protocol.py 26.6 KB
Newer Older
1
2
# Adapted from
# https://github.com/lm-sys/FastChat/blob/168ccc29d3f7edc50823016105c024fe2282732a/fastchat/protocol/openai_api_protocol.py
Zhuohan Li's avatar
Zhuohan Li committed
3
import time
4
from typing import Any, Dict, List, Literal, Optional, Union
Zhuohan Li's avatar
Zhuohan Li committed
5

6
import openai.types.chat
7
import torch
8
from pydantic import BaseModel, ConfigDict, Field, model_validator
9
10
# pydantic needs the TypedDict from typing_extensions
from typing_extensions import Annotated, Required, TypedDict
Zhuohan Li's avatar
Zhuohan Li committed
11

12
from vllm.pooling_params import PoolingParams
13
from vllm.sampling_params import SamplingParams
14
from vllm.utils import random_uuid
15

Zhuohan Li's avatar
Zhuohan Li committed
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
class CustomChatCompletionContentPartParam(TypedDict, total=False):
    __pydantic_config__ = ConfigDict(extra="allow")  # type: ignore

    type: Required[str]
    """The type of the content part."""


ChatCompletionContentPartParam = Union[
    openai.types.chat.ChatCompletionContentPartParam,
    CustomChatCompletionContentPartParam]


class CustomChatCompletionMessageParam(TypedDict, total=False):
    """Enables custom roles in the Chat Completion API."""
    role: Required[str]
    """The role of the message's author."""

    content: Union[str, List[ChatCompletionContentPartParam]]
    """The contents of the message."""

    name: str
    """An optional name for the participant.

    Provides the model information to differentiate between participants of the
    same role.
    """


ChatCompletionMessageParam = Union[
    openai.types.chat.ChatCompletionMessageParam,
    CustomChatCompletionMessageParam]


50
51
52
53
54
55
class OpenAIBaseModel(BaseModel):
    # OpenAI API does not allow extra fields
    model_config = ConfigDict(extra="forbid")


class ErrorResponse(OpenAIBaseModel):
Zhuohan Li's avatar
Zhuohan Li committed
56
57
58
59
    object: str = "error"
    message: str
    type: str
    param: Optional[str] = None
60
    code: int
Zhuohan Li's avatar
Zhuohan Li committed
61
62


63
class ModelPermission(OpenAIBaseModel):
Zhuohan Li's avatar
Zhuohan Li committed
64
65
66
67
68
69
70
71
72
73
74
    id: str = Field(default_factory=lambda: f"modelperm-{random_uuid()}")
    object: str = "model_permission"
    created: int = Field(default_factory=lambda: int(time.time()))
    allow_create_engine: bool = False
    allow_sampling: bool = True
    allow_logprobs: bool = True
    allow_search_indices: bool = False
    allow_view: bool = True
    allow_fine_tuning: bool = False
    organization: str = "*"
    group: Optional[str] = None
75
    is_blocking: bool = False
Zhuohan Li's avatar
Zhuohan Li committed
76
77


78
class ModelCard(OpenAIBaseModel):
Zhuohan Li's avatar
Zhuohan Li committed
79
80
81
    id: str
    object: str = "model"
    created: int = Field(default_factory=lambda: int(time.time()))
Woosuk Kwon's avatar
Woosuk Kwon committed
82
    owned_by: str = "vllm"
Zhuohan Li's avatar
Zhuohan Li committed
83
84
    root: Optional[str] = None
    parent: Optional[str] = None
85
    max_model_len: Optional[int] = None
Zhuohan Li's avatar
Zhuohan Li committed
86
87
88
    permission: List[ModelPermission] = Field(default_factory=list)


89
class ModelList(OpenAIBaseModel):
Zhuohan Li's avatar
Zhuohan Li committed
90
91
92
93
    object: str = "list"
    data: List[ModelCard] = Field(default_factory=list)


94
class UsageInfo(OpenAIBaseModel):
Zhuohan Li's avatar
Zhuohan Li committed
95
96
97
98
99
    prompt_tokens: int = 0
    total_tokens: int = 0
    completion_tokens: Optional[int] = 0


100
class ResponseFormat(OpenAIBaseModel):
101
    # type must be "json_object" or "text"
102
    type: Literal["text", "json_object"]
103
104


