deepset-ai--haystack/haystack/components/embedders/openai_text_embedder.py

# SPDX-FileCopyrightText: 2022-present deepset GmbH <info@deepset.ai>
#
# SPDX-License-Identifier: Apache-2.0

import os
from typing import Any

from openai import AsyncOpenAI, OpenAI
from openai.types import CreateEmbeddingResponse

from haystack import component, default_from_dict, default_to_dict
from haystack.utils import Secret
from haystack.utils.http_client import init_http_client


@component
class OpenAITextEmbedder:
    """
    Embeds strings using OpenAI models.

    You can use it to embed user query and send it to an embedding Retriever.

    ### Usage example
    <!-- test-ignore -->
    ```python
    from haystack.components.embedders import OpenAITextEmbedder

    text_to_embed = "I love pizza!"
    text_embedder = OpenAITextEmbedder()

    print(text_embedder.run(text_to_embed))

    # {'embedding': [0.017020374536514282, -0.023255806416273117, ...],
    # 'meta': {'model': 'text-embedding-ada-002-v2',
    #          'usage': {'prompt_tokens': 4, 'total_tokens': 4}}}
    ```
    """

    def __init__(
        self,
        api_key: Secret = Secret.from_env_var("OPENAI_API_KEY"),
        model: str = "text-embedding-ada-002",
        dimensions: int | None = None,
        api_base_url: str | None = None,
        organization: str | None = None,
        prefix: str = "",
        suffix: str = "",
        timeout: float | None = None,
        max_retries: int | None = None,
        http_client_kwargs: dict[str, Any] | None = None,
    ) -> None:
        """
        Creates an OpenAITextEmbedder component.

        Before initializing the component, you can set the 'OPENAI_TIMEOUT' and 'OPENAI_MAX_RETRIES'
        environment variables to override the `timeout` and `max_retries` parameters respectively
        in the OpenAI client.

        :param api_key:
            The OpenAI API key.
            You can set it with an environment variable `OPENAI_API_KEY`, or pass with this parameter
            during initialization.
        :param model:
            The name of the model to use for calculating embeddings.
            The default model is `text-embedding-ada-002`.
        :param dimensions:
            The number of dimensions of the resulting embeddings. Only `text-embedding-3` and
            later models support this parameter.
        :param api_base_url:
            Overrides default base URL for all HTTP requests.
        :param organization:
            Your organization ID. See OpenAI's
            [production best practices](https://platform.openai.com/docs/guides/production-best-practices/setting-up-your-organization)
            for more information.
        :param prefix:
            A string to add at the beginning of each text to embed.
        :param suffix:
            A string to add at the end of each text to embed.
        :param timeout:
            Timeout for OpenAI client calls. If not set, it defaults to either the
            `OPENAI_TIMEOUT` environment variable, or 30 seconds.
        :param max_retries:
            Maximum number of retries to contact OpenAI after an internal error.
            If not set, it defaults to either the `OPENAI_MAX_RETRIES` environment variable, or set to 5.
        :param http_client_kwargs:
            A dictionary of keyword arguments to configure a custom `httpx.Client`or `httpx.AsyncClient`.
            For more information, see the [HTTPX documentation](https://www.python-httpx.org/api/#client).
        """
        self.model = model
        self.dimensions = dimensions
        self.api_base_url = api_base_url
        self.organization = organization
        self.prefix = prefix
        self.suffix = suffix
        self.api_key = api_key
        self.timeout = timeout
        self.max_retries = max_retries
        self.http_client_kwargs = http_client_kwargs

        self.client: OpenAI | None = None
        self.async_client: AsyncOpenAI | None = None

    def _client_kwargs(self) -> dict[str, Any]:
        timeout = self.timeout if self.timeout is not None else float(os.environ.get("OPENAI_TIMEOUT", "30.0"))
        max_retries = (
            self.max_retries if self.max_retries is not None else int(os.environ.get("OPENAI_MAX_RETRIES", "5"))
        )
        return {
            "api_key": self.api_key.resolve_value(),
            "organization": self.organization,
            "base_url": self.api_base_url,
            "timeout": timeout,
            "max_retries": max_retries,
        }

