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

webscout/Provider/LambdaChat.py

@@ -1,4 +1,5 @@
-import requests
+from curl_cffi.requests import Session
+from curl_cffi import CurlError
 import json
 import time
 import random
@@ -6,8 +7,8 @@ import re
 import uuid
 from typing import Any, Dict, List, Optional, Union, Generator
 
-from webscout.AIutel import Conversation
-from webscout.AIbase import Provider
+from webscout.AIutel import Conversation, sanitize_stream
+from webscout.AIbase import Provider # Import sanitize_stream
 from webscout import exceptions
 from webscout.litagent import LitAgent
 
@@ -31,42 +32,41 @@ class LambdaChat(Provider):
     def __init__(
         self,
         is_conversation: bool = True,
-        max_tokens: int = 2000,
+        max_tokens: int = 2000, # Note: max_tokens is not used by this API
         timeout: int = 60,
         filepath: str = None,
         update_file: bool = True,
         proxies: dict = {},
         model: str = "deepseek-llama3.3-70b",
-        assistantId: str = None,
-        system_prompt: str = "You are a helpful assistant. Please answer the following question.",
+        assistantId: str = None, # Note: assistantId is not used by this API
+        system_prompt: str = "You are a helpful assistant. Please answer the following question.", # Note: system_prompt is not used by this API
     ):
         """Initialize the LambdaChat client."""
         if model not in self.AVAILABLE_MODELS:
             raise ValueError(f"Invalid model: {model}. Choose from: {self.AVAILABLE_MODELS}")
             
         self.model = model
-        self.session = requests.Session()
-        self.session.proxies.update(proxies)
+        # Initialize curl_cffi Session
+        self.session = Session()
         self.assistantId = assistantId
         self.system_prompt = system_prompt
         
         # Set up headers for all requests
         self.headers = {
-            "Content-Type": "application/json",
-            "User-Agent": LitAgent().random(),
-            "Accept": "*/*",
-            "Accept-Encoding": "gzip, deflate, br, zstd",
-            "Accept-Language": "en-US,en;q=0.9",
-            "Origin": self.url,
-            "Referer": f"{self.url}/",
-            "Sec-Ch-Ua": "\"Chromium\";v=\"120\"",
-            "Sec-Ch-Ua-Mobile": "?0",
-            "Sec-Ch-Ua-Platform": "\"Windows\"",
-            "Sec-Fetch-Dest": "empty",
+            "Content-Type": "application/json", # Keep Content-Type for JSON posts
+            "Accept": "*/*", # Keep Accept
+            # "User-Agent": LitAgent().random(), # Removed, handled by impersonate
+            "Accept-Language": "en-US,en;q=0.9", # Keep Accept-Language
+            "Origin": self.url, # Keep Origin
+            "Referer": f"{self.url}/", # Keep Referer (will be updated per request)
+            # "Sec-Ch-Ua": "\"Chromium\";v=\"120\"", # Removed, handled by impersonate
+            # "Sec-Ch-Ua-Mobile": "?0", # Removed, handled by impersonate
+            # "Sec-Ch-Ua-Platform": "\"Windows\"", # Removed, handled by impersonate
+            "Sec-Fetch-Dest": "empty", # Keep Sec-Fetch-* headers
             "Sec-Fetch-Mode": "cors",
             "Sec-Fetch-Site": "same-origin",
-            "DNT": "1",
-            "Priority": "u=1, i"
+            "DNT": "1", # Keep DNT
+            "Priority": "u=1, i" # Keep Priority
         }
         
         # Provider settings
@@ -81,11 +81,17 @@ class LambdaChat(Provider):
         # Store conversation data for different models
         self._conversation_data = {}
 
+        # Update curl_cffi session headers and proxies
+        self.session.headers.update(self.headers)
+        self.session.proxies = proxies # Assign proxies directly
+
     def create_conversation(self, model: str):
         """Create a new conversation with the specified model."""
         url = f"{self.url}/conversation"
         payload = {
-            "model": model
+            "model": model,
+            "preprompt": self.system_prompt,
+
         }
         