105
class StreamOptions(OpenAIBaseModel):
106
107
    include_usage: Optional[bool] = True
    continuous_usage_stats: Optional[bool] = True
108
109


110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
class FunctionDefinition(OpenAIBaseModel):
    name: str
    description: Optional[str] = None
    parameters: Optional[Dict[str, Any]] = None


class ChatCompletionToolsParam(OpenAIBaseModel):
    type: Literal["function"] = "function"
    function: FunctionDefinition


class ChatCompletionNamedFunction(OpenAIBaseModel):
    name: str


class ChatCompletionNamedToolChoiceParam(OpenAIBaseModel):
    function: ChatCompletionNamedFunction
    type: Literal["function"] = "function"


130
class ChatCompletionRequest(OpenAIBaseModel):
131
132
    # Ordered by official OpenAI API documentation
    # https://platform.openai.com/docs/api-reference/chat/create
133
    messages: List[ChatCompletionMessageParam]
134
135
136
137
    model: str
    frequency_penalty: Optional[float] = 0.0
    logit_bias: Optional[Dict[str, float]] = None
    logprobs: Optional[bool] = False
138
    top_logprobs: Optional[int] = 0
139
    max_tokens: Optional[int] = None
140
141
142
    n: Optional[int] = 1
    presence_penalty: Optional[float] = 0.0
    response_format: Optional[ResponseFormat] = None
143
144
145
    seed: Optional[int] = Field(None,
                                ge=torch.iinfo(torch.long).min,
                                le=torch.iinfo(torch.long).max)
146
    stop: Optional[Union[str, List[str]]] = Field(default_factory=list)
Zhuohan Li's avatar
Zhuohan Li committed
147
    stream: Optional[bool] = False
148
    stream_options: Optional[StreamOptions] = None
149
150
    temperature: Optional[float] = 0.7
    top_p: Optional[float] = 1.0
151
152
153
    tools: Optional[List[ChatCompletionToolsParam]] = None
    tool_choice: Optional[Union[Literal["none"],
                                ChatCompletionNamedToolChoiceParam]] = "none"
Zhuohan Li's avatar
Zhuohan Li committed
154
    user: Optional[str] = None
155
156

    # doc: begin-chat-completion-sampling-params
157
158
    best_of: Optional[int] = None
    use_beam_search: Optional[bool] = False
159
160
161
162
    top_k: Optional[int] = -1
    min_p: Optional[float] = 0.0
    repetition_penalty: Optional[float] = 1.0
    length_penalty: Optional[float] = 1.0
163
    early_stopping: Optional[bool] = False
164
    ignore_eos: Optional[bool] = False
165
    min_tokens: Optional[int] = 0
166
    stop_token_ids: Optional[List[int]] = Field(default_factory=list)
167
    skip_special_tokens: Optional[bool] = True
168
    spaces_between_special_tokens: Optional[bool] = True
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
    # doc: end-chat-completion-sampling-params

    # doc: begin-chat-completion-extra-params
    echo: Optional[bool] = Field(
        default=False,
        description=(
            "If true, the new message will be prepended with the last message "
            "if they belong to the same role."),
    )
    add_generation_prompt: Optional[bool] = Field(
        default=True,
        description=
        ("If true, the generation prompt will be added to the chat template. "
         "This is a parameter used by chat template in tokenizer config of the "
         "model."),
    )
185
186
187
188
189
190
191
192
193
    add_special_tokens: Optional[bool] = Field(
        default=False,
        description=(
            "If true, special tokens (e.g. BOS) will be added to the prompt "
            "on top of what is added by the chat template. "
            "For most models, the chat template takes care of adding the "
            "special tokens so this should be set to False (as is the "
            "default)."),
    )
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
    documents: Optional[List[Dict[str, str]]] = Field(
        default=None,
        description=
        ("A list of dicts representing documents that will be accessible to "
         "the model if it is performing RAG (retrieval-augmented generation)."
         " If the template does not support RAG, this argument will have no "
         "effect. We recommend that each document should be a dict containing "
         "\"title\" and \"text\" keys."),
    )
    chat_template: Optional[str] = Field(
        default=None,
        description=(
            "A Jinja template to use for this conversion. "
            "If this is not passed, the model's default chat template will be "
            "used instead."),
    )
    chat_template_kwargs: Optional[Dict[str, Any]] = Field(
        default=None,
        description=("Additional kwargs to pass to the template renderer. "
                     "Will be accessible by the chat template."),
    )
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
    include_stop_str_in_output: Optional[bool] = Field(
        default=False,
        description=(
            "Whether to include the stop string in the output. "
            "This is only applied when the stop or stop_token_ids is set."),
    )
    guided_json: Optional[Union[str, dict, BaseModel]] = Field(
        default=None,
        description=("If specified, the output will follow the JSON schema."),
    )
    guided_regex: Optional[str] = Field(
        default=None,
        description=(
            "If specified, the output will follow the regex pattern."),
    )
    guided_choice: Optional[List[str]] = Field(
        default=None,
        description=(
            "If specified, the output will be exactly one of the choices."),
    )
    guided_grammar: Optional[str] = Field(
        default=None,
        description=(
            "If specified, the output will follow the context free grammar."),
    )
240
241
242
243
244
245
    guided_decoding_backend: Optional[str] = Field(
        default=None,
        description=(
            "If specified, will override the default guided decoding backend "
            "of the server for this specific request. If set, must be either "
            "'outlines' / 'lm-format-enforcer'"))
246
247
248
249
250
    guided_whitespace_pattern: Optional[str] = Field(
        default=None,
        description=(
            "If specified, will override the default whitespace pattern "
            "for guided json decoding."))
251
252

