Skip to content

vllm.entrypoints.openai.chat_completion.protocol

ChatCompletionRequest

Bases: OpenAIBaseModel

Source code in vllm/entrypoints/openai/chat_completion/protocol.py
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
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
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
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
class ChatCompletionRequest(OpenAIBaseModel):
    # Ordered by official OpenAI API documentation
    # https://platform.openai.com/docs/api-reference/chat/create
    messages: list[ChatCompletionMessageParam]
    model: str | None = None
    frequency_penalty: float | None = 0.0
    logit_bias: dict[str, float] | None = None
    logprobs: bool | None = False
    top_logprobs: int | None = 0
    max_tokens: int | None = Field(
        default=None,
        deprecated="max_tokens is deprecated in favor of "
        "the max_completion_tokens field",
    )
    max_completion_tokens: int | None = None
    n: int | None = 1
    presence_penalty: float | None = 0.0
    response_format: AnyResponseFormat | None = None
    seed: int | None = Field(None, ge=_LONG_INFO.min, le=_LONG_INFO.max)
    stop: str | list[str] | None = []
    stream: bool | None = False
    stream_options: StreamOptions | None = None
    temperature: float | None = None
    top_p: float | None = None
    tools: list[ChatCompletionToolsParam] | None = None
    tool_choice: (
        Literal["none"]
        | Literal["auto"]
        | Literal["required"]
        | ChatCompletionNamedToolChoiceParam
        | None
    ) = "none"
    reasoning_effort: Literal["low", "medium", "high"] | None = None
    include_reasoning: bool = True
    parallel_tool_calls: bool | None = True

    # NOTE this will be ignored by vLLM
    user: str | None = None

    # --8<-- [start:chat-completion-sampling-params]
    use_beam_search: bool = False
    top_k: int | None = None
    min_p: float | None = None
    repetition_penalty: float | None = None
    length_penalty: float = 1.0
    stop_token_ids: list[int] | None = []
    include_stop_str_in_output: bool = False
    ignore_eos: bool = False
    min_tokens: int = 0
    skip_special_tokens: bool = True
    spaces_between_special_tokens: bool = True
    truncate_prompt_tokens: Annotated[int, Field(ge=-1, le=_LONG_INFO.max)] | None = (
        None
    )
    prompt_logprobs: int | None = None
    allowed_token_ids: list[int] | None = None
    bad_words: list[str] = Field(default_factory=list)
    # --8<-- [end:chat-completion-sampling-params]

    # --8<-- [start:chat-completion-extra-params]
    echo: 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: 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."
        ),
    )
    continue_final_message: bool = Field(
        default=False,
        description=(
            "If this is set, the chat will be formatted so that the final "
            "message in the chat is open-ended, without any EOS tokens. The "
            "model will continue this message rather than starting a new one. "
            'This allows you to "prefill" part of the model\'s response for it. '
            "Cannot be used at the same time as `add_generation_prompt`."
        ),
    )
    add_special_tokens: 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)."
        ),
    )
    documents: list[dict[str, str]] | None = 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: str | None = Field(
        default=None,
        description=(
            "A Jinja template to use for this conversion. "
            "As of transformers v4.44, default chat template is no longer "
            "allowed, so you must provide a chat template if the tokenizer "
            "does not define one."
        ),
    )
    chat_template_kwargs: dict[str, Any] | None = Field(
        default=None,
        description=(
            "Additional keyword args to pass to the template renderer. "
            "Will be accessible by the chat template."
        ),
    )
    mm_processor_kwargs: dict[str, Any] | None = Field(
        default=None,
        description=("Additional kwargs to pass to the HF processor."),
    )
    structured_outputs: StructuredOutputsParams | None = Field(
        default=None,
        description="Additional kwargs for structured outputs",
    )
    priority: int = Field(
        default=0,
        description=(
            "The priority of the request (lower means earlier handling; "
            "default: 0). Any priority other than 0 will raise an error "
            "if the served model does not use priority scheduling."
        ),
    )
    request_id: str = Field(
        default_factory=random_uuid,
        description=(
            "The request_id related to this request. If the caller does "
            "not set it, a random_uuid will be generated. This id is used "
            "through out the inference process and return in response."
        ),
    )

