.`i)dddlZddlZddlmZddlmZmZmZmZddl Z ddl m Z ddl mZddlmZmZddlmZddlmZmZmZmZmZmZmZmZmZmZm Z m!Z!dd l"m#Z#dd l$m%Z%dd l&m'Z'dd l(m)Z)m*Z*m+Z+m,Z,dd l-m.Z.e%e/Z0e j1e j2Z3GddeZ4GddeZ5Gdde5Z6GddeZ7GddeZ8GddeZ9GddeZ:GddeZ;GddeZ<Gd d!eZ=Gd"d#eZ>Gd$d%eZ?dS)&N)replace) AnnotatedAnyClassVarLiteral)ChatCompletionAudio) Annotation)Fieldmodel_validator)ChatCompletionMessageParam) AnyResponseFormat DeltaMessage FunctionCallFunctionDefinition!LegacyStructuralTagResponseFormatLogitsProcessorsOpenAIBaseModel StreamOptionsStructuralTagResponseFormatToolCall UsageInfoget_logits_processors)VLLMValidationError) init_logger)Logprob)BeamSearchParamsRequestOutputKindSamplingParamsStructuredOutputsParams random_uuidceZdZUeed<dZedzed<dZedzed<dZedzed<dZ e dzed<dZ e dzed<e eZeeed <dZedzed <dZedzed < ed dZdS) ChatMessageroleNcontentrefusal annotationsaudio function_calldefault_factory tool_calls reasoningreasoning_contentaftermodec|j|_|S)z?Copy reasoning to reasoning_content for backward compatibility.)r-r.)selfs /home/jaya/work/projects/VOICE-AGENT/VIET/agent-env/lib/python3.11/site-packages/vllm/entrypoints/openai/chat_completion/protocol.py#handle_deprecated_reasoning_contentz/ChatMessage.handle_deprecated_reasoning_contentCs"& )__name__ __module__ __qualname__str__annotations__r%r&r'OpenAIAnnotationr(OpenAIChatCompletionAudior)rr listr,rr-r.r r5r6r4r#r#5s IIIGS4ZGS4Z+/K!D(///.2E $t +222)-M<$&---!&t!intr?r6r4rArAJsD JJJGU"E49t """""r6rAcjeZdZUdZeeedzed<ee Z e e ed<dS)ChatCompletionLogProbsContentN field_namesr* top_logprobs) r7r8r9rIrsetr:r;r r>rJrAr?r6r4rHrHPsT.2K#c(T/*11105d0K0K0KL$,-KKKKKr6rHc0eZdZUdZeedzed<dS)ChatCompletionLogProbsNr%)r7r8r9r%r>rHr;r?r6r4rMrMWs-:>GT/ 04 7>>>>>r6rMceZdZUeed<eed<dZedzed<dZe dzed<dZ ee zdzed<dZ e edzed<dS) ChatCompletionResponseChoiceindexmessageNlogprobsstop finish_reason stop_reason token_ids) r7r8r9rFr;r#rRrMrTr:rUrVr>r?r6r4rOrO[s JJJ .2H$t+222 &M3:&&&$(KsT!(((#'ItCy4&&&&&r6rOc`eZdZUedZeed<dZeded<edZ e ed<eed<e e ed <d Z ed d zed <d Zed zed <eed<d Ze ee efd zd zed<d Ze e d zed<ed dZeeefd zed<d S)ChatCompletionResponsec$dtSNz chatcmpl-r r?r6r4zChatCompletionResponse.i,G ,G,Gr6r*idzchat.completionobjectcBttjSNrFtimer?r6r4r[zChatCompletionResponse.kTY[[1A1Ar6createdmodelchoicesN)autodefaultflexscalepriority service_tiersystem_fingerprintusageprompt_logprobsprompt_token_idszKVTransfer parameters.rh descriptionkv_transfer_params)r7r8r9r r]r:r;r^rrdrFr>rOrlrmrrodictrrprsrr?r6r4rXrXhsEe$G$GHHHBHHH):FG% &:::5)A)ABBBGSBBB JJJ . ////SWL'HIDPWWW%)d ))) ?COT$sG|,t34t;BBB)-d3i$&---05":111S#X-r6rXceZdZUeed<eed<dZedzed<dZe dzed<dZ ee zdzed<dZ e edzed<dS)"ChatCompletionResponseStreamChoicerPdeltaNrRrTrUrV) r7r8r9rFr;rrRrMrTr:rUrVr>r?r6r4rvrvzs JJJ .2H$t+222 $M3:$$$$(KsT!