    # doc: end-chat-completion-extra-params
Zhuohan Li's avatar
Zhuohan Li committed
253

254
    def to_sampling_params(self) -> SamplingParams:
255
        # We now allow logprobs being true without top_logrobs.
256
257
258

        logits_processors = None
        if self.logit_bias:
259
260
261
262
263
264
265
266
267
268
            logit_bias: Dict[int, float] = {}
            try:
                for token_id, bias in self.logit_bias.items():
                    # Convert token_id to integer before we add to LLMEngine
                    # Clamp the bias between -100 and 100 per OpenAI API spec
                    logit_bias[int(token_id)] = min(100, max(-100, bias))
            except ValueError as exc:
                raise ValueError(f"Found token_id `{token_id}` in logit_bias "
                                 f"but token_id must be an integer or string "
                                 f"representing an integer") from exc
269
270
271
272

            def logit_bias_logits_processor(
                    token_ids: List[int],
                    logits: torch.Tensor) -> torch.Tensor:
273
274
                for token_id, bias in logit_bias.items():
                    logits[token_id] += bias
275
276
277
278
                return logits

            logits_processors = [logit_bias_logits_processor]

279
280
281
282
283
284
285
286
        return SamplingParams(
            n=self.n,
            presence_penalty=self.presence_penalty,
            frequency_penalty=self.frequency_penalty,
            repetition_penalty=self.repetition_penalty,
            temperature=self.temperature,
            top_p=self.top_p,
            min_p=self.min_p,
Nick Hill's avatar
Nick Hill committed
287
            seed=self.seed,
288
289
290
            stop=self.stop,
            stop_token_ids=self.stop_token_ids,
            max_tokens=self.max_tokens,
291
            min_tokens=self.min_tokens,
292
293
            logprobs=self.top_logprobs if self.logprobs else None,
            prompt_logprobs=self.top_logprobs if self.echo else None,
294
295
296
297
            best_of=self.best_of,
            top_k=self.top_k,
            ignore_eos=self.ignore_eos,
            use_beam_search=self.use_beam_search,
298
            early_stopping=self.early_stopping,
299
300
            skip_special_tokens=self.skip_special_tokens,
            spaces_between_special_tokens=self.spaces_between_special_tokens,
301
302
            include_stop_str_in_output=self.include_stop_str_in_output,
            length_penalty=self.length_penalty,
303
            logits_processors=logits_processors,
304
305
        )

306
307
308
309
310
311
312
313
314
    @model_validator(mode='before')
    @classmethod
    def validate_stream_options(cls, values):
        if (values.get('stream_options') is not None
                and not values.get('stream')):
            raise ValueError(
                "stream_options can only be set if stream is true")
        return values

