File size: 19,717 Bytes
4962437
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
from __future__ import annotations

import logging
import warnings
from typing import (
    Any,
    Callable,
    Dict,
    List,
    Literal,
    Optional,
    Sequence,
    Set,
    Tuple,
    Union,
)

import numpy as np
from pydantic import BaseModel, Extra, Field, root_validator
from tenacity import (
    AsyncRetrying,
    before_sleep_log,
    retry,
    retry_if_exception_type,
    stop_after_attempt,
    wait_exponential,
)
from swarms.embeddings.base import Embeddings


def get_from_dict_or_env(values: dict, key: str, env_key: str, default: Any = None) -> Any:
    import os

    return values.get(key) or os.getenv(env_key) or default


def get_pydantic_field_names(cls: Any) -> Set[str]:
    return set(cls.__annotations__.keys())


logger = logging.getLogger(__name__)


def _create_retry_decorator(embeddings: OpenAIEmbeddings) -> Callable[[Any], Any]:
    import llm

    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(embeddings.max_retries),
        wait=wait_exponential(multiplier=1, min=min_seconds, max=max_seconds),
        retry=(
            retry_if_exception_type(llm.error.Timeout)
            | retry_if_exception_type(llm.error.APIError)
            | retry_if_exception_type(llm.error.APIConnectionError)
            | retry_if_exception_type(llm.error.RateLimitError)
            | retry_if_exception_type(llm.error.ServiceUnavailableError)
        ),
        before_sleep=before_sleep_log(logger, logging.WARNING),
    )


def _async_retry_decorator(embeddings: OpenAIEmbeddings) -> Any:
    import llm

    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
    async_retrying = AsyncRetrying(
        reraise=True,
        stop=stop_after_attempt(embeddings.max_retries),
        wait=wait_exponential(multiplier=1, min=min_seconds, max=max_seconds),
        retry=(
            retry_if_exception_type(llm.error.Timeout)
            | retry_if_exception_type(llm.error.APIError)
            | retry_if_exception_type(llm.error.APIConnectionError)
            | retry_if_exception_type(llm.error.RateLimitError)
            | retry_if_exception_type(llm.error.ServiceUnavailableError)
        ),
        before_sleep=before_sleep_log(logger, logging.WARNING),
    )

    def wrap(func: Callable) -> Callable:
        async def wrapped_f(*args: Any, **kwargs: Any) -> Callable:
            async for _ in async_retrying:
                return await func(*args, **kwargs)
            raise AssertionError("this is unreachable")

        return wrapped_f

    return wrap


# https://stackoverflow.com/questions/76469415/getting-embeddings-of-length-1-from-langchain-openaiembeddings
def _check_response(response: dict) -> dict:
    if any(len(d["embedding"]) == 1 for d in response["data"]):
        import llm

        raise llm.error.APIError("OpenAI API returned an empty embedding")
    return response


def embed_with_retry(embeddings: OpenAIEmbeddings, **kwargs: Any) -> Any:
    """Use tenacity to retry the embedding call."""
    retry_decorator = _create_retry_decorator(embeddings)

    @retry_decorator
    def _embed_with_retry(**kwargs: Any) -> Any:
        response = embeddings.client.create(**kwargs)
        return _check_response(response)

    return _embed_with_retry(**kwargs)


async def async_embed_with_retry(embeddings: OpenAIEmbeddings, **kwargs: Any) -> Any:
    """Use tenacity to retry the embedding call."""

    @_async_retry_decorator(embeddings)
    async def _async_embed_with_retry(**kwargs: Any) -> Any:
        response = await embeddings.client.acreate(**kwargs)
        return _check_response(response)

    return await _async_embed_with_retry(**kwargs)


class OpenAIEmbeddings(BaseModel, Embeddings):
    """OpenAI embedding models.

    To use, you should have the ``openai`` python package installed, and the
    environment variable ``OPENAI_API_KEY`` set with your API key or pass it
    as a named parameter to the constructor.