    return_tokens_as_token_ids: bool | None = Field(
        default=None,
        description=(
            "If specified with 'logprobs', tokens are represented "
            " as strings of the form 'token_id:{token_id}' so that tokens "
            "that are not JSON-encodable can be identified."
        ),
    )
    return_token_ids: bool | None = Field(
        default=None,
        description=(
            "If specified, the result will include token IDs alongside the "
            "generated text. In streaming mode, prompt_token_ids is included "
            "only in the first chunk, and token_ids contains the delta tokens "
            "for each chunk. This is useful for debugging or when you "
            "need to map generated text back to input tokens."
        ),
    )

    cache_salt: str | None = Field(
        default=None,
        description=(
            "If specified, the prefix cache will be salted with the provided "
            "string to prevent an attacker to guess prompts in multi-user "
            "environments. The salt should be random, protected from "
            "access by 3rd parties, and long enough to be "
            "unpredictable (e.g., 43 characters base64-encoded, corresponding "
            "to 256 bit)."
        ),
    )

    kv_transfer_params: dict[str, Any] | None = Field(
        default=None,
        description="KVTransfer parameters used for disaggregated serving.",
    )

    vllm_xargs: dict[str, str | int | float | list[str | int | float]] | None = Field(
        default=None,
        description=(
            "Additional request parameters with (list of) string or "
            "numeric values, used by custom extensions."
        ),
    )

    # --8<-- [end:chat-completion-extra-params]

    def build_chat_params(
        self,
        default_template: str | None,
        default_template_content_format: ChatTemplateContentFormatOption,
    ) -> ChatParams:
        return ChatParams(
            chat_template=self.chat_template or default_template,
            chat_template_content_format=default_template_content_format,
            chat_template_kwargs=merge_kwargs(
                self.chat_template_kwargs,
                dict(
                    add_generation_prompt=self.add_generation_prompt,
                    continue_final_message=self.continue_final_message,
                    documents=self.documents,
                    reasoning_effort=self.reasoning_effort,
                ),
            ),
        )

    def build_tok_params(self, model_config: ModelConfig) -> TokenizeParams:
        if self.max_completion_tokens is not None:
            max_output_tokens: int | None = self.max_completion_tokens
            max_output_tokens_param = "max_completion_tokens"
        else:
            max_output_tokens = self.max_tokens
            max_output_tokens_param = "max_tokens"

        return TokenizeParams(
            max_total_tokens=model_config.max_model_len,
            max_output_tokens=max_output_tokens or 0,
            truncate_prompt_tokens=self.truncate_prompt_tokens,
            add_special_tokens=self.add_special_tokens,
            needs_detokenization=bool(self.echo and not self.return_token_ids),
            max_total_tokens_param="max_model_len",
            max_output_tokens_param=max_output_tokens_param,
        )

    # Default sampling parameters for chat completion requests
    _DEFAULT_SAMPLING_PARAMS: dict = {
        "repetition_penalty": 1.0,
        "temperature": 1.0,
        "top_p": 1.0,
        "top_k": 0,
        "min_p": 0.0,
    }

    def to_beam_search_params(
        self, max_tokens: int, default_sampling_params: dict
    ) -> BeamSearchParams:
        n = self.n if self.n is not None else 1
        if (temperature := self.temperature) is None:
            temperature = default_sampling_params.get(
                "temperature", self._DEFAULT_SAMPLING_PARAMS["temperature"]
            )

        return BeamSearchParams(
            beam_width=n,
            max_tokens=max_tokens,
            ignore_eos=self.ignore_eos,
            temperature=temperature,
            length_penalty=self.length_penalty,
            include_stop_str_in_output=self.include_stop_str_in_output,
        )

