langchain/langchain/llms/openai.py

"""Wrapper around OpenAI APIs."""
from __future__ import annotations

import logging
import sys
import warnings
from typing import (
    Any,
    Callable,
    Dict,
    Generator,
    List,
    Mapping,
    Optional,
    Set,
    Tuple,
    Union,
)

from pydantic import Extra, Field, root_validator
from tenacity import (
    before_sleep_log,
    retry,
    retry_if_exception_type,
    stop_after_attempt,
    wait_exponential,
)

from langchain.llms.base import BaseLLM
from langchain.schema import Generation, LLMResult
from langchain.utils import get_from_dict_or_env

logger = logging.getLogger(__name__)


def update_token_usage(
    keys: Set[str], response: Dict[str, Any], token_usage: Dict[str, Any]
) -> None:
    """Update token usage."""
    _keys_to_use = keys.intersection(response["usage"])
    for _key in _keys_to_use:
        if _key not in token_usage:
            token_usage[_key] = response["usage"][_key]
        else:
            token_usage[_key] += response["usage"][_key]


def _update_response(response: Dict[str, Any], stream_response: Dict[str, Any]) -> None:
    """Update response from the stream response."""
    response["choices"][0]["text"] += stream_response["choices"][0]["text"]
    response["choices"][0]["finish_reason"] = stream_response["choices"][0][
        "finish_reason"
    ]
    response["choices"][0]["logprobs"] = stream_response["choices"][0]["logprobs"]


def _streaming_response_template() -> Dict[str, Any]:
    return {
        "choices": [
            {
                "text": "",
                "finish_reason": None,
                "logprobs": None,
            }
        ]
    }


def _create_retry_decorator(llm: Union[BaseOpenAI, OpenAIChat]) -> Callable[[Any], Any]:
    import openai

    min_seconds = 4
    max_seconds = 10
    # Wait 2^x * 1 second between each retry starting with
    # 4 seconds, then up to 10 seconds, then 10 seconds afterwards
    return retry(
        reraise=True,
        stop=stop_after_attempt(llm.max_retries),
        wait=wait_exponential(multiplier=1, min=min_seconds, max=max_seconds),
        retry=(
            retry_if_exception_type(openai.error.Timeout)
            | retry_if_exception_type(openai.error.APIError)
            | retry_if_exception_type(openai.error.APIConnectionError)
            | retry_if_exception_type(openai.error.RateLimitError)
            | retry_if_exception_type(openai.error.ServiceUnavailableError)
        ),
        before_sleep=before_sleep_log(logger, logging.WARNING),
    )


def completion_with_retry(llm: Union[BaseOpenAI, OpenAIChat], **kwargs: Any) -> Any:
    """Use tenacity to retry the completion call."""
    retry_decorator = _create_retry_decorator(llm)

    @retry_decorator
    def _completion_with_retry(**kwargs: Any) -> Any:
        return llm.client.create(**kwargs)

    return _completion_with_retry(**kwargs)


async def acompletion_with_retry(
    llm: Union[BaseOpenAI, OpenAIChat], **kwargs: Any
) -> Any:
    """Use tenacity to retry the async completion call."""
    retry_decorator = _create_retry_decorator(llm)

    @retry_decorator
    async def _completion_with_retry(**kwargs: Any) -> Any:
        # Use OpenAI's async api https://github.com/openai/openai-python#async-api
        return await llm.client.acreate(**kwargs)

    return await _completion_with_retry(**kwargs)


class BaseOpenAI(BaseLLM):
    """Wrapper around OpenAI large language models.

    To use, you should have the ``openai`` python package installed, and the
    environment variable ``OPENAI_API_KEY`` set with your API key.

    Any parameters that are valid to be passed to the openai.create call can be passed
    in, even if not explicitly saved on this class.

