webscout 8.2.7__py3-none-any.whl → 8.2.9__py3-none-any.whl

webscout/Provider/Koboldai.py

@@ -0,0 +1,384 @@
+import requests
+import json
+from ..AIutel import Optimizers
+from ..AIutel import Conversation
+from ..AIutel import AwesomePrompts, sanitize_stream
+from ..AIbase import  Provider, AsyncProvider
+from webscout import exceptions
+from typing import Union, Any, AsyncGenerator, Dict
+import httpx
+#------------------------------------------------------KOBOLDAI-----------------------------------------------------------
+class KOBOLDAI(Provider):
+    def __init__(
+        self,
+        is_conversation: bool = True,
+        max_tokens: int = 600,
+        temperature: float = 1,
+        top_p: float = 1,
+        timeout: int = 30,
+        intro: str = None,
+        filepath: str = None,
+        update_file: bool = True,
+        proxies: dict = {},
+        history_offset: int = 10250,
+        act: str = None,
+    ):
+        """Instantiate TGPT
+
+        Args:
+            is_conversation (str, optional): Flag for chatting conversationally. Defaults to True.
+            max_tokens (int, optional): Maximum number of tokens to be generated upon completion. Defaults to 600.
+            temperature (float, optional): Charge of the generated text's randomness. Defaults to 0.2.
+            top_p (float, optional): Sampling threshold during inference time. Defaults to 0.999.
+            timeout (int, optional): Http requesting timeout. Defaults to 30
+            intro (str, optional): Conversation introductory prompt. Defaults to `Conversation.intro`.
+            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 reqiuest proxies (socks). 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.
+        """
+        self.session = requests.Session()
+        self.is_conversation = is_conversation
+        self.max_tokens_to_sample = max_tokens
+        self.temperature = temperature
+        self.top_p = top_p
+        self.chat_endpoint = (
+            "https://koboldai-koboldcpp-tiefighter.hf.space/api/extra/generate/stream"
+        )
+        self.stream_chunk_size = 64
+        self.timeout = timeout
+        self.last_response = {}
+        self.headers = {
+            "Content-Type": "application/json",
+            "Accept": "application/json",
+        }
+
+        self.__available_optimizers = (
+            method
+            for method in dir(Optimizers)
+            if callable(getattr(Optimizers, method)) and not method.startswith("__")
+        )
+        self.session.headers.update(self.headers)
+        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
+
+    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): 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:
+           dict : {}
+        ```json
+        {
+           "token" : "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}"
+                )
+
+        self.session.headers.update(self.headers)
+        payload = {
+            "prompt": conversation_prompt,
+            "temperature": self.temperature,
+            "top_p": self.top_p,
+        }
+
+        def for_stream():
+            response = self.session.post(
+                self.chat_endpoint, json=payload, stream=True, timeout=self.timeout
+            )
+            if not response.ok:
+                raise Exception(
+                    f"Failed to generate response - ({response.status_code}, {response.reason}) - {response.text}"
+                )
+
+            message_load = ""
+            final_resp = None
+            for value in response.iter_lines(
+                decode_unicode=True,
+                delimiter="" if raw else "event: message\ndata:",
+                chunk_size=self.stream_chunk_size,
+            ):
+                try:
+                    resp = json.loads(value)
+                    message_load += self.get_message(resp)
+                    resp["token"] = message_load
+                    self.last_response.update(resp)
+                    final_resp = resp  # Always keep the latest
+                except json.decoder.JSONDecodeError:
+                    pass
+            if final_resp:
+                yield final_resp if not raw else json.dumps(final_resp)
+                self.conversation.update_chat_history(
+                    prompt, self.get_message(self.last_response)
+                )
+
+        def for_non_stream():
+            # let's make use of stream
+            for _ in for_stream():
+                pass
+            return self.last_response
+
+        return for_stream() if stream else for_non_stream()
+
+    def chat(
+        self,
+        prompt: str,
+        stream: bool = False,
+        optimizer: str = None,
+        conversationally: bool = False,
+    ) -> str:
+        """Generate response `str`
+        Args:
+            prompt (str): Prompt to be send.
