webscout 8.2.7__py3-none-any.whl → 8.2.8__py3-none-any.whl

webscout/Provider/geminiapi.py

@@ -0,0 +1,208 @@
+"""
+Install the Google AI Python SDK
+
+$ pip install google-generativeai
+"""
+
+import os
+import google.generativeai as genai
+
+from google.generativeai.types import HarmCategory, HarmBlockThreshold
+import requests
+from webscout.AIutel import Optimizers
+from webscout.AIutel import Conversation
+from webscout.AIutel import AwesomePrompts
+from webscout.AIbase import Provider
+
+
+class GEMINIAPI(Provider):
+    """
+    A class to interact with the Gemini API using the google-generativeai library.
+    """
+
+    def __init__(
+        self,
+        api_key,
+        model_name: str = "gemini-1.5-flash-latest",
+        temperature: float = 1,
+        top_p: float = 0.95,
+        top_k: int = 64,
+        max_output_tokens: int = 8192,
+        is_conversation: bool = True,
+        timeout: int = 30,
+        intro: str = None,
+        filepath: str = None,
+        update_file: bool = True,
+        proxies: dict = {},
+        history_offset: int = 10250,
+        act: str = None,
+        system_instruction: str = "You are a helpful and informative AI assistant.",
+        safety_settings: dict = None,
+    ):
+        """
+        Initializes the Gemini API with the given parameters.
+
+        Args:
+            api_key (str, optional): Your Gemini API key. If None, it will use the environment variable "GEMINI_API_KEY". 
+                                      Defaults to None.
+            model_name (str, optional): The name of the Gemini model to use. 
+                                        Defaults to "gemini-1.5-flash-exp-0827".
+            temperature (float, optional): The temperature parameter for the model. Defaults to 1.
+            top_p (float, optional): The top_p parameter for the model. Defaults to 0.95.
+            top_k (int, optional): The top_k parameter for the model. Defaults to 64.
+            max_output_tokens (int, optional): The maximum number of output tokens. Defaults to 8192.
+            is_conversation (bool, optional): Flag for chatting conversationally. Defaults to True.
+            timeout (int, optional): Http request timeout. Defaults to 30.
+            intro (str, optional): Conversation introductory prompt. Defaults to None.
+            filepath (str, optional): Path to file containing conversation history. Defaults to None.
+            update_file (bool, optional): Add new prompts and responses to the file. Defaults to True.
+            proxies (dict, optional): Http request proxies. Defaults to {}.
+            history_offset (int, optional): Limit conversation history to this number of last texts. Defaults to 10250.
+            act (str|int, optional): Awesome prompt key or index. (Used as intro). Defaults to None.
+            system_instruction (str, optional): System instruction to guide the AI's behavior. 
+                                                Defaults to "You are a helpful and informative AI assistant.".
+        """
+        self.api_key = api_key
+        self.model_name = model_name
+        self.temperature = temperature
+        self.top_p = top_p
+        self.top_k = top_k
+        self.max_output_tokens = max_output_tokens
+        self.system_instruction = system_instruction
+        self.safety_settings = safety_settings if safety_settings else {}
+        self.session = requests.Session()  # Not directly used for Gemini API calls, but can be used for other requests
+        self.is_conversation = is_conversation
+        self.max_tokens_to_sample = max_output_tokens
+        self.timeout = timeout
+        self.last_response = {}
+
+        self.__available_optimizers = (
+            method
+            for method in dir(Optimizers)
+            if callable(getattr(Optimizers, method)) and not method.startswith("__")
+        )
+        Conversation.intro = (
+            AwesomePrompts().get_act(
+                act, raise_not_found=True, default=None, case_insensitive=True
+            )
+            if act
+            else intro or Conversation.intro
+        )
+        self.conversation = Conversation(
+            is_conversation, self.max_tokens_to_sample, filepath, update_file
+        )
+        self.conversation.history_offset = history_offset
+        self.session.proxies = proxies
+
+        # Configure the Gemini API
+        genai.configure(api_key=self.api_key)
+
+        # Create the model with generation config
+        self.generation_config = {
+            "temperature": self.temperature,
+            "top_p": self.top_p,
+            "top_k": self.top_k,
+            "max_output_tokens": self.max_output_tokens,
+            "response_mime_type": "text/plain",
+        }
+
+        self.model = genai.GenerativeModel(
+            model_name=self.model_name,
+            generation_config=self.generation_config,
+            safety_settings=self.safety_settings,
+            system_instruction=self.system_instruction,
+        )
+
+        # Start the chat session
+        self.chat_session = self.model.start_chat()
+
+    def ask(
+        self,
+        prompt: str,
+        stream: bool = False,
+        raw: bool = False,
+        optimizer: str = None,
+        conversationally: bool = False,
+    ) -> dict:
+        """Chat with AI
+
+        Args:
+            prompt (str): Prompt to be send.
+            stream (bool, optional): Not used for Gemini API. Defaults to False.
+            raw (bool, optional): Not used for Gemini API. Defaults to False.