    Example:
        .. code-block:: python

            from langchain.llms import OpenAI
            openai = OpenAI(model_name="text-davinci-003")
    """

    client: Any  #: :meta private:
    model_name: str = "text-davinci-003"
    """Model name to use."""
    temperature: float = 0.7
    """What sampling temperature to use."""
    max_tokens: int = 256
    """The maximum number of tokens to generate in the completion.
    -1 returns as many tokens as possible given the prompt and
    the models maximal context size."""
    top_p: float = 1
    """Total probability mass of tokens to consider at each step."""
    frequency_penalty: float = 0
    """Penalizes repeated tokens according to frequency."""
    presence_penalty: float = 0
    """Penalizes repeated tokens."""
    n: int = 1
    """How many completions to generate for each prompt."""
    best_of: int = 1
    """Generates best_of completions server-side and returns the "best"."""
    model_kwargs: Dict[str, Any] = Field(default_factory=dict)
    """Holds any model parameters valid for `create` call not explicitly specified."""
    openai_api_key: Optional[str] = None
    batch_size: int = 20
    """Batch size to use when passing multiple documents to generate."""
    request_timeout: Optional[Union[float, Tuple[float, float]]] = None
    """Timeout for requests to OpenAI completion API. Default is 600 seconds."""
    logit_bias: Optional[Dict[str, float]] = Field(default_factory=dict)
    """Adjust the probability of specific tokens being generated."""
    max_retries: int = 6
    """Maximum number of retries to make when generating."""
    streaming: bool = False
    """Whether to stream the results or not."""

    def __new__(cls, **data: Any) -> Union[OpenAIChat, BaseOpenAI]:  # type: ignore
        """Initialize the OpenAI object."""
        model_name = data.get("model_name", "")
        if model_name.startswith("gpt-3.5-turbo") or model_name.startswith("gpt-4"):
            warnings.warn(
                "You are trying to use a chat model. This way of initializing it is "
                "no longer supported. Instead, please use: "
                "`from langchain.chat_models import ChatOpenAI`"
            )
            return OpenAIChat(**data)
        return super().__new__(cls)

    class Config:
        """Configuration for this pydantic object."""

        extra = Extra.ignore

    @root_validator(pre=True)
    def build_extra(cls, values: Dict[str, Any]) -> Dict[str, Any]:
        """Build extra kwargs from additional params that were passed in."""
        all_required_field_names = {field.alias for field in cls.__fields__.values()}

        extra = values.get("model_kwargs", {})
        for field_name in list(values):
            if field_name not in all_required_field_names:
                if field_name in extra:
                    raise ValueError(f"Found {field_name} supplied twice.")
                logger.warning(
                    f"""WARNING! {field_name} is not default parameter.
                    {field_name} was transfered to model_kwargs.
                    Please confirm that {field_name} is what you intended."""
                )
                extra[field_name] = values.pop(field_name)
        values["model_kwargs"] = extra
        return values

    @root_validator()
    def validate_environment(cls, values: Dict) -> Dict:
        """Validate that api key and python package exists in environment."""
        openai_api_key = get_from_dict_or_env(
            values, "openai_api_key", "OPENAI_API_KEY"
        )
        openai_organization = get_from_dict_or_env(
            values, "openai_organization", "OPENAI_ORGANIZATION", default=None
        )
        try:
            import openai

            openai.api_key = openai_api_key
            if openai_organization:
                print("USING ORGANIZATION: ")
                print(openai_organization)
                openai.organization = openai_organization
            values["client"] = openai.Completion
        except ImportError:
            raise ValueError(
                "Could not import openai python package. "
                "Please it install it with `pip install openai`."
            )
        if values["streaming"] and values["n"] > 1:
            raise ValueError("Cannot stream results when n > 1.")
        if values["streaming"] and values["best_of"] > 1:
            raise ValueError("Cannot stream results when best_of > 1.")
        return values

    @property
    def _default_params(self) -> Dict[str, Any]:
        """Get the default parameters for calling OpenAI API."""
        normal_params = {
            "temperature": self.temperature,
            "max_tokens": self.max_tokens,
            "top_p": self.top_p,
            "frequency_penalty": self.frequency_penalty,
            "presence_penalty": self.presence_penalty,
            "n": self.n,
            "best_of": self.best_of,
            "request_timeout": self.request_timeout,
            "logit_bias": self.logit_bias,
        }
        return {**normal_params, **self.model_kwargs}

    def _generate(
        self, prompts: List[str], stop: Optional[List[str]] = None
    ) -> LLMResult:
        """Call out to OpenAI's endpoint with k unique prompts.