315
316
317
318
319
320
321
322
    @model_validator(mode="before")
    @classmethod
    def check_guided_decoding_count(cls, data):
        guide_count = sum([
            "guided_json" in data and data["guided_json"] is not None,
            "guided_regex" in data and data["guided_regex"] is not None,
            "guided_choice" in data and data["guided_choice"] is not None
        ])
323
        # you can only use one kind of guided decoding
324
325
326
327
        if guide_count > 1:
            raise ValueError(
                "You can only use one kind of guided decoding "
                "('guided_json', 'guided_regex' or 'guided_choice').")
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
        # you can only either use guided decoding or tools, not both
        if guide_count > 1 and "tool_choice" in data and data[
                "tool_choice"] != "none":
            raise ValueError(
                "You can only either use guided decoding or tools, not both.")
        return data

    @model_validator(mode="before")
    @classmethod
    def check_tool_choice(cls, data):
        if "tool_choice" in data and data["tool_choice"] != "none":
            if not isinstance(data["tool_choice"], dict):
                raise ValueError("Currently only named tools are supported.")
            if "tools" not in data or data["tools"] is None:
                raise ValueError(
                    "When using `tool_choice`, `tools` must be set.")
344
345
        return data

346
347
348
349
350
351
352
353
    @model_validator(mode="before")
    @classmethod
    def check_logprobs(cls, data):
        if "top_logprobs" in data and data["top_logprobs"] is not None:
            if "logprobs" not in data or data["logprobs"] is False:
                raise ValueError(
                    "when using `top_logprobs`, `logprobs` must be set to true."
                )
354
            elif data["top_logprobs"] < 0:
355
                raise ValueError(
356
                    "`top_logprobs` must be a value a positive value.")
357
358
        return data

Zhuohan Li's avatar
Zhuohan Li committed
359

360
class CompletionRequest(OpenAIBaseModel):
361
362
    # Ordered by official OpenAI API documentation
    # https://platform.openai.com/docs/api-reference/completions/create
Zhuohan Li's avatar
Zhuohan Li committed
363
    model: str
364
    prompt: Union[List[int], List[List[int]], str, List[str]]
365
    best_of: Optional[int] = None
Zhuohan Li's avatar
Zhuohan Li committed
366
367
368
    echo: Optional[bool] = False
    frequency_penalty: Optional[float] = 0.0
    logit_bias: Optional[Dict[str, float]] = None
369
370
    logprobs: Optional[int] = None
    max_tokens: Optional[int] = 16
371
    n: int = 1
372
    presence_penalty: Optional[float] = 0.0
373
374
375
    seed: Optional[int] = Field(None,
                                ge=torch.iinfo(torch.long).min,
                                le=torch.iinfo(torch.long).max)
376
377
    stop: Optional[Union[str, List[str]]] = Field(default_factory=list)
    stream: Optional[bool] = False
378
    stream_options: Optional[StreamOptions] = None
379
380
381
    suffix: Optional[str] = None
    temperature: Optional[float] = 1.0
    top_p: Optional[float] = 1.0
Zhuohan Li's avatar
Zhuohan Li committed
382
    user: Optional[str] = None
383
384

    # doc: begin-completion-sampling-params
Zhuohan Li's avatar
Zhuohan Li committed
385
    use_beam_search: Optional[bool] = False
386
387
388
389
    top_k: Optional[int] = -1
    min_p: Optional[float] = 0.0
    repetition_penalty: Optional[float] = 1.0
    length_penalty: Optional[float] = 1.0
390
    early_stopping: Optional[bool] = False
391
    stop_token_ids: Optional[List[int]] = Field(default_factory=list)
392
    ignore_eos: Optional[bool] = False
393
    min_tokens: Optional[int] = 0
394
    skip_special_tokens: Optional[bool] = True
395
    spaces_between_special_tokens: Optional[bool] = True
396
    truncate_prompt_tokens: Optional[Annotated[int, Field(ge=1)]] = None
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
    # doc: end-completion-sampling-params