         # Update referer for this specific request
@@ -93,7 +99,13 @@ class LambdaChat(Provider):
         headers["Referer"] = f"{self.url}/models/{model}"
         
         try:
-            response = self.session.post(url, json=payload, headers=headers)
+            # Use curl_cffi session post with impersonate
+            response = self.session.post(
+                url, 
+                json=payload, 
+                headers=headers, # Use updated headers with specific Referer
+                impersonate="chrome110" # Use a common impersonation profile
+            )
             
             if response.status_code == 401:
                 raise exceptions.AuthenticationError("Authentication failed.")
@@ -113,14 +125,21 @@ class LambdaChat(Provider):
                 }
                 
             return conversation_id
-        except requests.exceptions.RequestException:
+        except CurlError as e: # Catch CurlError
+            # Log or handle CurlError specifically if needed
+            return None
+        except Exception: # Catch other potential exceptions (like JSONDecodeError, HTTPError)
             return None
     
     def fetch_message_id(self, conversation_id: str) -> str:
         """Fetch the latest message ID for a conversation."""
         try:
             url = f"{self.url}/conversation/{conversation_id}/__data.json?x-sveltekit-invalidated=11"
-            response = self.session.get(url, headers=self.headers)
+            response = self.session.get(
+                url, 
+                headers=self.headers, # Use base headers
+                impersonate="chrome110" # Use a common impersonation profile
+            )
             response.raise_for_status()
             
             # Parse the JSON data from the response
@@ -150,7 +169,9 @@ class LambdaChat(Provider):
             
             return message_id
             
-        except Exception:
+        except CurlError: # Catch CurlError
+            return str(uuid.uuid4()) # Fallback on CurlError
+        except Exception: # Catch other potential exceptions
             # Fall back to a UUID if there's an error
             return str(uuid.uuid4())
     
@@ -161,70 +182,29 @@ class LambdaChat(Provider):
         boundary += "".join(random.choice(boundary_chars) for _ in range(16))
         return boundary
     
-    def process_response(self, response, prompt: str):
-        """Process streaming response and extract content."""
-        full_text = ""
-        sources = None
+    @staticmethod
+    def _lambdachat_extractor(chunk: Union[str, Dict[str, Any]]) -> Optional[str]:
+        """Extracts content from LambdaChat stream JSON objects."""
+        if not isinstance(chunk, dict) or "type" not in chunk:
+            return None
+
         reasoning_text = ""
-        has_reasoning = False
-        
-        for line in response.iter_lines(decode_unicode=True):
-            if not line:
-                continue
-                
-            try:
-                # Parse each line as JSON
-                data = json.loads(line)
-                
-                # Handle different response types
-                if "type" not in data:
-                    continue
-                    
-                if data["type"] == "stream" and "token" in data:
-                    token = data["token"].replace("\u0000", "")
-                    full_text += token
-                    resp = {"text": token}
-                    yield resp
-                elif data["type"] == "finalAnswer":
-                    final_text = data.get("text", "")
-                    if final_text and not full_text:
-                        full_text = final_text
-                        resp = {"text": final_text}
-                        yield resp
-                elif data["type"] == "webSearch" and "sources" in data:
-                    sources = data["sources"]
-                elif data["type"] == "reasoning":
-                    has_reasoning = True
-                    if data.get("subtype") == "stream" and "token" in data:
-                        reasoning_text += data["token"]
-                    
-                    # If we have reasoning, prepend it to the next text output
-                    if reasoning_text and not full_text:
-                        resp = {"text": f"<think>\n{reasoning_text}\n</think>\n", "is_reasoning": True}
-                        yield resp
-                    
-            except json.JSONDecodeError:
-                continue
-        
-        # Update conversation history only for saving to file if needed
-        if full_text and self.conversation.file:
-            if has_reasoning:
-                full_text_with_reasoning = f"<think>\n{reasoning_text}\n</think>\n{full_text}"
-                self.last_response = {"text": full_text_with_reasoning}
-                self.conversation.update_chat_history(prompt, full_text_with_reasoning)
-            else:
-                self.last_response = {"text": full_text}
-                self.conversation.update_chat_history(prompt, full_text)
-        
-        return full_text
+        if chunk["type"] == "stream" and "token" in chunk:
+            return chunk["token"].replace("\u0000", "")
+        elif chunk["type"] == "finalAnswer":
+            return chunk.get("text")
+        elif chunk["type"] == "reasoning" and chunk.get("subtype") == "stream" and "token" in chunk:
+            # Prepend reasoning with <think> tags? Or handle separately? For now, just return token.
+            return chunk["token"] # Or potentially format as f"<think>{chunk['token']}</think>"
+        return None
     