        Args:
            prompts: The prompts to pass into the model.
            stop: Optional list of stop words to use when generating.

        Returns:
            The full LLM output.

        Example:
            .. code-block:: python

                response = openai.generate(["Tell me a joke."])
        """
        # TODO: write a unit test for this
        params = self._invocation_params
        sub_prompts = self.get_sub_prompts(params, prompts, stop)
        choices = []
        token_usage: Dict[str, int] = {}
        # Get the token usage from the response.
        # Includes prompt, completion, and total tokens used.
        _keys = {"completion_tokens", "prompt_tokens", "total_tokens"}
        for _prompts in sub_prompts:
            if self.streaming:
                if len(_prompts) > 1:
                    raise ValueError("Cannot stream results with multiple prompts.")
                params["stream"] = True
                response = _streaming_response_template()
                for stream_resp in completion_with_retry(
                    self, prompt=_prompts, **params
                ):
                    self.callback_manager.on_llm_new_token(
                        stream_resp["choices"][0]["text"],
                        verbose=self.verbose,
                        logprobs=stream_resp["choices"][0]["logprobs"],
                    )
                    _update_response(response, stream_resp)
                choices.extend(response["choices"])
            else:
                response = completion_with_retry(self, prompt=_prompts, **params)
                choices.extend(response["choices"])
            if not self.streaming:
                # Can't update token usage if streaming
                update_token_usage(_keys, response, token_usage)
        return self.create_llm_result(choices, prompts, token_usage)

    async def _agenerate(
        self, prompts: List[str], stop: Optional[List[str]] = None
    ) -> LLMResult:
        """Call out to OpenAI's endpoint async with k unique prompts."""
        params = self._invocation_params
        sub_prompts = self.get_sub_prompts(params, prompts, stop)
        choices = []
        token_usage: Dict[str, int] = {}
        # Get the token usage from the response.
        # Includes prompt, completion, and total tokens used.
        _keys = {"completion_tokens", "prompt_tokens", "total_tokens"}
        for _prompts in sub_prompts:
            if self.streaming:
                if len(_prompts) > 1:
                    raise ValueError("Cannot stream results with multiple prompts.")
                params["stream"] = True
                response = _streaming_response_template()
                async for stream_resp in await acompletion_with_retry(
                    self, prompt=_prompts, **params
                ):
                    if self.callback_manager.is_async:
                        await self.callback_manager.on_llm_new_token(
                            stream_resp["choices"][0]["text"],
                            verbose=self.verbose,
                            logprobs=stream_resp["choices"][0]["logprobs"],
                        )
                    else:
                        self.callback_manager.on_llm_new_token(
                            stream_resp["choices"][0]["text"],
                            verbose=self.verbose,
                            logprobs=stream_resp["choices"][0]["logprobs"],
                        )
                    _update_response(response, stream_resp)
                choices.extend(response["choices"])
            else:
                response = await acompletion_with_retry(self, prompt=_prompts, **params)
                choices.extend(response["choices"])
            if not self.streaming:
                # Can't update token usage if streaming
                update_token_usage(_keys, response, token_usage)
        return self.create_llm_result(choices, prompts, token_usage)

    def get_sub_prompts(
        self,
        params: Dict[str, Any],
        prompts: List[str],
        stop: Optional[List[str]] = None,
    ) -> List[List[str]]:
        """Get the sub prompts for llm call."""
        if stop is not None:
            if "stop" in params:
                raise ValueError("`stop` found in both the input and default params.")
            params["stop"] = stop
        if params["max_tokens"] == -1:
            if len(prompts) != 1:
                raise ValueError(
                    "max_tokens set to -1 not supported for multiple inputs."
                )
            params["max_tokens"] = self.max_tokens_for_prompt(prompts[0])
        sub_prompts = [
            prompts[i : i + self.batch_size]
            for i in range(0, len(prompts), self.batch_size)
        ]
        return sub_prompts