    # doc: begin-completion-extra-params
    include_stop_str_in_output: Optional[bool] = Field(
        default=False,
        description=(
            "Whether to include the stop string in the output. "
            "This is only applied when the stop or stop_token_ids is set."),
    )
    response_format: Optional[ResponseFormat] = Field(
        default=None,
        description=
        ("Similar to chat completion, this parameter specifies the format of "
         "output. Only {'type': 'json_object'} or {'type': 'text' } is "
         "supported."),
    )
    guided_json: Optional[Union[str, dict, BaseModel]] = Field(
        default=None,
        description=("If specified, the output will follow the JSON schema."),
    )
    guided_regex: Optional[str] = Field(
        default=None,
        description=(
            "If specified, the output will follow the regex pattern."),
    )
    guided_choice: Optional[List[str]] = Field(
        default=None,
        description=(
            "If specified, the output will be exactly one of the choices."),
    )
    guided_grammar: Optional[str] = Field(
        default=None,
        description=(
            "If specified, the output will follow the context free grammar."),
    )
432
433
434
435
436
437
    guided_decoding_backend: Optional[str] = Field(
        default=None,
        description=(
            "If specified, will override the default guided decoding backend "
            "of the server for this specific request. If set, must be one of "
            "'outlines' / 'lm-format-enforcer'"))
438
439
440
441
442
    guided_whitespace_pattern: Optional[str] = Field(
        default=None,
        description=(
            "If specified, will override the default whitespace pattern "
            "for guided json decoding."))
443
444

    # doc: end-completion-extra-params
Zhuohan Li's avatar
Zhuohan Li committed
445

446
447
448
    def to_sampling_params(self):
        echo_without_generation = self.echo and self.max_tokens == 0

449
450
        logits_processors = None
        if self.logit_bias:
451
452
453
454
455
456
457
458
459
460
            logit_bias: Dict[int, float] = {}
            try:
                for token_id, bias in self.logit_bias.items():
                    # Convert token_id to integer
                    # Clamp the bias between -100 and 100 per OpenAI API spec
                    logit_bias[int(token_id)] = min(100, max(-100, bias))
            except ValueError as exc:
                raise ValueError(f"Found token_id `{token_id}` in logit_bias "
                                 f"but token_id must be an integer or string "
                                 f"representing an integer") from exc
461
462
463
464

            def logit_bias_logits_processor(
                    token_ids: List[int],
                    logits: torch.Tensor) -> torch.Tensor:
465
466
                for token_id, bias in logit_bias.items():
                    logits[token_id] += bias
467
468
469
470
                return logits

            logits_processors = [logit_bias_logits_processor]

471
472
473
474
475
476
477
478
479
480
        return SamplingParams(
            n=self.n,
            best_of=self.best_of,
            presence_penalty=self.presence_penalty,
            frequency_penalty=self.frequency_penalty,
            repetition_penalty=self.repetition_penalty,
            temperature=self.temperature,
            top_p=self.top_p,
            top_k=self.top_k,
            min_p=self.min_p,
Nick Hill's avatar
Nick Hill committed
481
            seed=self.seed,
482
483
484
485
            stop=self.stop,
            stop_token_ids=self.stop_token_ids,
            ignore_eos=self.ignore_eos,
            max_tokens=self.max_tokens if not echo_without_generation else 1,
486
            min_tokens=self.min_tokens,
487
488
            logprobs=self.logprobs,
            use_beam_search=self.use_beam_search,
489
            early_stopping=self.early_stopping,
490
491
492
            prompt_logprobs=self.logprobs if self.echo else None,
            skip_special_tokens=self.skip_special_tokens,
            spaces_between_special_tokens=(self.spaces_between_special_tokens),
493
494
            include_stop_str_in_output=self.include_stop_str_in_output,
            length_penalty=self.length_penalty,
495
            logits_processors=logits_processors,
496
            truncate_prompt_tokens=self.truncate_prompt_tokens,
497
498
        )

499
500
501
502
503
504
505
506
507
508
509
510
511
512
    @model_validator(mode="before")
    @classmethod
    def check_guided_decoding_count(cls, data):
        guide_count = sum([
            "guided_json" in data and data["guided_json"] is not None,
            "guided_regex" in data and data["guided_regex"] is not None,
            "guided_choice" in data and data["guided_choice"] is not None
        ])
        if guide_count > 1:
            raise ValueError(
                "You can only use one kind of guided decoding "
                "('guided_json', 'guided_regex' or 'guided_choice').")
        return data

513
514
515
516
    @model_validator(mode="before")
    @classmethod
    def check_logprobs(cls, data):
        if "logprobs" in data and data[
517
518
                "logprobs"] is not None and not data["logprobs"] >= 0:
            raise ValueError("if passed, `logprobs` must be a positive value.")
519
520
        return data

521
522
523
524
525
526
527
528
    @model_validator(mode="before")
    @classmethod
    def validate_stream_options(cls, data):
        if data.get("stream_options") and not data.get("stream"):
            raise ValueError(
                "Stream options can only be defined when stream is True.")
        return data