    def to_sampling_params(
        self,
        max_tokens: int,
        default_sampling_params: dict,
    ) -> SamplingParams:
        # Default parameters
        if (repetition_penalty := self.repetition_penalty) is None:
            repetition_penalty = default_sampling_params.get(
                "repetition_penalty",
                self._DEFAULT_SAMPLING_PARAMS["repetition_penalty"],
            )
        if (temperature := self.temperature) is None:
            temperature = default_sampling_params.get(
                "temperature", self._DEFAULT_SAMPLING_PARAMS["temperature"]
            )
        if (top_p := self.top_p) is None:
            top_p = default_sampling_params.get(
                "top_p", self._DEFAULT_SAMPLING_PARAMS["top_p"]
            )
        if (top_k := self.top_k) is None:
            top_k = default_sampling_params.get(
                "top_k", self._DEFAULT_SAMPLING_PARAMS["top_k"]
            )
        if (min_p := self.min_p) is None:
            min_p = default_sampling_params.get(
                "min_p", self._DEFAULT_SAMPLING_PARAMS["min_p"]
            )

        prompt_logprobs = self.prompt_logprobs
        if prompt_logprobs is None and self.echo:
            prompt_logprobs = self.top_logprobs

        response_format = self.response_format
        if response_format is not None:
            structured_outputs_kwargs = dict[str, Any]()

            # Set structured output params for response format
            if response_format.type == "json_object":
                structured_outputs_kwargs["json_object"] = True
            elif response_format.type == "json_schema":
                json_schema = response_format.json_schema
                assert json_schema is not None
                structured_outputs_kwargs["json"] = json_schema.json_schema
            elif response_format.type == "structural_tag":
                structural_tag = response_format
                assert structural_tag is not None and isinstance(
                    structural_tag,
                    (
                        LegacyStructuralTagResponseFormat,
                        StructuralTagResponseFormat,
                    ),
                )
                s_tag_obj = structural_tag.model_dump(by_alias=True)
                structured_outputs_kwargs["structural_tag"] = json.dumps(s_tag_obj)

            # If structured outputs wasn't already enabled,
            # we must enable it for these features to work
            if len(structured_outputs_kwargs) > 0:
                self.structured_outputs = (
                    StructuredOutputsParams(**structured_outputs_kwargs)
                    if self.structured_outputs is None
                    else replace(self.structured_outputs, **structured_outputs_kwargs)
                )

        extra_args: dict[str, Any] = self.vllm_xargs if self.vllm_xargs else {}
        if self.kv_transfer_params:
            # Pass in kv_transfer_params via extra_args
            extra_args["kv_transfer_params"] = self.kv_transfer_params
        return SamplingParams.from_optional(
            n=self.n,
            presence_penalty=self.presence_penalty,
            frequency_penalty=self.frequency_penalty,
            repetition_penalty=repetition_penalty,
            temperature=temperature,
            top_p=top_p,
            top_k=top_k,
            min_p=min_p,
            seed=self.seed,
            stop=self.stop,
            stop_token_ids=self.stop_token_ids,
            logprobs=self.top_logprobs if self.logprobs else None,
            prompt_logprobs=prompt_logprobs,
            ignore_eos=self.ignore_eos,
            max_tokens=max_tokens,
            min_tokens=self.min_tokens,
            skip_special_tokens=self.skip_special_tokens,
            spaces_between_special_tokens=self.spaces_between_special_tokens,
            include_stop_str_in_output=self.include_stop_str_in_output,
            truncate_prompt_tokens=self.truncate_prompt_tokens,
            output_kind=RequestOutputKind.DELTA
            if self.stream
            else RequestOutputKind.FINAL_ONLY,
            structured_outputs=self.structured_outputs,
            logit_bias=self.logit_bias,
            bad_words=self.bad_words,
            allowed_token_ids=self.allowed_token_ids,
            extra_args=extra_args or None,
            skip_clone=True,  # Created fresh per request, safe to skip clone
        )

    @model_validator(mode="before")
    @classmethod
    def validate_stream_options(cls, data):
        if data.get("stream_options") and not data.get("stream"):
            raise VLLMValidationError(
                "Stream options can only be defined when `stream=True`.",
                parameter="stream_options",
            )

        return data

    @model_validator(mode="before")
    @classmethod
    def check_logprobs(cls, data):
        if (prompt_logprobs := data.get("prompt_logprobs")) is not None:
            if data.get("stream") and (prompt_logprobs > 0 or prompt_logprobs == -1):
                raise VLLMValidationError(
                    "`prompt_logprobs` are not available when `stream=True`.",
                    parameter="prompt_logprobs",
                )