    Example:
        .. code-block:: python

            from langchain.embeddings import OpenAIEmbeddings
            openai = OpenAIEmbeddings(openai_api_key="my-api-key")

    In order to use the library with Microsoft Azure endpoints, you need to set
    the OPENAI_API_TYPE, OPENAI_API_BASE, OPENAI_API_KEY and OPENAI_API_VERSION.
    The OPENAI_API_TYPE must be set to 'azure' and the others correspond to
    the properties of your endpoint.
    In addition, the deployment name must be passed as the model parameter.

    Example:
        .. code-block:: python

            import os
            os.environ["OPENAI_API_TYPE"] = "azure"
            os.environ["OPENAI_API_BASE"] = "https://<your-endpoint.openai.azure.com/"
            os.environ["OPENAI_API_KEY"] = "your AzureOpenAI key"
            os.environ["OPENAI_API_VERSION"] = "2023-05-15"
            os.environ["OPENAI_PROXY"] = "http://your-corporate-proxy:8080"

            from langchain.embeddings.openai import OpenAIEmbeddings
            embeddings = OpenAIEmbeddings(
                deployment="your-embeddings-deployment-name",
                model="your-embeddings-model-name",
                openai_api_base="https://your-endpoint.openai.azure.com/",
                openai_api_type="azure",
            )
            text = "This is a test query."
            query_result = embeddings.embed_query(text)

    """

    client: Any  #: :meta private:
    model: str = "text-embedding-ada-002"
    deployment: str = model  # to support Azure OpenAI Service custom deployment names
    openai_api_version: Optional[str] = None
    # to support Azure OpenAI Service custom endpoints
    openai_api_base: Optional[str] = None
    # to support Azure OpenAI Service custom endpoints
    openai_api_type: Optional[str] = None
    # to support explicit proxy for OpenAI
    openai_proxy: Optional[str] = None
    embedding_ctx_length: int = 8191
    """The maximum number of tokens to embed at once."""
    openai_api_key: Optional[str] = None
    openai_organization: Optional[str] = None
    allowed_special: Union[Literal["all"], Set[str]] = set()
    disallowed_special: Union[Literal["all"], Set[str], Sequence[str]] = "all"
    chunk_size: int = 1000
    """Maximum number of texts to embed in each batch"""
    max_retries: int = 6
    """Maximum number of retries to make when generating."""
    request_timeout: Optional[Union[float, Tuple[float, float]]] = None
    """Timeout in seconds for the OpenAPI request."""
    headers: Any = None
    tiktoken_model_name: Optional[str] = None
    """The model name to pass to tiktoken when using this class. 
    Tiktoken is used to count the number of tokens in documents to constrain 
    them to be under a certain limit. By default, when set to None, this will 
    be the same as the embedding model name. However, there are some cases 
    where you may want to use this Embedding class with a model name not 
    supported by tiktoken. This can include when using Azure embeddings or 
    when using one of the many model providers that expose an OpenAI-like 
    API but with different models. In those cases, in order to avoid erroring 
    when tiktoken is called, you can specify a model name to use here."""
    show_progress_bar: bool = False
    """Whether to show a progress bar when embedding."""
    model_kwargs: Dict[str, Any] = Field(default_factory=dict)
    """Holds any model parameters valid for `create` call not explicitly specified."""

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

        extra = Extra.forbid

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

        invalid_model_kwargs = all_required_field_names.intersection(extra.keys())
        if invalid_model_kwargs:
            raise ValueError(
                f"Parameters {invalid_model_kwargs} should be specified explicitly. "
                f"Instead they were passed in as part of `model_kwargs` parameter."
            )

        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."""
        values["openai_api_key"] = get_from_dict_or_env(
            values, "openai_api_key", "OPENAI_API_KEY"
        )
        values["openai_api_base"] = get_from_dict_or_env(
            values,
            "openai_api_base",
            "OPENAI_API_BASE",
            default="",
        )
        values["openai_api_type"] = get_from_dict_or_env(
            values,
            "openai_api_type",
            "OPENAI_API_TYPE",
            default="",
        )
        values["openai_proxy"] = get_from_dict_or_env(
            values,
            "openai_proxy",
            "OPENAI_PROXY",
            default="",
        )
        if values["openai_api_type"] in ("azure", "azure_ad", "azuread"):
            default_api_version = "2022-12-01"
        else:
            default_api_version = ""
        values["openai_api_version"] = get_from_dict_or_env(
            values,
            "openai_api_version",
            "OPENAI_API_VERSION",
            default=default_api_version,
        )
        values["openai_organization"] = get_from_dict_or_env(
            values,
            "openai_organization",
            "OPENAI_ORGANIZATION",
            default="",
        )
        try:
            import llm

            values["client"] = llm.Embedding
        except ImportError:
            raise ImportError(
                "Could not import openai python package. "
                "Please install it with `pip install openai`."
            )
        return values