Zhuohan Li's avatar
Zhuohan Li committed
529

530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
class EmbeddingRequest(BaseModel):
    # Ordered by official OpenAI API documentation
    # https://platform.openai.com/docs/api-reference/embeddings
    model: str
    input: Union[List[int], List[List[int]], str, List[str]]
    encoding_format: Optional[str] = Field('float', pattern='^(float|base64)$')
    dimensions: Optional[int] = None
    user: Optional[str] = None

    # doc: begin-embedding-pooling-params
    additional_data: Optional[Any] = None

    # doc: end-embedding-pooling-params

    def to_pooling_params(self):
        return PoolingParams(additional_data=self.additional_data)


548
class CompletionLogProbs(OpenAIBaseModel):
Zhuohan Li's avatar
Zhuohan Li committed
549
550
551
    text_offset: List[int] = Field(default_factory=list)
    token_logprobs: List[Optional[float]] = Field(default_factory=list)
    tokens: List[str] = Field(default_factory=list)
552
553
    top_logprobs: List[Optional[Dict[str,
                                     float]]] = Field(default_factory=list)
Zhuohan Li's avatar
Zhuohan Li committed
554
555


556
class CompletionResponseChoice(OpenAIBaseModel):
Zhuohan Li's avatar
Zhuohan Li committed
557
558
    index: int
    text: str
559
    logprobs: Optional[CompletionLogProbs] = None
560
561
    finish_reason: Optional[str] = None
    stop_reason: Optional[Union[int, str]] = Field(
562
563
564
565
566
567
        default=None,
        description=(
            "The stop string or token id that caused the completion "
            "to stop, None if the completion finished for some other reason "
            "including encountering the EOS token"),
    )
Zhuohan Li's avatar
Zhuohan Li committed
568
569


570
class CompletionResponse(OpenAIBaseModel):
Zhuohan Li's avatar
Zhuohan Li committed
571
572
573
574
575
576
577
578
    id: str = Field(default_factory=lambda: f"cmpl-{random_uuid()}")
    object: str = "text_completion"
    created: int = Field(default_factory=lambda: int(time.time()))
    model: str
    choices: List[CompletionResponseChoice]
    usage: UsageInfo


579
class CompletionResponseStreamChoice(OpenAIBaseModel):
Zhuohan Li's avatar
Zhuohan Li committed
580
581
    index: int
    text: str
582
    logprobs: Optional[CompletionLogProbs] = None
583
584
    finish_reason: Optional[str] = None
    stop_reason: Optional[Union[int, str]] = Field(
585
586
587
588
589
590
        default=None,
        description=(
            "The stop string or token id that caused the completion "
            "to stop, None if the completion finished for some other reason "
            "including encountering the EOS token"),
    )
Zhuohan Li's avatar
Zhuohan Li committed
591
592


593
class CompletionStreamResponse(OpenAIBaseModel):
Zhuohan Li's avatar
Zhuohan Li committed
594
595
596
597
598
    id: str = Field(default_factory=lambda: f"cmpl-{random_uuid()}")
    object: str = "text_completion"
    created: int = Field(default_factory=lambda: int(time.time()))
    model: str
    choices: List[CompletionResponseStreamChoice]
599
    usage: Optional[UsageInfo] = Field(default=None)
600
601


602
603
604
class EmbeddingResponseData(BaseModel):
    index: int
    object: str = "embedding"
605
    embedding: Union[List[float], str]
606
607
608
609
610
611
612
613
614
615
616


class EmbeddingResponse(BaseModel):
    id: str = Field(default_factory=lambda: f"cmpl-{random_uuid()}")
    object: str = "list"
    created: int = Field(default_factory=lambda: int(time.time()))
    model: str
    data: List[EmbeddingResponseData]
    usage: UsageInfo


617
618
619
620
621
622
623
624
625
626
627
class FunctionCall(OpenAIBaseModel):
    name: str
    arguments: str


class ToolCall(OpenAIBaseModel):
    id: str = Field(default_factory=lambda: f"chatcmpl-tool-{random_uuid()}")
    type: Literal["function"] = "function"
    function: FunctionCall


628
class ChatMessage(OpenAIBaseModel):
629
630
    role: str
    content: str
631
    tool_calls: List[ToolCall] = Field(default_factory=list)
632
633