            if prompt_logprobs < 0 and prompt_logprobs != -1:
                raise VLLMValidationError(
                    "`prompt_logprobs` must be a positive value or -1.",
                    parameter="prompt_logprobs",
                    value=prompt_logprobs,
                )
        if (top_logprobs := data.get("top_logprobs")) is not None:
            if top_logprobs < 0 and top_logprobs != -1:
                raise VLLMValidationError(
                    "`top_logprobs` must be a positive value or -1.",
                    parameter="top_logprobs",
                    value=top_logprobs,
                )

            if (top_logprobs == -1 or top_logprobs > 0) and not data.get("logprobs"):
                raise VLLMValidationError(
                    "when using `top_logprobs`, `logprobs` must be set to true.",
                    parameter="top_logprobs",
                )

        return data

    @model_validator(mode="before")
    @classmethod
    def check_structured_outputs_count(cls, data):
        if isinstance(data, ValueError):
            raise data

        if data.get("structured_outputs", None) is None:
            return data

        structured_outputs_kwargs = data["structured_outputs"]
        count = sum(
            structured_outputs_kwargs.get(k) is not None
            for k in ("json", "regex", "choice")
        )
        # you can only use one kind of constraints for structured outputs
        if count > 1:
            raise ValueError(
                "You can only use one kind of constraints for structured "
                "outputs ('json', 'regex' or 'choice')."
            )
        # you can only either use structured outputs or tools, not both
        if count > 1 and data.get("tool_choice", "none") not in (
            "none",
            "auto",
            "required",
        ):
            raise ValueError(
                "You can only either use constraints for structured outputs "
                "or tools, not both."
            )
        return data

    @model_validator(mode="before")
    @classmethod
    def check_tool_usage(cls, data):
        # if "tool_choice" is not specified but tools are provided,
        # default to "auto" tool_choice
        if "tool_choice" not in data and data.get("tools"):
            data["tool_choice"] = "auto"

        # if "tool_choice" is "none" -- no validation is needed for tools
        if "tool_choice" in data and data["tool_choice"] == "none":
            return data

        # if "tool_choice" is specified -- validation
        if "tool_choice" in data and data["tool_choice"] is not None:
            # ensure that if "tool choice" is specified, tools are present
            if "tools" not in data or data["tools"] is None:
                raise ValueError("When using `tool_choice`, `tools` must be set.")

            # make sure that tool choice is either a named tool
            # OR that it's set to "auto" or "required"
            if data["tool_choice"] not in ["auto", "required"] and not isinstance(
                data["tool_choice"], dict
            ):
                raise ValueError(
                    f"Invalid value for `tool_choice`: {data['tool_choice']}! "
                    'Only named tools, "none", "auto" or "required" '
                    "are supported."
                )

            # if tool_choice is "required" but the "tools" list is empty,
            # override the data to behave like "none" to align with
            # OpenAI’s behavior.
            if (
                data["tool_choice"] == "required"
                and isinstance(data["tools"], list)
                and len(data["tools"]) == 0
            ):
                data["tool_choice"] = "none"
                del data["tools"]
                return data

            # ensure that if "tool_choice" is specified as an object,
            # it matches a valid tool
            correct_usage_message = (
                'Correct usage: `{"type": "function",'
                ' "function": {"name": "my_function"}}`'
            )
            if isinstance(data["tool_choice"], dict):
                valid_tool = False
                function = data["tool_choice"].get("function")
                if not isinstance(function, dict):
                    raise ValueError(
                        f"Invalid value for `function`: `{function}` in "
                        f"`tool_choice`! {correct_usage_message}"
                    )
                if "name" not in function:
                    raise ValueError(
                        f"Expected field `name` in `function` in "
                        f"`tool_choice`! {correct_usage_message}"
                    )
                function_name = function["name"]
                if not isinstance(function_name, str) or len(function_name) == 0:
                    raise ValueError(
                        f"Invalid `name` in `function`: `{function_name}`"
                        f" in `tool_choice`! {correct_usage_message}"
                    )
                for tool in data["tools"]:
                    if tool["function"]["name"] == function_name:
                        valid_tool = True
                        break
                if not valid_tool:
                    raise ValueError(
                        "The tool specified in `tool_choice` does not match any"
                        " of the specified `tools`"
                    )
        return data