     def ask(
         self,
         prompt: str,
-        stream: bool = False,
+        stream: bool = False, # API supports streaming
         raw: bool = False,
-        optimizer: str = None,
-        conversationally: bool = False,
+        optimizer: str = None, # Note: optimizer is not used by this API
+        conversationally: bool = False, # Note: conversationally is not used by this API
         web_search: bool = False,
     ) -> Union[Dict[str, Any], Generator]:
         """Send a message to the Lambda Chat API"""
@@ -275,43 +255,59 @@ class LambdaChat(Provider):
         multipart_headers["Content-Length"] = str(len(body))
         
         def for_stream():
+            streaming_text = "" # Initialize for history
             try:
                 # Try with multipart/form-data first
                 response = None
                 try:
+                    # Use curl_cffi session post with impersonate
                     response = self.session.post(
                         url, 
                         data=body,
-                        headers=multipart_headers,
+                        headers=multipart_headers, # Use multipart headers
                         stream=True,
-                        timeout=self.timeout
+                        timeout=self.timeout,
+                        impersonate="chrome110" # Use a common impersonation profile
                     )
-                except requests.exceptions.RequestException:
-                    pass
-                
+                    response.raise_for_status() # Check status after potential error
+                except (CurlError, exceptions.FailedToGenerateResponseError, Exception): # Catch potential errors
+                    response = None # Ensure response is None if multipart fails
+
                 # If multipart fails or returns error, try with regular JSON
                 if not response or response.status_code != 200:
+                    # Use curl_cffi session post with impersonate
                     response = self.session.post(
                         url, 
-                        json=request_data,
-                        headers=headers,
+                        json=request_data, # Use JSON payload
+                        headers=headers, # Use regular headers
                         stream=True,
-                        timeout=self.timeout
+                        timeout=self.timeout,
+                        impersonate="chrome110" # Use a common impersonation profile
                     )
                 
-                # If both methods fail, raise exception
-                if response.status_code != 200:
-                    raise exceptions.FailedToGenerateResponseError(f"Request failed with status code {response.status_code}")
+                response.raise_for_status() # Check status after potential fallback
                 
-                # Process the streaming response
-                yield from self.process_response(response, prompt)
+                # 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
+                    content_extractor=self._lambdachat_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 _lambdachat_extractor
+                    if content_chunk and isinstance(content_chunk, str):
+                        streaming_text += content_chunk # Aggregate text for history
+                        resp = {"text": content_chunk}
+                        yield resp if not raw else content_chunk
                 
-            except Exception as e:
-                if isinstance(e, requests.exceptions.RequestException):
-                    if hasattr(e, 'response') and e.response is not None:
-                        status_code = e.response.status_code 
-                        if status_code == 401:
-                            raise exceptions.AuthenticationError("Authentication failed.")
+            except (CurlError, exceptions.FailedToGenerateResponseError, Exception) as e: # Catch errors from both attempts
+                # Handle specific exceptions if needed
+                if isinstance(e, CurlError):
+                     # Log or handle CurlError specifically
+                     pass
                 
                 # Try another model if current one fails
                 if len(self.AVAILABLE_MODELS) > 1:
@@ -328,15 +324,33 @@ class LambdaChat(Provider):
                         return
                 
                 # If we get here, all models failed
-                raise exceptions.FailedToGenerateResponseError(f"Request failed: {str(e)}")
+                raise exceptions.FailedToGenerateResponseError(f"Request failed after trying fallback: {str(e)}") from e
+
+            # Update history after stream finishes
+            if streaming_text and self.conversation.file:
+                self.last_response = {"text": streaming_text}
+                self.conversation.update_chat_history(prompt, streaming_text)
 
