File size: 15,585 Bytes
9c6594c
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
from functools import wraps

import sentry_sdk
from sentry_sdk import consts
from sentry_sdk.ai.monitoring import record_token_usage
from sentry_sdk.ai.utils import set_data_normalized
from sentry_sdk.consts import SPANDATA
from sentry_sdk.integrations import DidNotEnable, Integration
from sentry_sdk.scope import should_send_default_pii
from sentry_sdk.utils import (
    capture_internal_exceptions,
    event_from_exception,
)

from typing import TYPE_CHECKING

if TYPE_CHECKING:
    from typing import Any, Iterable, List, Optional, Callable, AsyncIterator, Iterator
    from sentry_sdk.tracing import Span

try:
    from openai.resources.chat.completions import Completions, AsyncCompletions
    from openai.resources import Embeddings, AsyncEmbeddings

    if TYPE_CHECKING:
        from openai.types.chat import ChatCompletionMessageParam, ChatCompletionChunk
except ImportError:
    raise DidNotEnable("OpenAI not installed")


class OpenAIIntegration(Integration):
    identifier = "openai"
    origin = f"auto.ai.{identifier}"

    def __init__(self, include_prompts=True, tiktoken_encoding_name=None):
        # type: (OpenAIIntegration, bool, Optional[str]) -> None
        self.include_prompts = include_prompts

        self.tiktoken_encoding = None
        if tiktoken_encoding_name is not None:
            import tiktoken  # type: ignore

            self.tiktoken_encoding = tiktoken.get_encoding(tiktoken_encoding_name)

    @staticmethod
    def setup_once():
        # type: () -> None
        Completions.create = _wrap_chat_completion_create(Completions.create)
        Embeddings.create = _wrap_embeddings_create(Embeddings.create)

        AsyncCompletions.create = _wrap_async_chat_completion_create(
            AsyncCompletions.create
        )
        AsyncEmbeddings.create = _wrap_async_embeddings_create(AsyncEmbeddings.create)

    def count_tokens(self, s):
        # type: (OpenAIIntegration, str) -> int
        if self.tiktoken_encoding is not None:
            return len(self.tiktoken_encoding.encode_ordinary(s))
        return 0


def _capture_exception(exc):
    # type: (Any) -> None
    event, hint = event_from_exception(
        exc,
        client_options=sentry_sdk.get_client().options,
        mechanism={"type": "openai", "handled": False},
    )
    sentry_sdk.capture_event(event, hint=hint)


def _calculate_chat_completion_usage(
    messages, response, span, streaming_message_responses, count_tokens
):
    # type: (Iterable[ChatCompletionMessageParam], Any, Span, Optional[List[str]], Callable[..., Any]) -> None
    completion_tokens = 0  # type: Optional[int]
    prompt_tokens = 0  # type: Optional[int]
    total_tokens = 0  # type: Optional[int]
    if hasattr(response, "usage"):
        if hasattr(response.usage, "completion_tokens") and isinstance(
            response.usage.completion_tokens, int
        ):
            completion_tokens = response.usage.completion_tokens
        if hasattr(response.usage, "prompt_tokens") and isinstance(
            response.usage.prompt_tokens, int
        ):
            prompt_tokens = response.usage.prompt_tokens
        if hasattr(response.usage, "total_tokens") and isinstance(
            response.usage.total_tokens, int
        ):
            total_tokens = response.usage.total_tokens

    if prompt_tokens == 0:
        for message in messages:
            if "content" in message:
                prompt_tokens += count_tokens(message["content"])

    if completion_tokens == 0:
        if streaming_message_responses is not None:
            for message in streaming_message_responses:
                completion_tokens += count_tokens(message)
        elif hasattr(response, "choices"):
            for choice in response.choices:
                if hasattr(choice, "message"):
                    completion_tokens += count_tokens(choice.message)

    if prompt_tokens == 0:
        prompt_tokens = None
    if completion_tokens == 0:
        completion_tokens = None
    if total_tokens == 0:
        total_tokens = None
    record_token_usage(span, prompt_tokens, completion_tokens, total_tokens)


def _new_chat_completion_common(f, *args, **kwargs):
    # type: (Any, *Any, **Any) -> Any
    integration = sentry_sdk.get_client().get_integration(OpenAIIntegration)
    if integration is None:
        return f(*args, **kwargs)

    if "messages" not in kwargs:
        # invalid call (in all versions of openai), let it return error
        return f(*args, **kwargs)

    try:
        iter(kwargs["messages"])
    except TypeError:
        # invalid call (in all versions), messages must be iterable
        return f(*args, **kwargs)