((("&ItCy4&&&&&r6rvceZdZUedZeed<dZeded<edZ e ed<eed<e e ed <ed Z ed zed <d Ze e d zed <d S)ChatCompletionStreamResponsec$dtSrZr r?r6r4r[z%ChatCompletionStreamResponse.r\r6r*r]zchat.completion.chunkr^cBttjSr`rar?r6r4r[z%ChatCompletionStreamResponse.rcr6rdrerfN)rhrnrp)r7r8r9r r]r:r;r^rrdrFr>rvrnrrpr?r6r4ryryse$G$GHHHBHHH/FFG+ ,FFF5)A)ABBBGSBBB JJJ 4 5555#eD111E9t 111)-d3i$&-----r6ryc4eZdZUdZeded<eed<dS)ChatCompletionToolsParamfunctiontypeN)r7r8r9rrr;rr?r6r4r}r}s3 *D'* ***      r6r}ceZdZUeed<dS)ChatCompletionNamedFunctionnameN)r7r8r9r:r;r?r6r4rrs IIIIIr6rc4eZdZUeed<dZeded<dS)"ChatCompletionNamedToolChoiceParamr~rN)r7r8r9rr;rrr?r6r4rrs3)))) *D'* *****r6rceZdZUeeed<dZedzed<dZe dzed<dZ e ee fdzed<dZ e dzed<d Zedzed <edd Zedzed <dZedzed<dZedzed<dZe dzed<dZedzed<edejejZedzed<gZeeezdzed<dZe dzed<dZedzed<dZe dzed<dZ e dzed<dZ!ee"dzed<dZ#e$de$dze$dze%zdzed<dZ&e$ddzed <d!Z'e ed"<d!Z(e dzed#<dZ)edzed$<dZ*e ed%<dZ+edzed&<dZ,e dzed'<dZ-e dzed(<d)Z.e ed*<gZ/eedzed+<dZ0e ed,<dZ1e ed-<d Z2eed.<d!Z3e ed/<d!Z4e ed0<dZ5e6eed1ejfdzed2<dZ7edzed3<dZ8eedzed4<ee5Z9eeed6<edd78Z:e ed9<ed!d:8Z;e ed;<edd<8Z8Z=e ed?<edd@8Z>ee eefdzedA<eddB8Z?edzedC<eddD8Z@e eeAfdzedE<eddF8ZBe eeAfdzedG<eddH8ZCeDdzedI<ed dJ8ZEeedK<eeFdLMZGeedN<eddO8ZHeIdzedP<eddQ8ZJe dzedR<eddS8ZKe dzedT<eddU8ZLedzedV<eddW8ZMe eeAfdzedX<eddY8ZNe eeeze zeeeze zzfdzedZ<d)d)d)d dd[ZOe ed\<d ed]e d^ePfd_ZQd ed`edzd]e d^eRfdaZSeTdbceUddZVeTdbceUdeZWeTdbceUdfZXeTdbceUdgZYeTdbceUdhZZeTdbceUdiZ[dS)jChatCompletionRequestmessagesNregfrequency_penalty logit_biasFrRrrJzDmax_tokens is deprecated in favor of the max_completion_tokens field)rh deprecated max_tokensmax_completion_tokensnpresence_penaltyresponse_format)geleseedrSstreamstream_options temperaturetop_ptoolsnonergrequired tool_choice)lowmediumhighreasoning_effortTinclude_reasoningparallel_tool_callsuseruse_beam_searchtop_kmin_prepetition_penaltyg?length_penaltystop_token_idsinclude_stop_str_in_output ignore_eos min_tokensskip_special_tokensspaces_between_special_tokenstruncate_prompt_tokensroallowed_token_idsr* bad_wordszaIf true, the new message will be prepended with the last message if they belong to the same role.rqechozIf 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.add_generation_prompta: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`.continue_final_messagezIf 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).add_special_tokensa#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. documentszA 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_templatezbAdditional keyword args to pass to the template renderer. Will be accessible by the chat template.chat_template_kwargsz.Additional kwargs to pass to the HF processor.mm_processor_kwargsz(Additional kwargs for structured outputsstructured_outputszThe 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.rkzThe 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.)r+rr request_idaA list of either qualified