    def create_llm_result(
        self, choices: Any, prompts: List[str], token_usage: Dict[str, int]
    ) -> LLMResult:
        """Create the LLMResult from the choices and prompts."""
        generations = []
        for i, _ in enumerate(prompts):
            sub_choices = choices[i * self.n : (i + 1) * self.n]
            generations.append(
                [
                    Generation(
                        text=choice["text"],
                        generation_info=dict(
                            finish_reason=choice.get("finish_reason"),
                            logprobs=choice.get("logprobs"),
                        ),
                    )
                    for choice in sub_choices
                ]
            )
        llm_output = {"token_usage": token_usage, "model_name": self.model_name}
        return LLMResult(generations=generations, llm_output=llm_output)

    def stream(self, prompt: str, stop: Optional[List[str]] = None) -> Generator:
        """Call OpenAI with streaming flag and return the resulting generator.

        BETA: this is a beta feature while we figure out the right abstraction.
        Once that happens, this interface could change.

        Args:
            prompt: The prompts to pass into the model.
            stop: Optional list of stop words to use when generating.

        Returns:
            A generator representing the stream of tokens from OpenAI.

        Example:
            .. code-block:: python

                generator = openai.stream("Tell me a joke.")
                for token in generator:
                    yield token
        """
        params = self.prep_streaming_params(stop)
        generator = self.client.create(prompt=prompt, **params)

        return generator

    def prep_streaming_params(self, stop: Optional[List[str]] = None) -> Dict[str, Any]:
        """Prepare the params for streaming."""
        params = self._invocation_params
        if params["best_of"] != 1:
            raise ValueError("OpenAI only supports best_of == 1 for streaming")
        if stop is not None:
            if "stop" in params:
                raise ValueError("`stop` found in both the input and default params.")
            params["stop"] = stop
        params["stream"] = True
        return params

    @property
    def _invocation_params(self) -> Dict[str, Any]:
        """Get the parameters used to invoke the model."""
        return self._default_params

    @property
    def _identifying_params(self) -> Mapping[str, Any]:
        """Get the identifying parameters."""
        return {**{"model_name": self.model_name}, **self._default_params}

    @property
    def _llm_type(self) -> str:
        """Return type of llm."""
        return "openai"

    def get_num_tokens(self, text: str) -> int:
        """Calculate num tokens with tiktoken package."""
        # tiktoken NOT supported for Python 3.8 or below
        if sys.version_info[1] <= 8:
            return super().get_num_tokens(text)
        try:
            import tiktoken
        except ImportError:
            raise ValueError(
                "Could not import tiktoken python package. "
                "This is needed in order to calculate get_num_tokens. "
                "Please it install it with `pip install tiktoken`."
            )
        encoder = "gpt2"
        if self.model_name in ("text-davinci-003", "text-davinci-002"):
            encoder = "p50k_base"
        if self.model_name.startswith("code"):
            encoder = "p50k_base"
        # create a GPT-3 encoder instance
        enc = tiktoken.get_encoding(encoder)

        # encode the text using the GPT-3 encoder
        tokenized_text = enc.encode(text)

        # calculate the number of tokens in the encoded text
        return len(tokenized_text)

    def modelname_to_contextsize(self, modelname: str) -> int:
        """Calculate the maximum number of tokens possible to generate for a model.

        text-davinci-003: 4,097 tokens
        text-curie-001: 2,048 tokens
        text-babbage-001: 2,048 tokens
        text-ada-001: 2,048 tokens
        code-davinci-002: 8,000 tokens
        code-cushman-001: 2,048 tokens

        Args:
            modelname: The modelname we want to know the context size for.