         def for_non_stream():
+            # Aggregate the stream using the updated for_stream logic
             response_text = ""
-            for response in for_stream():
-                if "text" in response:
-                    response_text += response["text"]
-            self.last_response = {"text": response_text}
-            return self.last_response
+            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:
+                        response_text += chunk_data["text"]
+                    # Handle raw string case if raw=True was passed
+                    elif raw and isinstance(chunk_data, str):
+                         response_text += chunk_data
+            except Exception as e:
+                 # If aggregation fails but some text was received, use it. Otherwise, re-raise.
+                 if not response_text:
+                     raise exceptions.FailedToGenerateResponseError(f"Failed to get non-stream response: {str(e)}") from e
+
+            # last_response and history are updated within process_response called by for_stream
+            # Return the final aggregated response dict or raw string
+            return response_text if raw else {"text": response_text} # Return dict for consistency
+
 
         return for_stream() if stream else for_non_stream()
 
@@ -344,25 +358,29 @@ class LambdaChat(Provider):
         self,
         prompt: str,
         stream: bool = False,
-        optimizer: str = None,
-        conversationally: bool = False,
+        optimizer: str = None, # Note: optimizer is not used by this API
+        conversationally: bool = False, # Note: conversationally is not used by this API
         web_search: bool = False
     ) -> Union[str, Generator]:
         """Generate a response to a prompt"""
-        def for_stream():
-            for response in self.ask(
-                prompt, True, optimizer=optimizer, conversationally=conversationally, web_search=web_search
-            ):
-                yield self.get_message(response)
+        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, web_search=web_search
+            )
+            for response_dict in gen:
+                yield self.get_message(response_dict) # get_message expects dict
                 
-        def for_non_stream():
-            return self.get_message(
-                self.ask(
-                    prompt, False, optimizer=optimizer, conversationally=conversationally, web_search=web_search
-                )
+        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, web_search=web_search
             )
+            return self.get_message(response_data) # get_message expects dict
             
-        return for_stream() if stream else for_non_stream()
+        return for_stream_chat() if stream else for_non_stream_chat()
 
     def get_message(self, response: dict) -> str:
         """Extract message text from response"""
@@ -370,6 +388,7 @@ class LambdaChat(Provider):
         return response.get("text", "")
 
 if __name__ == "__main__":
+    # Ensure curl_cffi is installed
     print("-" * 80)
     print(f"{'Model':<50} {'Status':<10} {'Response'}")
     print("-" * 80)
@@ -389,4 +408,4 @@ if __name__ == "__main__":
                 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)}") 
+            print(f"{model:<50} {'✗':<10} {str(e)}")

webscout/Provider/Llama3.py

@@ -1,4 +1,5 @@
-import requests
+from curl_cffi.requests import Session
+from curl_cffi import CurlError
 import json
 from typing import Union, Any, Dict, Generator
 
@@ -52,8 +53,8 @@ class Sambanova(Provider):
         self.model = model
         self.system_prompt = system_prompt
 
-        self.session = requests.Session()
-        self.session.proxies = proxies
+        # Initialize curl_cffi Session
+        self.session = Session()
         self.is_conversation = is_conversation
         self.max_tokens_to_sample = max_tokens
         self.timeout = timeout
@@ -80,8 +81,13 @@ class Sambanova(Provider):
         self.base_url = "https://api.sambanova.ai/v1/chat/completions"
         self.headers = {
             "Authorization": f"Bearer {self.api_key}",
-            "Content-Type": "application/json"
+            "Content-Type": "application/json",
+            # Add User-Agent or sec-ch-ua headers if needed, or rely on impersonate
         }
+        
+        # Update curl_cffi session headers and proxies
+        self.session.headers.update(self.headers)
+        self.session.proxies = proxies # Assign proxies directly
 
     def ask(
         self,
@@ -105,36 +111,42 @@ class Sambanova(Provider):
 
         payload = {
             "model": self.model,
-            "stream": stream,
             "messages": [
                 {"role": "system", "content": self.system_prompt},
                 {"role": "user", "content": conversation_prompt},
             ],
             "max_tokens": self.max_tokens_to_sample,
+            "stream": True # API seems to always stream based on endpoint name
         }
 