    kwargs["messages"] = list(kwargs["messages"])
    messages = kwargs["messages"]
    model = kwargs.get("model")
    streaming = kwargs.get("stream")

    span = sentry_sdk.start_span(
        op=consts.OP.OPENAI_CHAT_COMPLETIONS_CREATE,
        name="Chat Completion",
        origin=OpenAIIntegration.origin,
    )
    span.__enter__()

    res = yield f, args, kwargs

    with capture_internal_exceptions():
        if should_send_default_pii() and integration.include_prompts:
            set_data_normalized(span, SPANDATA.AI_INPUT_MESSAGES, messages)

        set_data_normalized(span, SPANDATA.AI_MODEL_ID, model)
        set_data_normalized(span, SPANDATA.AI_STREAMING, streaming)

        if hasattr(res, "choices"):
            if should_send_default_pii() and integration.include_prompts:
                set_data_normalized(
                    span,
                    SPANDATA.AI_RESPONSES,
                    list(map(lambda x: x.message, res.choices)),
                )
            _calculate_chat_completion_usage(
                messages, res, span, None, integration.count_tokens
            )
            span.__exit__(None, None, None)
        elif hasattr(res, "_iterator"):
            data_buf: list[list[str]] = []  # one for each choice

            old_iterator = res._iterator

            def new_iterator():
                # type: () -> Iterator[ChatCompletionChunk]
                with capture_internal_exceptions():
                    for x in old_iterator:
                        if hasattr(x, "choices"):
                            choice_index = 0
                            for choice in x.choices:
                                if hasattr(choice, "delta") and hasattr(
                                    choice.delta, "content"
                                ):
                                    content = choice.delta.content
                                    if len(data_buf) <= choice_index:
                                        data_buf.append([])
                                    data_buf[choice_index].append(content or "")
                                choice_index += 1
                        yield x
                    if len(data_buf) > 0:
                        all_responses = list(
                            map(lambda chunk: "".join(chunk), data_buf)
                        )
                        if should_send_default_pii() and integration.include_prompts:
                            set_data_normalized(
                                span, SPANDATA.AI_RESPONSES, all_responses
                            )
                        _calculate_chat_completion_usage(
                            messages,
                            res,
                            span,
                            all_responses,
                            integration.count_tokens,
                        )
                span.__exit__(None, None, None)

            async def new_iterator_async():
                # type: () -> AsyncIterator[ChatCompletionChunk]
                with capture_internal_exceptions():
                    async for x in old_iterator:
                        if hasattr(x, "choices"):
                            choice_index = 0
                            for choice in x.choices:
                                if hasattr(choice, "delta") and hasattr(
                                    choice.delta, "content"
                                ):
                                    content = choice.delta.content
                                    if len(data_buf) <= choice_index:
                                        data_buf.append([])
                                    data_buf[choice_index].append(content or "")
                                choice_index += 1
                        yield x
                    if len(data_buf) > 0:
                        all_responses = list(
                            map(lambda chunk: "".join(chunk), data_buf)
                        )
                        if should_send_default_pii() and integration.include_prompts:
                            set_data_normalized(
                                span, SPANDATA.AI_RESPONSES, all_responses
                            )
                        _calculate_chat_completion_usage(
                            messages,
                            res,
                            span,
                            all_responses,
                            integration.count_tokens,
                        )
                span.__exit__(None, None, None)

            if str(type(res._iterator)) == "<class 'async_generator'>":
                res._iterator = new_iterator_async()
            else:
                res._iterator = new_iterator()

        else:
            set_data_normalized(span, "unknown_response", True)
            span.__exit__(None, None, None)
    return res


def _wrap_chat_completion_create(f):
    # type: (Callable[..., Any]) -> Callable[..., Any]
    def _execute_sync(f, *args, **kwargs):
        # type: (Any, *Any, **Any) -> Any
        gen = _new_chat_completion_common(f, *args, **kwargs)

        try:
            f, args, kwargs = next(gen)
        except StopIteration as e:
            return e.value

        try:
            try:
                result = f(*args, **kwargs)
            except Exception as e:
                _capture_exception(e)
                raise e from None

            return gen.send(result)
        except StopIteration as e:
            return e.value