        Returns:
            The maximum context size

        Example:
            .. code-block:: python

                max_tokens = openai.modelname_to_contextsize("text-davinci-003")
        """
        if modelname == "text-davinci-003":
            return 4097
        elif modelname == "text-curie-001":
            return 2048
        elif modelname == "text-babbage-001":
            return 2048
        elif modelname == "text-ada-001":
            return 2048
        elif modelname == "code-davinci-002":
            return 8000
        elif modelname == "code-cushman-001":
            return 2048
        else:
            return 4097

    def max_tokens_for_prompt(self, prompt: str) -> int:
        """Calculate the maximum number of tokens possible to generate for a prompt.

        Args:
            prompt: The prompt to pass into the model.

        Returns:
            The maximum number of tokens to generate for a prompt.

        Example:
            .. code-block:: python

                max_tokens = openai.max_token_for_prompt("Tell me a joke.")
        """
        num_tokens = self.get_num_tokens(prompt)

        # get max context size for model by name
        max_size = self.modelname_to_contextsize(self.model_name)
        return max_size - num_tokens


class OpenAI(BaseOpenAI):
    """Generic OpenAI class that uses model name."""

    @property
    def _invocation_params(self) -> Dict[str, Any]:
        return {**{"model": self.model_name}, **super()._invocation_params}


class AzureOpenAI(BaseOpenAI):
    """Azure specific OpenAI class that uses deployment name."""

    deployment_name: str = ""
    """Deployment name to use."""

    @property
    def _identifying_params(self) -> Mapping[str, Any]:
        return {
            **{"deployment_name": self.deployment_name},
            **super()._identifying_params,
        }

    @property
    def _invocation_params(self) -> Dict[str, Any]:
        return {**{"engine": self.deployment_name}, **super()._invocation_params}


class OpenAIChat(BaseLLM):
    """Wrapper around OpenAI Chat large language models.

    To use, you should have the ``openai`` python package installed, and the
    environment variable ``OPENAI_API_KEY`` set with your API key.

    Any parameters that are valid to be passed to the openai.create call can be passed
    in, even if not explicitly saved on this class.

    Example:
        .. code-block:: python

            from langchain.llms import OpenAIChat
            openaichat = OpenAIChat(model_name="gpt-3.5-turbo")
    """

    client: Any  #: :meta private:
    model_name: str = "gpt-3.5-turbo"
    """Model name to use."""
    model_kwargs: Dict[str, Any] = Field(default_factory=dict)
    """Holds any model parameters valid for `create` call not explicitly specified."""
    openai_api_key: Optional[str] = None
    max_retries: int = 6
    """Maximum number of retries to make when generating."""
    prefix_messages: List = Field(default_factory=list)
    """Series of messages for Chat input."""
    streaming: bool = False
    """Whether to stream the results or not."""

    class Config:
        """Configuration for this pydantic object."""

        extra = Extra.ignore

    @root_validator(pre=True)
    def build_extra(cls, values: Dict[str, Any]) -> Dict[str, Any]:
        """Build extra kwargs from additional params that were passed in."""
        all_required_field_names = {field.alias for field in cls.__fields__.values()}

        extra = values.get("model_kwargs", {})
        for field_name in list(values):
            if field_name not in all_required_field_names:
                if field_name in extra:
                    raise ValueError(f"Found {field_name} supplied twice.")
                extra[field_name] = values.pop(field_name)
        values["model_kwargs"] = extra
        return values

    @root_validator()
    def validate_environment(cls, values: Dict) -> Dict:
        """Validate that api key and python package exists in environment."""
        openai_api_key = get_from_dict_or_env(
            values, "openai_api_key", "OPENAI_API_KEY"
        )
        openai_organization = get_from_dict_or_env(
            values, "openai_organization", "OPENAI_ORGANIZATION", default=None
        )
        try:
            import openai

            openai.api_key = openai_api_key
            if openai_organization:
                openai.organization = openai_organization
        except ImportError:
            raise ValueError(
                "Could not import openai python package. "
                "Please it install it with `pip install openai`."
            )
        try:
            values["client"] = openai.ChatCompletion
        except AttributeError:
            raise ValueError(
                "`openai` has no `ChatCompletion` attribute, this is likely "
                "due to an old version of the openai package. Try upgrading it "
                "with `pip install --upgrade openai`."
            )
        warnings.warn(
            "You are trying to use a chat model. This way of initializing it is "
            "no longer supported. Instead, please use: "
            "`from langchain.chat_models import ChatOpenAI`"
        )
        return values