    @model_validator(mode="before")
    @classmethod
    def check_generation_prompt(cls, data):
        if data.get("continue_final_message") and data.get("add_generation_prompt"):
            raise ValueError(
                "Cannot set both `continue_final_message` and "
                "`add_generation_prompt` to True."
            )
        return data

    @model_validator(mode="before")
    @classmethod
    def check_cache_salt_support(cls, data):
        if data.get("cache_salt") is not None and (
            not isinstance(data["cache_salt"], str) or not data["cache_salt"]
        ):
            raise ValueError(
                "Parameter 'cache_salt' must be a non-empty string if provided."
            )
        return data

    @model_validator(mode="before")
    @classmethod
    def check_system_message_content_type(cls, data):
        """Warn if system messages contain non-text content.

        According to OpenAI API spec, system messages can only be of type
        'text'. We log a warning instead of rejecting to avoid breaking
        users who intentionally send multimodal system messages.
        See: https://platform.openai.com/docs/api-reference/chat/create#chat_create-messages-system_message
        """
        if not isinstance(data, dict):
            return data
        messages = data.get("messages", [])
        for msg in messages:
            # Check if this is a system message
            if isinstance(msg, dict) and msg.get("role") == "system":
                content = msg.get("content")

                # If content is a list (multimodal format)
                if isinstance(content, list):
                    for part in content:
                        if isinstance(part, dict):
                            part_type = part.get("type")
                            # Infer type when 'type' field is not explicit
                            if part_type is None:
                                if "image_url" in part or "image_pil" in part:
                                    part_type = "image_url"
                                elif "image_embeds" in part:
                                    part_type = "image_embeds"
                                elif "audio_url" in part:
                                    part_type = "audio_url"
                                elif "input_audio" in part:
                                    part_type = "input_audio"
                                elif "audio_embeds" in part:
                                    part_type = "audio_embeds"
                                elif "video_url" in part:
                                    part_type = "video_url"

                            # Warn about non-text content in system messages
                            if part_type and part_type != "text":
                                logger.warning_once(
                                    "System messages should only contain text "
                                    "content according to the OpenAI API spec. "
                                    "Found content type: '%s'.",
                                    part_type,
                                )

        return data

check_system_message_content_type classmethod

check_system_message_content_type(data)

Warn if system messages contain non-text content.

According to OpenAI API spec, system messages can only be of type 'text'. We log a warning instead of rejecting to avoid breaking users who intentionally send multimodal system messages. See: https://platform.openai.com/docs/api-reference/chat/create#chat_create-messages-system_message

Source code in vllm/entrypoints/openai/chat_completion/protocol.py
@model_validator(mode="before")
@classmethod
def check_system_message_content_type(cls, data):
    """Warn if system messages contain non-text content.

    According to OpenAI API spec, system messages can only be of type
    'text'. We log a warning instead of rejecting to avoid breaking
    users who intentionally send multimodal system messages.
    See: https://platform.openai.com/docs/api-reference/chat/create#chat_create-messages-system_message
    """
    if not isinstance(data, dict):
        return data
    messages = data.get("messages", [])
    for msg in messages:
        # Check if this is a system message
        if isinstance(msg, dict) and msg.get("role") == "system":
            content = msg.get("content")

            # If content is a list (multimodal format)
            if isinstance(content, list):
                for part in content:
                    if isinstance(part, dict):
                        part_type = part.get("type")
                        # Infer type when 'type' field is not explicit
                        if part_type is None:
                            if "image_url" in part or "image_pil" in part:
                                part_type = "image_url"
                            elif "image_embeds" in part:
                                part_type = "image_embeds"
                            elif "audio_url" in part:
                                part_type = "audio_url"
                            elif "input_audio" in part:
                                part_type = "input_audio"
                            elif "audio_embeds" in part:
                                part_type = "audio_embeds"
                            elif "video_url" in part:
                                part_type = "video_url"

                        # Warn about non-text content in system messages
                        if part_type and part_type != "text":
                            logger.warning_once(
                                "System messages should only contain text "
                                "content according to the OpenAI API spec. "
                                "Found content type: '%s'.",
                                part_type,
                            )

    return data