
                 # Use curl_cffi session post with impersonate
                 # Note: The API expects 'text/plain' but we send JSON.
@@ -159,40 +171,32 @@ class Elmo(Provider):
                 )
                 response.raise_for_status()  # Check for HTTP errors
 
-                # Iterate over bytes and decode manually
-                for line_bytes in response.iter_lines():
-                    if line_bytes:
-                        try:
-                            line = line_bytes.decode('utf-8')
-                            if line.startswith('0:'):
-                                # Extract content after '0:"' and before the closing '"'
-                                match = line.split(':"', 1)
-                                if len(match) > 1:
-                                    chunk = match[1]
-                                    if chunk.endswith('"'):
-                                        chunk = chunk[:-1]  # Remove trailing quote
-
-                                    # Handle potential escape sequences like \\n
-                                    formatted_output = chunk.encode().decode('unicode_escape')
+                # 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 text after extraction
+                    content_extractor=self._elmo_extractor, # Use the specific extractor
+                    yield_raw_on_error=True
+                )
 
-                                    if formatted_output:  # Ensure content is not None or empty
-                                        full_response += formatted_output
-                                        resp = dict(text=formatted_output)
-                                        # Yield dict or raw string chunk
-                                        yield resp if not raw else formatted_output
-                        except (UnicodeDecodeError, IndexError):
-                            continue  # Ignore lines that cannot be decoded or parsed
+                for content_chunk in processed_stream:
+                    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=full_response)
-                self.conversation.update_chat_history(
-                    prompt, full_response
-                )
             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)
                 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
+            finally:
+                # Update history after stream finishes
+                self.last_response = 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
@@ -210,7 +214,9 @@ class Elmo(Provider):
                  if not collected_text:
                      raise exceptions.FailedToGenerateResponseError(f"Failed to get non-stream response: {str(e)}") from e
 
-            # last_response and history are updated within for_stream
+            # Update last_response and history *after* aggregation for non-stream
+            self.last_response = {"text": collected_text}
+            self.conversation.update_chat_history(prompt, collected_text)
             # Return the final aggregated response dict or raw string
             return collected_text if raw else self.last_response
 
@@ -265,7 +271,7 @@ class Elmo(Provider):
             str: Message extracted
         """
         assert isinstance(response, dict), "Response should be of dict data-type only"
-        return response["text"]
+        return response.get("text", "") # Use .get for safety
 
 
 if __name__ == "__main__":

webscout/Provider/granite.py

@@ -1,9 +1,9 @@
 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
@@ -77,6 +77,13 @@ 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,
@@ -127,25 +134,21 @@ class IBMGranite(Provider):
                 )
                 response.raise_for_status() # Check for HTTP errors
 
-                # Iterate over bytes and decode manually
-                for line_bytes in response.iter_lines():
-                    if line_bytes:
-                        try:
-                            line = line_bytes.decode('utf-8')
-                            data = json.loads(line)
-                            # Check the specific format [3, "text_chunk"]
-                            if isinstance(data, list) and len(data) == 2 and data[0] == 3 and isinstance(data[1], str):
-                                content = data[1]
-                                if content: # Ensure content is not None or empty
-                                    streaming_text += content
-                                    resp = dict(text=content)
-                                    # Yield dict or raw string chunk
-                                    yield resp if not raw else content
-                            else:
-                                # Skip unrecognized lines/formats
-                                pass
-                        except (json.JSONDecodeError, UnicodeDecodeError):
-                            continue # Ignore lines that are not valid JSON or cannot be decoded
+                # 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
+                )
+
+                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)

webscout/Provider/hermes.py

@@ -4,7 +4,7 @@ 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
@@ -102,6 +102,13 @@ class NousHermes(Provider):
             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,
@@ -145,36 +152,36 @@ class NousHermes(Provider):
             "top_p": self.top_p,
         }
         def for_stream():
-            full_response = ""
+            streaming_text = "" # Initialize outside try block
             try:
                 response = self.session.post(
                     self.api_endpoint, 
                     json=payload, 
                     stream=True, 
                     timeout=self.timeout,
-                    impersonate="chrome110"
+                    impersonate="chrome110" # Keep impersonate
                 )
                 response.raise_for_status()
 
-                for line_bytes in response.iter_lines():
-                    if line_bytes:
-                        try:
-                            decoded_line = line_bytes.decode('utf-8')
-                            if decoded_line.startswith('data: '):
-                                data_str = decoded_line.replace('data: ', '', 1)
-                                data = json.loads(data_str)
-                                if data.get('type') == 'llm_response':
-                                    content = data.get('content', '')
-                                    if content:
-                                        full_response += content
-                                        resp = dict(text=content)
-                                        yield resp if not raw else content
-                        except (json.JSONDecodeError, UnicodeDecodeError):
-                            continue
+                # 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=full_response)
+                self.last_response = dict(text=streaming_text) # Use streaming_text
                 self.conversation.update_chat_history(
-                    prompt, full_response
+                    prompt, streaming_text # Use streaming_text
                 )
 
             except CurlError as e:

webscout/Provider/learnfastai.py

@@ -1,12 +1,12 @@
 import os
 import json
-from typing import Optional, Union, Generator
+from typing import Any, Dict, Optional, Union, Generator
 import uuid
 from curl_cffi.requests import Session
 from curl_cffi import CurlError
 
 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
@@ -79,6 +79,13 @@ class LearnFast(Provider):
         )
         self.conversation.history_offset = history_offset
 
+    @staticmethod
+    def _learnfast_extractor(chunk: Union[str, Dict[str, Any]]) -> Optional[str]:
+        """Extracts message content from LearnFast stream JSON objects."""
+        if isinstance(chunk, dict) and chunk.get('code') == 200 and chunk.get('data'):
+            return chunk['data'].get('message')
+        return None
+
     def generate_unique_id(self) -> str:
         """Generate a 32-character hexadecimal unique ID."""
         return uuid.uuid4().hex
@@ -209,24 +216,22 @@ class LearnFast(Provider):
                 )
                 response.raise_for_status()  # Check for HTTP errors
 
-                # Process the streamed response
-                # Iterate over bytes and decode manually
-                for line_bytes in response.iter_lines():
-                    if line_bytes:
-                        try:
-                            line = line_bytes.decode('utf-8').strip()
-                            if line == "[DONE]":
-                                break
-                            json_response = json.loads(line)
-                            if json_response.get('code') == 200 and json_response.get('data'):
-                                message = json_response['data'].get('message', '')
-                                if message:
-                                    full_response += message
-                                    resp = {"text": message}
-                                    # Yield dict or raw string chunk
-                                    yield resp if not raw else message
-                        except (json.JSONDecodeError, UnicodeDecodeError):
-                            pass # Ignore lines that are not valid JSON or cannot be decoded
+                # 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
+                    skip_markers=["[DONE]"],
+                    content_extractor=self._learnfast_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 _learnfast_extractor
+                    if content_chunk and isinstance(content_chunk, str):
+                        full_response += content_chunk
+                        resp = {"text": content_chunk}
+                        yield resp if not raw else content_chunk
                 
                 # Update history after stream finishes
                 self.last_response = {"text": full_response}