    @wraps(f)
    def _sentry_patched_create_sync(*args, **kwargs):
        # type: (*Any, **Any) -> Any
        integration = sentry_sdk.get_client().get_integration(OpenAIIntegration)
        if integration is None or "messages" not in kwargs:
            # no "messages" means invalid call (in all versions of openai), let it return error
            return f(*args, **kwargs)

        return _execute_sync(f, *args, **kwargs)

    return _sentry_patched_create_sync


def _wrap_async_chat_completion_create(f):
    # type: (Callable[..., Any]) -> Callable[..., Any]
    async def _execute_async(f, *args, **kwargs):
        # type: (Any, *Any, **Any) -> Any
        gen = _new_chat_completion_common(f, *args, **kwargs)

        try:
            f, args, kwargs = next(gen)
        except StopIteration as e:
            return await e.value

        try:
            try:
                result = await f(*args, **kwargs)
            except Exception as e:
                _capture_exception(e)
                raise e from None

            return gen.send(result)
        except StopIteration as e:
            return e.value

    @wraps(f)
    async def _sentry_patched_create_async(*args, **kwargs):
        # type: (*Any, **Any) -> Any
        integration = sentry_sdk.get_client().get_integration(OpenAIIntegration)
        if integration is None or "messages" not in kwargs:
            # no "messages" means invalid call (in all versions of openai), let it return error
            return await f(*args, **kwargs)

        return await _execute_async(f, *args, **kwargs)

    return _sentry_patched_create_async


def _new_embeddings_create_common(f, *args, **kwargs):
    # type: (Any, *Any, **Any) -> Any
    integration = sentry_sdk.get_client().get_integration(OpenAIIntegration)
    if integration is None:
        return f(*args, **kwargs)

    with sentry_sdk.start_span(
        op=consts.OP.OPENAI_EMBEDDINGS_CREATE,
        description="OpenAI Embedding Creation",
        origin=OpenAIIntegration.origin,
    ) as span:
        if "input" in kwargs and (
            should_send_default_pii() and integration.include_prompts
        ):
            if isinstance(kwargs["input"], str):
                set_data_normalized(span, SPANDATA.AI_INPUT_MESSAGES, [kwargs["input"]])
            elif (
                isinstance(kwargs["input"], list)
                and len(kwargs["input"]) > 0
                and isinstance(kwargs["input"][0], str)
            ):
                set_data_normalized(span, SPANDATA.AI_INPUT_MESSAGES, kwargs["input"])
        if "model" in kwargs:
            set_data_normalized(span, SPANDATA.AI_MODEL_ID, kwargs["model"])

        response = yield f, args, kwargs

        prompt_tokens = 0
        total_tokens = 0
        if hasattr(response, "usage"):
            if hasattr(response.usage, "prompt_tokens") and isinstance(
                response.usage.prompt_tokens, int
            ):
                prompt_tokens = response.usage.prompt_tokens
            if hasattr(response.usage, "total_tokens") and isinstance(
                response.usage.total_tokens, int
            ):
                total_tokens = response.usage.total_tokens

        if prompt_tokens == 0:
            prompt_tokens = integration.count_tokens(kwargs["input"] or "")

        record_token_usage(span, prompt_tokens, None, total_tokens or prompt_tokens)

        return response


def _wrap_embeddings_create(f):
    # type: (Any) -> Any
    def _execute_sync(f, *args, **kwargs):
        # type: (Any, *Any, **Any) -> Any
        gen = _new_embeddings_create_common(f, *args, **kwargs)

        try:
            f, args, kwargs = next(gen)
        except StopIteration as e:
            return e.value

        try:
            try:
                result = f(*args, **kwargs)
            except Exception as e:
                _capture_exception(e)
                raise e from None

            return gen.send(result)
        except StopIteration as e:
            return e.value

    @wraps(f)
    def _sentry_patched_create_sync(*args, **kwargs):
        # type: (*Any, **Any) -> Any
        integration = sentry_sdk.get_client().get_integration(OpenAIIntegration)
        if integration is None:
            return f(*args, **kwargs)

        return _execute_sync(f, *args, **kwargs)

    return _sentry_patched_create_sync


def _wrap_async_embeddings_create(f):
    # type: (Any) -> Any
    async def _execute_async(f, *args, **kwargs):
        # type: (Any, *Any, **Any) -> Any
        gen = _new_embeddings_create_common(f, *args, **kwargs)

        try:
            f, args, kwargs = next(gen)
        except StopIteration as e:
            return await e.value

        try:
            try:
                result = await f(*args, **kwargs)
            except Exception as e:
                _capture_exception(e)
                raise e from None

            return gen.send(result)
        except StopIteration as e:
            return e.value

    @wraps(f)
    async def _sentry_patched_create_async(*args, **kwargs):
        # type: (*Any, **Any) -> Any
        integration = sentry_sdk.get_client().get_integration(OpenAIIntegration)
        if integration is None:
            return await f(*args, **kwargs)

        return await _execute_async(f, *args, **kwargs)

    return _sentry_patched_create_async