names of logits processors, or constructor objects, to apply when sampling. A constructor is a JSON object with a required 'qualname' field specifying the qualified name of the processor class/factory, and optional 'args' and 'kwargs' fields containing positional and keyword arguments. For example: {'qualname': 'my_module.MyLogitsProcessor', 'args': [1, 2], 'kwargs': {'param': 'value'}}.logits_processorszIf 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_tokens_as_token_idsa(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.return_token_idsa/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). cache_saltz5KVTransfer parameters used for disaggregated serving.rszaAdditional request parameters with (list of) string or numeric values, used by custom extensions. vllm_xargs)rrrrr_DEFAULT_SAMPLING_PARAMSdefault_sampling_paramsreturnc|j|jnd}|jx}!|d|jd}t |||j||j|jS)Nrr) beam_widthrrrrr)rrgetrrrrr)r3rrrrs r4to_beam_search_paramsz+ChatCompletionRequest.to_beam_search_paramspsyf(DFFa+ +K 4155t<]KK !#.'+'F     r6logits_processor_patternc |jx}!|d|jd}|jx}!|d|jd}|jx}!|d|jd}|jx}!|d|jd}|jx}!|d|jd}|j} | |jr|j } |j } | tttf} | jdkrd| d<n{| jdkr| j} | J| j| d <nZ| jd krO| } | t!| t"t$fsJ| d }t)j|| d <t-| d kr*|j t1d%i| nt3|jfi| |_|jr|jni}|jr |j|d <t9jd%id|jd|jd|j d|d|d|d|d|d|j!d|j"d|j#d|j$r|j ndd| d|j%d|d|j&d|j'd|j(dtS|j*|d|j+d|j,d|j-r t\j/n t\j0d|jd |j1d!|j2d"|j3d#|pdd$dS)&Nrrrrr json_objectT json_schemajsonstructural_tag)by_aliasrrsrrrrrSrrRrorrrrrrrr output_kindrrrr extra_args skip_cloner?)4rrrrrrrrorrJrrtr:rrr isinstancerr model_dumprdumpslenrrrrrsr from_optionalrrrrrSrrRrrrrrrrrrrDELTA FINAL_ONLYrrr)r3rrrrrrrrrorstructured_outputs_kwargsrr s_tag_objrs r4to_sampling_paramsz(ChatCompletionRequest.to_sampling_paramssY#'"9 9  B!8!Bj>S>S)*:;,--11.6,HH.GHHH !8VV#>/! 2{.)//"-5! 6 $667! 89! :nn;! <#44=! >")T?! @tA! ! r6beforer0c||dr&|dstdd|S)Nrrz6Stream options can only be defined when `stream=True`. parameterrrclsdatas r4validate_stream_optionsz-ChatCompletionRequest.validate_stream_optionssP 88$ % % dhhx.@.@ %H*   r6c|dx}P|dr|dks|dkrtdd|dkr|dkrtdd||d x}P|dkr|dkrtd d ||dks|dkr&|d std d |S) Nrorrrz7`prompt_logprobs` are not available when `stream=True`.rz1`prompt_logprobs` must be a positive value or -1.)rvaluerJz.`top_logprobs` must be a positive value or -1.rRz:when using `top_logprobs`, `logprobs` must be set to true.r)rrrorJs r4check_logprobsz$ChatCompletionRequest.check_logprobss3 $xx(9:: :O Gxx!! ':':oQS>S>S)M/ """'<'<)G/) !HH^44 4L AaLB$6$6)D,& ""lQ&6&6@T@T&6)P,  r6c0t|tr||dd|S|dtfddD}|dkrtd|dkr'|dddvrtd |S) Nrc3FK|]}|duVdSr`)r).0krs r4 zGChatCompletionRequest.check_structured_outputs_count..sJ   & ) )! , ,D 8      r6)rregexchoicerz^You can only use one kind of constraints for structured outputs ('json', 'regex' or 'choice').rr)rrgrzNYou can only either use constraints for structured outputs or tools, not both.)r ValueErrorrsum)rrcountrs @r4check_structured_outputs_countz4ChatCompletionRequest.check_structured_outputs_counts dJ ' ' J 88($ / / 7K$()=$>!    