         def for_stream():
-            streaming_text = ""
+            streaming_text = "" # Initialize outside try block
             try:
+                # Use curl_cffi session post with impersonate
                 response = self.session.post(
-                    self.base_url, headers=self.headers, json=payload, stream=True, timeout=self.timeout
+                    self.base_url, 
+                    # headers are set on the session
+                    json=payload, 
+                    stream=True, 
+                    timeout=self.timeout,
+                    # proxies are set on the session
+                    impersonate="chrome110" # Use a common impersonation profile
                 )
-                if not response.ok:
-                    raise exceptions.FailedToGenerateResponseError(
-                        f"Request failed: {response.status_code} - {response.text}"
-                    )
-
-                for line in response.iter_lines():
-                    if line:
-                        # Remove the "data:" prefix and extra whitespace if present
-                        line_str = line.decode('utf-8').strip() if isinstance(line, bytes) else line.strip()
-                        if line_str.startswith("data:"):
-                            data = line_str[5:].strip()
-                        else:
-                            data = line_str
-                        if data == "[DONE]":
-                            break
+                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_str = line_bytes.decode('utf-8').strip()
+                            if line_str.startswith("data:"):
+                                data = line_str[5:].strip()
+                            else:
+                                data = line_str # Handle cases where 'data:' prefix might be missing
+                            
+                            if data == "[DONE]":
+                                break
+                            
                             json_data = json.loads(data)
                             # Skip entries without valid choices
                             if not json_data.get("choices"):
@@ -143,26 +155,52 @@ class Sambanova(Provider):
                             delta = choice.get("delta", {})
                             if "content" in delta:
                                 content = delta["content"]
-                                streaming_text += content
-                                # Yield content directly as a string for consistency
-                                yield content
+                                if content: # Ensure content is not None or empty
+                                    streaming_text += content
+                                    resp = {"text": content}
+                                    # Yield dict or raw string chunk
+                                    yield resp if not raw else content
                             # If finish_reason is provided, consider the stream complete
                             if choice.get("finish_reason"):
                                 break
-                        except json.JSONDecodeError:
-                            continue
-                self.last_response = streaming_text
+                        except (json.JSONDecodeError, UnicodeDecodeError):
+                            continue # Ignore lines that are not valid JSON or cannot be decoded
+                
+                # Update history after stream finishes
+                self.last_response = streaming_text # Store aggregated text
                 self.conversation.update_chat_history(
                     prompt, self.last_response
                 )
-            except requests.exceptions.RequestException as e:
-                raise exceptions.ProviderConnectionError(f"Request failed: {e}")
+            except CurlError as e: # Catch CurlError
+                raise exceptions.ProviderConnectionError(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.ProviderConnectionError(f"Request failed ({type(e).__name__}): {e} - {err_text}") from e
+
 
         def for_non_stream():
-            for _ in for_stream():
-                pass
-            return self.last_response
+            # 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:
+                        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 {"text": self.last_response} # Return dict for consistency
+
+
+        # Since the API endpoint suggests streaming, always call the stream generator.
+        # The non-stream wrapper will handle aggregation if stream=False.
         return for_stream() if stream else for_non_stream()
 
     def chat(
@@ -173,12 +211,28 @@ class Sambanova(Provider):
         conversationally: bool = False,
     ) -> Union[str, Generator[str, None, None]]:
         """Generate response `str`"""
-        if stream:
-            # For stream mode, yield the text chunks directly
-            return self.ask(prompt, stream=True, optimizer=optimizer, conversationally=conversationally)
-        else:
-            # For non-stream mode, return the complete text response
-            return self.ask(prompt, stream=False, optimizer=optimizer, conversationally=conversationally)
+        
+        def for_stream_chat():
+             # ask() yields dicts or strings when streaming
+             gen = self.ask(
+                 prompt, stream=True, raw=False, # Ensure ask yields dicts
+                 optimizer=optimizer, conversationally=conversationally
+             )
+             for response_dict in gen:
+                 yield self.get_message(response_dict) # get_message expects dict or string
+
+        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 or string
+
+        return for_stream_chat() if stream else for_non_stream_chat()
 
     def get_message(self, response: Any) -> str:
         """
@@ -197,6 +251,7 @@ class Sambanova(Provider):
         return ""
 
 if __name__ == "__main__":
+    # Ensure curl_cffi is installed
     from rich import print
     ai = Sambanova(api_key='')
     response = ai.chat(input(">>> "), stream=True)