    def warm_up(self) -> None:
        """
        Initializes the synchronous OpenAI client.
        """
        if self.client is None:
            self.client = OpenAI(
                http_client=init_http_client(self.http_client_kwargs, async_client=False), **self._client_kwargs()
            )

    async def warm_up_async(self) -> None:  # noqa: RUF029
        """
        Initializes the asynchronous OpenAI client on the serving event loop.
        """
        if self.async_client is None:
            self.async_client = AsyncOpenAI(
                http_client=init_http_client(self.http_client_kwargs, async_client=True), **self._client_kwargs()
            )

    def close(self) -> None:
        """
        Releases the synchronous OpenAI client.
        """
        if self.client is not None:
            self.client.close()
            self.client = None

    async def close_async(self) -> None:
        """
        Releases the asynchronous OpenAI client.
        """
        if self.async_client is not None:
            await self.async_client.close()
            self.async_client = None

    def _get_telemetry_data(self) -> dict[str, Any]:
        """
        Data that is sent to Posthog for usage analytics.
        """
        return {"model": self.model}

    def to_dict(self) -> dict[str, Any]:
        """
        Serializes the component to a dictionary.

        :returns:
            Dictionary with serialized data.
        """
        return default_to_dict(
            self,
            api_key=self.api_key,
            model=self.model,
            dimensions=self.dimensions,
            api_base_url=self.api_base_url,
            organization=self.organization,
            prefix=self.prefix,
            suffix=self.suffix,
            timeout=self.timeout,
            max_retries=self.max_retries,
            http_client_kwargs=self.http_client_kwargs,
        )

    @classmethod
    def from_dict(cls, data: dict[str, Any]) -> "OpenAITextEmbedder":
        """
        Deserializes the component from a dictionary.

        :param data:
            Dictionary to deserialize from.
        :returns:
            Deserialized component.
        """
        return default_from_dict(cls, data)

    def _prepare_input(self, text: str) -> dict[str, Any]:
        if not isinstance(text, str):
            raise TypeError(
                "OpenAITextEmbedder expects a string as an input."
                "In case you want to embed a list of Documents, please use the OpenAIDocumentEmbedder."
            )

        text_to_embed = self.prefix + text + self.suffix

        kwargs: dict[str, Any] = {"model": self.model, "input": text_to_embed, "encoding_format": "float"}
        if self.dimensions is not None:
            kwargs["dimensions"] = self.dimensions
        return kwargs

    def _prepare_output(self, result: CreateEmbeddingResponse) -> dict[str, Any]:
        return {"embedding": result.data[0].embedding, "meta": {"model": result.model, "usage": dict(result.usage)}}

    @component.output_types(embedding=list[float], meta=dict[str, Any])
    def run(self, text: str) -> dict[str, Any]:
        """
        Embeds a single string.

        :param text:
            Text to embed.

        :returns:
            A dictionary with the following keys:
            - `embedding`: The embedding of the input text.
            - `meta`: Information about the usage of the model.
        """
        self.warm_up()
        create_kwargs = self._prepare_input(text=text)
        assert self.client is not None  # mypy: client is built by warm_up above
        response = self.client.embeddings.create(**create_kwargs)
        return self._prepare_output(result=response)

    @component.output_types(embedding=list[float], meta=dict[str, Any])
    async def run_async(self, text: str) -> dict[str, Any]:
        """
        Asynchronously embed a single string.

        This is the asynchronous version of the `run` method. It has the same parameters and return values
        but can be used with `await` in async code.

        :param text:
            Text to embed.

        :returns:
            A dictionary with the following keys:
            - `embedding`: The embedding of the input text.
            - `meta`: Information about the usage of the model.
        """
        await self.warm_up_async()
        create_kwargs = self._prepare_input(text=text)
        assert self.async_client is not None  # mypy: async_client is built by warm_up_async above
        response = await self.async_client.embeddings.create(**create_kwargs)
        return self._prepare_output(result=response)