    @property
    def _invocation_params(self) -> Dict:
        openai_args = {
            "model": self.model,
            "request_timeout": self.request_timeout,
            "headers": self.headers,
            "api_key": self.openai_api_key,
            "organization": self.openai_organization,
            "api_base": self.openai_api_base,
            "api_type": self.openai_api_type,
            "api_version": self.openai_api_version,
            **self.model_kwargs,
        }
        if self.openai_api_type in ("azure", "azure_ad", "azuread"):
            openai_args["engine"] = self.deployment
        if self.openai_proxy:
            import llm

            llm.proxy = {
                "http": self.openai_proxy,
                "https": self.openai_proxy,
            }  # type: ignore[assignment]  # noqa: E501
        return openai_args

    def _get_len_safe_embeddings(
        self, texts: List[str], *, engine: str, chunk_size: Optional[int] = None
    ) -> List[List[float]]:
        embeddings: List[List[float]] = [[] for _ in range(len(texts))]
        try:
            import tiktoken
        except ImportError:
            raise ImportError(
                "Could not import tiktoken python package. "
                "This is needed in order to for OpenAIEmbeddings. "
                "Please install it with `pip install tiktoken`."
            )

        tokens = []
        indices = []
        model_name = self.tiktoken_model_name or self.model
        try:
            encoding = tiktoken.encoding_for_model(model_name)
        except KeyError:
            logger.warning("Warning: model not found. Using cl100k_base encoding.")
            model = "cl100k_base"
            encoding = tiktoken.get_encoding(model)
        for i, text in enumerate(texts):
            if self.model.endswith("001"):
                # See: https://github.com/openai/openai-python/issues/418#issuecomment-1525939500
                # replace newlines, which can negatively affect performance.
                text = text.replace("\n", " ")
            token = encoding.encode(
                text,
                allowed_special=self.allowed_special,
                disallowed_special=self.disallowed_special,
            )
            for j in range(0, len(token), self.embedding_ctx_length):
                tokens.append(token[j : j + self.embedding_ctx_length])
                indices.append(i)

        batched_embeddings: List[List[float]] = []
        _chunk_size = chunk_size or self.chunk_size

        if self.show_progress_bar:
            try:
                import tqdm

                _iter = tqdm.tqdm(range(0, len(tokens), _chunk_size))
            except ImportError:
                _iter = range(0, len(tokens), _chunk_size)
        else:
            _iter = range(0, len(tokens), _chunk_size)

        for i in _iter:
            response = embed_with_retry(
                self,
                input=tokens[i : i + _chunk_size],
                **self._invocation_params,
            )
            batched_embeddings.extend(r["embedding"] for r in response["data"])

        results: List[List[List[float]]] = [[] for _ in range(len(texts))]
        num_tokens_in_batch: List[List[int]] = [[] for _ in range(len(texts))]
        for i in range(len(indices)):
            results[indices[i]].append(batched_embeddings[i])
            num_tokens_in_batch[indices[i]].append(len(tokens[i]))

        for i in range(len(texts)):
            _result = results[i]
            if len(_result) == 0:
                average = embed_with_retry(
                    self,
                    input="",
                    **self._invocation_params,
                )[
                    "data"
                ][0]["embedding"]
            else:
                average = np.average(_result, axis=0, weights=num_tokens_in_batch[i])
            embeddings[i] = (average / np.linalg.norm(average)).tolist()