    @property
    def _default_params(self) -> Dict[str, Any]:
        """Get the default parameters for calling OpenAI API."""
        return self.model_kwargs

    def _get_chat_params(
        self, prompts: List[str], stop: Optional[List[str]] = None
    ) -> Tuple:
        if len(prompts) > 1:
            raise ValueError(
                f"OpenAIChat currently only supports single prompt, got {prompts}"
            )
        messages = self.prefix_messages + [{"role": "user", "content": prompts[0]}]
        params: Dict[str, Any] = {**{"model": self.model_name}, **self._default_params}
        if stop is not None:
            if "stop" in params:
                raise ValueError("`stop` found in both the input and default params.")
            params["stop"] = stop
        if params.get("max_tokens") == -1:
            # for ChatGPT api, omitting max_tokens is equivalent to having no limit
            del params["max_tokens"]
        return messages, params

    def _generate(
        self, prompts: List[str], stop: Optional[List[str]] = None
    ) -> LLMResult:
        messages, params = self._get_chat_params(prompts, stop)
        if self.streaming:
            response = ""
            params["stream"] = True
            for stream_resp in completion_with_retry(self, messages=messages, **params):
                token = stream_resp["choices"][0]["delta"].get("content", "")
                response += token
                self.callback_manager.on_llm_new_token(
                    token,
                    verbose=self.verbose,
                )
            return LLMResult(
                generations=[[Generation(text=response)]],
            )
        else:
            full_response = completion_with_retry(self, messages=messages, **params)
            llm_output = {
                "token_usage": full_response["usage"],
                "model_name": self.model_name,
            }
            return LLMResult(
                generations=[
                    [Generation(text=full_response["choices"][0]["message"]["content"])]
                ],
                llm_output=llm_output,
            )

    async def _agenerate(
        self, prompts: List[str], stop: Optional[List[str]] = None
    ) -> LLMResult:
        messages, params = self._get_chat_params(prompts, stop)
        if self.streaming:
            response = ""
            params["stream"] = True
            async for stream_resp in await acompletion_with_retry(
                self, messages=messages, **params
            ):
                token = stream_resp["choices"][0]["delta"].get("content", "")
                response += token
                if self.callback_manager.is_async:
                    await self.callback_manager.on_llm_new_token(
                        token,
                        verbose=self.verbose,
                    )
                else:
                    self.callback_manager.on_llm_new_token(
                        token,
                        verbose=self.verbose,
                    )
            return LLMResult(
                generations=[[Generation(text=response)]],
            )
        else:
            full_response = await acompletion_with_retry(
                self, messages=messages, **params
            )
            llm_output = {
                "token_usage": full_response["usage"],
                "model_name": self.model_name,
            }
            return LLMResult(
                generations=[
                    [Generation(text=full_response["choices"][0]["message"]["content"])]
                ],
                llm_output=llm_output,
            )

    @property
    def _identifying_params(self) -> Mapping[str, Any]:
        """Get the identifying parameters."""
        return {**{"model_name": self.model_name}, **self._default_params}

    @property
    def _llm_type(self) -> str:
        """Return type of llm."""
        return "openai-chat"

    def get_num_tokens(self, text: str) -> int:
        """Calculate num tokens with tiktoken package."""
        # tiktoken NOT supported for Python 3.8 or below
        if sys.version_info[1] <= 8:
            return super().get_num_tokens(text)
        try:
            import tiktoken
        except ImportError:
            raise ValueError(
                "Could not import tiktoken python package. "
                "This is needed in order to calculate get_num_tokens. "
                "Please it install it with `pip install tiktoken`."
            )
        # create a GPT-3.5-Turbo encoder instance
        enc = tiktoken.encoding_for_model("gpt-3.5-turbo")

        # encode the text using the GPT-3.5-Turbo encoder
        tokenized_text = enc.encode(text)

        # calculate the number of tokens in the encoded text
        return len(tokenized_text)