+            optimizer (str, optional): Prompt optimizer name - `[code, shell_command]`. Defaults to None.
+            conversationally (bool, optional): Chat conversationally when using optimizer. Defaults to False.
+        Returns:
+           dict : {}
+        ```json
+        {
+           "text" : "How may I assist you today?"
+        }
+        ```
+        """
+        conversation_prompt = self.conversation.gen_complete_prompt(prompt)
+        if optimizer:
+            if optimizer in self.__available_optimizers:
+                conversation_prompt = getattr(Optimizers, optimizer)(
+                    conversation_prompt if conversationally else prompt
+                )
+            else:
+                raise Exception(
+                    f"Optimizer is not one of {self.__available_optimizers}"
+                )
+
+        # Send the message to the chat session and get the response
+        response = self.chat_session.send_message(conversation_prompt)
+        self.last_response.update(dict(text=response.text))
+        self.conversation.update_chat_history(
+            prompt, self.get_message(self.last_response)
+        )
+        return self.last_response 
+
+    def chat(
+        self,
+        prompt: str,
+        stream: bool = False,  # Streaming not supported by the current google-generativeai library
+        optimizer: str = None,
+        conversationally: bool = False,
+    ) -> str:
+        """Generate response `str`
+
+        Args:
+            prompt (str): Prompt to be send.
+            stream (bool, optional): Not used for Gemini API. Defaults to False.
+            optimizer (str, optional): Prompt optimizer name - `[code, shell_command]`. Defaults to None.
+            conversationally (bool, optional): Chat conversationally when using optimizer. Defaults to False.
+        Returns:
+            str: Response generated
+        """
+        return self.get_message(
+            self.ask(
+                prompt,
+                optimizer=optimizer,
+                conversationally=conversationally,
+            )
+        )
+
+    def get_message(self, response: dict) -> str:
+        """Retrieves message only from response
+
+        Args:
+            response (dict): Response generated by `self.ask`
+
+        Returns:
+            str: Message extracted
+        """
+        assert isinstance(response, dict), "Response should be of dict data-type only"
+        return response["text"]
+if __name__ == "__main__":
+    safety_settings = {
+        HarmCategory.HARM_CATEGORY_HATE_SPEECH: HarmBlockThreshold.BLOCK_NONE,
+        HarmCategory.HARM_CATEGORY_HARASSMENT: HarmBlockThreshold.BLOCK_NONE,
+        HarmCategory.HARM_CATEGORY_SEXUALLY_EXPLICIT: HarmBlockThreshold.BLOCK_NONE,
+        HarmCategory.HARM_CATEGORY_DANGEROUS_CONTENT: HarmBlockThreshold.BLOCK_NONE,
+    }
+    ai = GEMINIAPI(api_key="" , safety_settings=safety_settings)
+    res = ai.chat(input(">>> "))
+    for r in res:
+        print(r, end="", flush=True)

webscout/Provider/granite.py

@@ -0,0 +1,235 @@
+from curl_cffi.requests import Session
+from curl_cffi import CurlError
+import json
+from typing import Optional, Union, Any, Dict, Generator
+
+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
+
+class IBMGranite(Provider):
+    """
+    A class to interact with the IBM Granite API (accessed via d18n68ssusgr7r.cloudfront.net)
+    using Lit agent for the user agent.
+    """
+
+    AVAILABLE_MODELS = ["granite-3-8b-instruct", "granite-3-2-8b-instruct"]
+
+    def __init__(
+        self,
+        api_key: str,
+        is_conversation: bool = True,
+        max_tokens: int = 600, # Note: max_tokens is not used by this API
+        timeout: int = 30,
+        intro: str = None,
+        filepath: str = None,
+        update_file: bool = True,
+        proxies: dict = {},
+        history_offset: int = 10250,
+        act: str = None,
+        model: str = "granite-3-2-8b-instruct",
+        system_prompt: str = "You are a helpful AI assistant.",
+        thinking: bool = False,
+    ):
+        """Initializes the IBMGranite API client using Lit agent for the user agent."""
+        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
+        self.api_endpoint = "https://d18n68ssusgr7r.cloudfront.net/v1/chat/completions"
+        self.stream_chunk_size = 64
+        self.timeout = timeout
+        self.last_response = {}
+        self.model = model
+        self.system_prompt = system_prompt
+        self.thinking = thinking
+
+        # Use Lit agent (keep if needed for other headers or logic)
+        self.headers = {
+            "authority": "d18n68ssusgr7r.cloudfront.net", # Keep authority
+            "accept": "application/json,application/jsonl", # Keep accept
+            "content-type": "application/json",
+            "origin": "https://www.ibm.com", # Keep origin
+            "referer": "https://www.ibm.com/", # Keep referer
+        }
+        self.headers["Authorization"] = f"Bearer {api_key}"
+        
+        # Update curl_cffi session headers and proxies
+        self.session.headers.update(self.headers)
+        self.session.proxies = proxies # Assign proxies directly
+
+        self.__available_optimizers = (
+            method for method in dir(Optimizers)
+            if callable(getattr(Optimizers, method)) and not method.startswith("__")
+        )
+
+        Conversation.intro = (
+            AwesomePrompts().get_act(
+                act, raise_not_found=True, default=None, case_insensitive=True
+            )
+            if act
+            else intro or Conversation.intro
+        )
+        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,
+        stream: bool = False, # API supports streaming
+        raw: bool = False,
+        optimizer: str = None,
+        conversationally: bool = False,
+    ) -> Union[Dict[str, Any], Generator[Any, None, None]]:
+        """Chat with AI
+        Args:
+            prompt (str): Prompt to be sent.