634
635
636
637
638
639
640
641
642
643
644
645
646
647
class ChatCompletionLogProb(OpenAIBaseModel):
    token: str
    logprob: float = -9999.0
    bytes: Optional[List[int]] = None


class ChatCompletionLogProbsContent(ChatCompletionLogProb):
    top_logprobs: List[ChatCompletionLogProb] = Field(default_factory=list)


class ChatCompletionLogProbs(OpenAIBaseModel):
    content: Optional[List[ChatCompletionLogProbsContent]] = None


648
class ChatCompletionResponseChoice(OpenAIBaseModel):
649
650
    index: int
    message: ChatMessage
651
    logprobs: Optional[ChatCompletionLogProbs] = None
652
    finish_reason: Optional[str] = None
653
    stop_reason: Optional[Union[int, str]] = None
654
655


656
class ChatCompletionResponse(OpenAIBaseModel):
657
    id: str = Field(default_factory=lambda: f"chatcmpl-{random_uuid()}")
658
    object: Literal["chat.completion"] = "chat.completion"
659
660
661
662
663
664
    created: int = Field(default_factory=lambda: int(time.time()))
    model: str
    choices: List[ChatCompletionResponseChoice]
    usage: UsageInfo


665
class DeltaMessage(OpenAIBaseModel):
666
667
    role: Optional[str] = None
    content: Optional[str] = None
668
    tool_calls: List[ToolCall] = Field(default_factory=list)
669
670


671
class ChatCompletionResponseStreamChoice(OpenAIBaseModel):
672
673
    index: int
    delta: DeltaMessage
674
    logprobs: Optional[ChatCompletionLogProbs] = None
675
    finish_reason: Optional[str] = None
676
    stop_reason: Optional[Union[int, str]] = None
677
678


679
class ChatCompletionStreamResponse(OpenAIBaseModel):
680
    id: str = Field(default_factory=lambda: f"chatcmpl-{random_uuid()}")
681
    object: Literal["chat.completion.chunk"] = "chat.completion.chunk"
682
683
684
    created: int = Field(default_factory=lambda: int(time.time()))
    model: str
    choices: List[ChatCompletionResponseStreamChoice]
685
    usage: Optional[UsageInfo] = Field(default=None)
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710


class BatchRequestInput(OpenAIBaseModel):
    """
    The per-line object of the batch input file.

    NOTE: Currently only the `/v1/chat/completions` endpoint is supported.
    """

    # A developer-provided per-request id that will be used to match outputs to
    # inputs. Must be unique for each request in a batch.
    custom_id: str

    # The HTTP method to be used for the request. Currently only POST is
    # supported.
    method: str

    # The OpenAI API relative URL to be used for the request. Currently
    # /v1/chat/completions is supported.
    url: str

    # The parameteters of the request.
    body: Union[ChatCompletionRequest, ]


711
712
713
714
715
716
717
718
class BatchResponseData(OpenAIBaseModel):
    # HTTP status code of the response.
    status_code: int = 200

    # An unique identifier for the API request.
    request_id: str

    # The body of the response.
719
    body: Optional[ChatCompletionResponse] = None
720
721


722
723
724
725
726
727
728
729
730
731
732
class BatchRequestOutput(OpenAIBaseModel):
    """
    The per-line object of the batch output and error files
    """

    id: str

    # A developer-provided per-request id that will be used to match outputs to
    # inputs.
    custom_id: str

733
    response: Optional[BatchResponseData]
734
735
736
737

    # For requests that failed with a non-HTTP error, this will contain more
    # information on the cause of the failure.
    error: Optional[Any]
738
739
740


class TokenizeRequest(OpenAIBaseModel):
741
742
743
744
    add_generation_prompt: bool = Field(default=True)
    add_special_tokens: bool = Field(default=False)
    prompt: Optional[str] = Field(default=None)
    messages: Optional[List[ChatCompletionMessageParam]] = Field(default=None)
745
746
747
748
749
750
    model: str


class TokenizeResponse(OpenAIBaseModel):
    count: int
    max_model_len: int
751
    tokens: List[int]
752
753
754
755
756
757
758
759
760


class DetokenizeRequest(OpenAIBaseModel):
    model: str
    tokens: List[int]


class DetokenizeResponse(OpenAIBaseModel):
    prompt: str