        return embeddings

    # please refer to
    # https://github.com/openai/openai-cookbook/blob/main/examples/Embedding_long_inputs.ipynb
    async def _aget_len_safe_embeddings(
        self, texts: List[str], *, engine: str, chunk_size: Optional[int] = None
    ) -> List[List[float]]:
        embeddings: List[List[float]] = [[] for _ in range(len(texts))]
        try:
            import tiktoken
        except ImportError:
            raise ImportError(
                "Could not import tiktoken python package. "
                "This is needed in order to for OpenAIEmbeddings. "
                "Please install it with `pip install tiktoken`."
            )

        tokens = []
        indices = []
        model_name = self.tiktoken_model_name or self.model
        try:
            encoding = tiktoken.encoding_for_model(model_name)
        except KeyError:
            logger.warning("Warning: model not found. Using cl100k_base encoding.")
            model = "cl100k_base"
            encoding = tiktoken.get_encoding(model)
        for i, text in enumerate(texts):
            if self.model.endswith("001"):
                # See: https://github.com/openai/openai-python/issues/418#issuecomment-1525939500
                # replace newlines, which can negatively affect performance.
                text = text.replace("\n", " ")
            token = encoding.encode(
                text,
                allowed_special=self.allowed_special,
                disallowed_special=self.disallowed_special,
            )
            for j in range(0, len(token), self.embedding_ctx_length):
                tokens.append(token[j : j + self.embedding_ctx_length])
                indices.append(i)

        batched_embeddings: List[List[float]] = []
        _chunk_size = chunk_size or self.chunk_size
        for i in range(0, len(tokens), _chunk_size):
            response = await async_embed_with_retry(
                self,
                input=tokens[i : i + _chunk_size],
                **self._invocation_params,
            )
            batched_embeddings.extend(r["embedding"] for r in response["data"])

        results: List[List[List[float]]] = [[] for _ in range(len(texts))]
        num_tokens_in_batch: List[List[int]] = [[] for _ in range(len(texts))]
        for i in range(len(indices)):
            results[indices[i]].append(batched_embeddings[i])
            num_tokens_in_batch[indices[i]].append(len(tokens[i]))

        for i in range(len(texts)):
            _result = results[i]
            if len(_result) == 0:
                average = (
                    await async_embed_with_retry(
                        self,
                        input="",
                        **self._invocation_params,
                    )
                )["data"][0]["embedding"]
            else:
                average = np.average(_result, axis=0, weights=num_tokens_in_batch[i])
            embeddings[i] = (average / np.linalg.norm(average)).tolist()

        return embeddings

    def embed_documents(
        self, texts: List[str], chunk_size: Optional[int] = 0
    ) -> List[List[float]]:
        """Call out to OpenAI's embedding endpoint for embedding search docs.

        Args:
            texts: The list of texts to embed.
            chunk_size: The chunk size of embeddings. If None, will use the chunk size
                specified by the class.

        Returns:
            List of embeddings, one for each text.
        """
        # NOTE: to keep things simple, we assume the list may contain texts longer
        #       than the maximum context and use length-safe embedding function.
        return self._get_len_safe_embeddings(texts, engine=self.deployment)

    async def aembed_documents(
        self, texts: List[str], chunk_size: Optional[int] = 0
    ) -> List[List[float]]:
        """Call out to OpenAI's embedding endpoint async for embedding search docs.

        Args:
            texts: The list of texts to embed.
            chunk_size: The chunk size of embeddings. If None, will use the chunk size
                specified by the class.

        Returns:
            List of embeddings, one for each text.
        """
        # NOTE: to keep things simple, we assume the list may contain texts longer
        #       than the maximum context and use length-safe embedding function.
        return await self._aget_len_safe_embeddings(texts, engine=self.deployment)

    def embed_query(self, text: str) -> List[float]:
        """Call out to OpenAI's embedding endpoint for embedding query text.

        Args:
            text: The text to embed.

        Returns:
            Embedding for the text.
        """
        return self.embed_documents([text])[0]

    async def aembed_query(self, text: str) -> List[float]:
        """Call out to OpenAI's embedding endpoint async for embedding query text.

        Args:
            text: The text to embed.

        Returns:
            Embedding for the text.
        """
        embeddings = await self.aembed_documents([text])
        return embeddings[0]