0      1999  199-88A   &  r6cd|vr|drd|d<d|vr|ddkr|Sd|vr|dd|vs|dtd|ddvr4t|dtstd|dd|dd kr>t|dtr#t |dd kr d|d<|d=|Sd }t|dtrd }|dd }t|tstd|d|d|vrtd||d}t|t rt |d krtd|d||dD]}|d d|krd}n|std|S)Nrrrgrz.When using `tool_choice`, `tools` must be set.)rgrz!Invalid value for `tool_choice`: z?! Only named tools, "none", "auto" or "required" are supported.rrzJCorrect usage: `{"type": "function", "function": {"name": "my_function"}}`Fr~zInvalid value for `function`: `z` in `tool_choice`! rz6Expected field `name` in `function` in `tool_choice`! zInvalid `name` in `function`: `TzOThe tool specified in `tool_choice` does not match any of the specified `tools`)rrrrtr>rr:)rrcorrect_usage_message valid_toolr~ function_nametools r4check_tool_usagez&ChatCompletionRequest.check_tool_usage5s  $ $'):): $"(D  D T-%8F%B%BK D T-%8%Dd""d7m&; !QRRRM"*>>>z]#THH>!%]8K%%%]#z11tG}d332W &&!++&,]#M 9 "$}-t44 "  .22:>>!(D11$B(BB*?BB))$B*?BB!) 0 !-55]9K9Kq9P9P$F-FF.CFF!MDJ'/=@@%) A"$4 r6cx|dr$|drtd|S)NrrzMCannot set both `continue_final_message` and `add_generation_prompt` to True.)rrrs r4check_generation_promptz-ChatCompletionRequest.check_generation_promptsK 88, - - $((;R2S2S 3  r6c|d2t|dtr|dstd|S)Nrz>Parameter 'cache_salt' must be a non-empty string if provided.)rrr:rrs r4check_cache_salt_supportz.ChatCompletionRequest.check_cache_salt_supportsX 88L ! ! -4 -s33 .;? ;M .P  r6)\r7r8r9r>r r;rer:rrErrtrRboolrJrFr rrrrrr _LONG_INFOminmaxrrSrrrrrrr}rrrrrrrrrrrrrrrrrrrrrorrrrrrrrrrrrrrkr!rrrrrrrsrrrrrrr classmethodrrrrrrr?r6r4rrsx -....E3:&)ut|)))*.JS%Z 4'...!HdTk!!! L#*   "U*Jd  )-3:,,,AsTz%(edl(((04O&-444uTjnHHHD#*HHH#%D#S /D %%%FD4K+/NMD(/// $K$$$E54<37E4( )D 0777   &/  *   - -   AEg56=DDD"t"""'++++D#*"OT!!!E3:E54<'+ +++NE')NDI$)))',,,,JJ $$$$*.!4... Ic55B:>+J+J+J&JKdR#'OS4Z&&&*.tCy4'... 5666ItCy666 /D$#(% ###4$)5 J $ $ $D    %u       .3U ' . . .ItDcN#d*   !& #!!!M3:38% 7333$sCx.4/27E222c3h$.:?>:::/$6E DHce# HJ27 # 2 2 2'$.   /4e =///t %*E ? % % %dTk   #U    Jd   16K111S#X- QVPU 9QQQJS#)e+d39u3D.EEEFM" &&d  8<     $f f #&*f "& f  f f f f P_(###[$#_(###[$#<_(###[$#<_(###II[$#IV_(###[$#_(###[$#r6r)@rrb dataclassesrtypingrrrrtorch'openai.types.chat.chat_completion_audiorr=)openai.types.chat.chat_completion_messager r<pydanticr r vllm.entrypoints.chat_utilsr 'vllm.entrypoints.openai.engine.protocolr rrrrrrrrrrrvllm.exceptionsr vllm.loggerr vllm.logprobsrvllm.sampling_paramsrrrr vllm.utilsr!r7loggeriinfolongrr#rArHrMrOrXrvryr}rrrr?r6r4rs  444444444444 UTTTTT CBBBBB                            0/////######!!!!!! #""""" X  U[ $ $ /*#####O### LLLLL$9LLL?????_??? ' ' ' ' '? ' ' '_$''''''''.....?...!!!!!!!! /++++++++ xxxxxOxxxxxr6