+            stream (bool, optional): Flag for streaming response. Defaults to False.
+            raw (bool, optional): Stream back raw response as received. Defaults to False.
+            optimizer (str, optional): Prompt optimizer name - `[code, shell_command]`. Defaults to None.
+            conversationally (bool, optional): Chat conversationally when using optimizer. Defaults to False.
+        Returns:
+            Union[Dict, Generator[Dict, None, None]]: Response generated
+        """
+        conversation_prompt = self.conversation.gen_complete_prompt(prompt)
+        if optimizer:
+            if optimizer in self.__available_optimizers:
+                conversation_prompt = getattr(Optimizers, optimizer)(
+                    conversation_prompt if conversationally else prompt
+                )
+            else:
+                raise Exception(f"Optimizer is not one of {self.__available_optimizers}")
+
+        payload = {
+            "model": self.model,
+            "messages": [
+                {"role": "system", "content": self.system_prompt},
+                {"role": "user", "content": conversation_prompt},
+            ],
+            "stream": True # API seems to require stream=True based on response format
+        }
+
+        def for_stream():
+            streaming_text = "" # Initialize outside try block
+            try:
+                # Use curl_cffi session post with impersonate
+                response = self.session.post(
+                    self.api_endpoint, 
+                    # headers are set on the session
+                    json=payload, 
+                    stream=True, 
+                    timeout=self.timeout,
+                    impersonate="chrome110" # Use a common impersonation profile
+                )
+                response.raise_for_status() # Check for HTTP errors
+
+                # Use sanitize_stream
+                processed_stream = sanitize_stream(
+                    data=response.iter_content(chunk_size=None), # Pass byte iterator
+                    intro_value=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)
+                self.conversation.update_chat_history(prompt, streaming_text)
+                
+            except CurlError as e: # Catch CurlError
+                raise exceptions.ProviderConnectionError(f"Request failed (CurlError): {e}") from e
+            except json.JSONDecodeError as e: # Keep specific JSON error handling
+                raise exceptions.InvalidResponseError(f"Failed to decode JSON response: {e}") from e
+            except Exception as e: # Catch other potential exceptions (like HTTPError)
+                err_text = getattr(e, 'response', None) and getattr(e.response, 'text', '')
+                # Use specific exception type if available, otherwise generic
+                ex_type = exceptions.FailedToGenerateResponseError if not isinstance(e, exceptions.ProviderConnectionError) else type(e)
+                raise ex_type(f"An unexpected error occurred ({type(e).__name__}): {e} - {err_text}") from e
+
+
+        def for_non_stream():
+            # Aggregate the stream using the updated for_stream logic
+            full_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_text += chunk_data["text"]
+                    # Handle raw string case if raw=True was passed
+                    elif raw and isinstance(chunk_data, str):
+                         full_text += chunk_data
+            except Exception as e:
+                 # If aggregation fails but some text was received, use it. Otherwise, re-raise.
+                 if not full_text:
+                     raise exceptions.FailedToGenerateResponseError(f"Failed to get non-stream response: {str(e)}") from e
+
+            # last_response and history are updated within for_stream
+            # Return the final aggregated response dict or raw string
+            return full_text if raw else self.last_response
+
+
+        # 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(
+        self,
+        prompt: str,
+        stream: bool = False,
+        optimizer: str = None,
+        conversationally: bool = False,
+    ) -> Union[str, Generator[str, None, None]]:
+        """Generate response as a string using chat method"""
+        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
+
+        def for_non_stream_chat():
+            # ask() returns dict or str when not streaming
+            response_data = self.ask(
+                prompt, stream=False, raw=False, # Ensure ask returns dict
+                optimizer=optimizer, conversationally=conversationally
+            )
+            return self.get_message(response_data) # get_message expects dict
+
+        return for_stream_chat() if stream else for_non_stream_chat()
+
+    def get_message(self, response: dict) -> str:
+        """Retrieves message only from response"""
+        assert isinstance(response, dict), "Response should be of dict data-type only"
+        return response["text"]
+
+if __name__ == "__main__":
+    # Ensure curl_cffi is installed
+    from rich import print
+    # Example usage: Initialize without logging.
+    ai = IBMGranite(
+        api_key="",  # press f12 to see the API key
+        thinking=True,
+    )
+    response = ai.chat("write a poem about AI", stream=True)
+    for chunk in response:
+        print(chunk, end="", flush=True)