+            stream (bool, optional): Flag for streaming response. 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
+        """
+
+        def for_stream():
+            for response in self.ask(
+                prompt, True, optimizer=optimizer, conversationally=conversationally
+            ):
+                yield self.get_message(response)
+
+        def for_non_stream():
+            return self.get_message(
+                self.ask(
+                    prompt,
+                    False,
+                    optimizer=optimizer,
+                    conversationally=conversationally,
+                )
+            )
+
+        return for_stream() if stream else for_non_stream()
+
+    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.get("token")
+class AsyncKOBOLDAI(AsyncProvider):
+    def __init__(
+        self,
+        is_conversation: bool = True,
+        max_tokens: int = 600,
+        temperature: float = 1,
+        top_p: float = 1,
+        timeout: int = 30,
+        intro: str = None,
+        filepath: str = None,
+        update_file: bool = True,
+        proxies: dict = {},
+        history_offset: int = 10250,
+        act: str = None,
+    ):
+        """Instantiate TGPT
+
+        Args:
+            is_conversation (str, optional): Flag for chatting conversationally. Defaults to True.
+            max_tokens (int, optional): Maximum number of tokens to be generated upon completion. Defaults to 600.
+            temperature (float, optional): Charge of the generated text's randomness. Defaults to 0.2.
+            top_p (float, optional): Sampling threshold during inference time. Defaults to 0.999.
+            timeout (int, optional): Http requesting timeout. Defaults to 30
+            intro (str, optional): Conversation introductory prompt. Defaults to `Conversation.intro`.
+            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 reqiuest proxies (socks). 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.
+        """
+        self.is_conversation = is_conversation
+        self.max_tokens_to_sample = max_tokens
+        self.temperature = temperature
+        self.top_p = top_p
+        self.chat_endpoint = (
+            "https://koboldai-koboldcpp-tiefighter.hf.space/api/extra/generate/stream"
+        )
+        self.stream_chunk_size = 64
+        self.timeout = timeout
+        self.last_response = {}
+        self.headers = {
+            "Content-Type": "application/json",
+            "Accept": "application/json",
+        }
+
+        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 = httpx.AsyncClient(headers=self.headers, proxies=proxies)
+
+    async def ask(
+        self,
+        prompt: str,
+        stream: bool = False,
+        raw: bool = False,
+        optimizer: str = None,
+        conversationally: bool = False,
+    ) -> Union[dict, AsyncGenerator]:
+        """Chat with AI asynchronously.
+
+        Args:
+            prompt (str): Prompt to be send.
+            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:
+           dict|AsyncGenerator : ai content
+        ```json
+        {
+           "token" : "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}"
+                )
+
+        payload = {
+            "prompt": conversation_prompt,
+            "temperature": self.temperature,
+            "top_p": self.top_p,
+        }
+
+        async def for_stream():
+            async with self.session.stream(
+                "POST", self.chat_endpoint, json=payload, timeout=self.timeout
+            ) as response:
+                if not response.is_success:
+                    raise exceptions.FailedToGenerateResponseError(
+                        f"Failed to generate response - ({response.status_code}, {response.reason_phrase})"
+                    )
+
+                message_load = ""
+                async for value in response.aiter_lines():
+                    try:
+                        resp = sanitize_stream(value)
+                        message_load += await self.get_message(resp)
+                        resp["token"] = message_load
+                        self.last_response.update(resp)
+                        yield value if raw else resp
+                    except json.decoder.JSONDecodeError:
+                        pass
+
+            self.conversation.update_chat_history(
+                prompt, await self.get_message(self.last_response)
+            )
+
+        async def for_non_stream():
+            # let's make use of stream
+            async for _ in for_stream():
+                pass
+            return self.last_response
+
+        return for_stream() if stream else await for_non_stream()
+
+    async def chat(
+        self,
+        prompt: str,
+        stream: bool = False,
+        optimizer: str = None,
+        conversationally: bool = False,
+    ) -> Union[str, AsyncGenerator]:
+        """Generate response `str` asynchronously.
+        Args:
+            prompt (str): Prompt to be send.
+            stream (bool, optional): Flag for streaming response. 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
+        """
+
+        async def for_stream():
+            async_ask = await self.ask(
+                prompt, True, optimizer=optimizer, conversationally=conversationally
+            )
+            async for response in async_ask:
+                yield await self.get_message(response)
+
+        async def for_non_stream():
+            return await self.get_message(
+                await self.ask(
+                    prompt,
+                    False,
+                    optimizer=optimizer,
+                    conversationally=conversationally,
+                )
+            )
+
+        return for_stream() if stream else await for_non_stream()
+
+    async 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.get("token")