Vercel.AI.Operations

Speech, image, video, embeddings, transcription, and other operations.

Classes

Type	Summary
AiSdkVersion	Package version used in the ai/ user-agent suffix.
AudioChunkStream	Audio the model may take ownership of.
Batch	Batch operations. Maps to startBatch, getBatchStatus, getBatchResults, listBatches, and cancelBatch.
BatchCancelResult	Cancellation result.
BatchContentPart	One content part inside a batch text result.
BatchImageGeneration	A provider image generation inside a batch item.
BatchItem	One raw batch item.
BatchItemResult	Normalized batch item.
BatchListResult	A page of batches.
BatchModels	Default provider used when a batch call omits one.
BatchModelUsage	Provider token usage before it is reshaped.
BatchOperationCall	Arguments for a batch operation.
BatchReference	A version-2 batch reference.
BatchRequest	One request in a batch.
BatchRequestCounts	Request counts.
BatchRequestOptions	Options for batch operations.
BatchRequestWarning	A warning tied to a request.
BatchStartResult	Start acknowledgement.
BatchStatus	Status fields shared by start, list, and status calls.
BatchTextGeneration	A provider text generation inside a batch item.
BatchTool	A tool definition forwarded to a batch without being executed.
BatchUsage	Usage returned to batch callers.
Decide	Decides typed questions. Maps to experimentaldecide.
DecideEvent	Lifecycle event for decide.
DecideRequest	Options for CancellationToken).
DecideTelemetry	Telemetry callbacks. Generic onStart and onEnd are not invoked.
DecisionAnswer	One model answer. Numbers are returned unchanged.
DecisionBooleanCriteria	Optional descriptions of the true and false outcomes of a boolean question. A side that is never set is left out of the wire request. A side set to null is sent as null.
DecisionBox	Edges in [0, 1], measured from the image’s top-left corner.
DecisionCaller	Helpers for IDecisionCaller.
DecisionChoiceCriteria	Criteria of a choice question: a nonempty map of option names to descriptions. A null description means none.
DecisionCriteria	Criteria of a decision question. Maps to the criteria member of DecisionModelV4Question.
DecisionFilePart	A file part of decision state. A URL is downloaded before the model call.
DecisionInput	Content for decision instructions and criteria descriptions: a string, a JSON object, or a JSON array. Maps to DecisionModelV4Input.
DecisionJsonPart	A JSON part of decision state.
DecisionLanguageModel	Adapts structured language-model output to Choice, Score, and Boolean decisions. Maps to DecisionLanguageModel.
DecisionLanguageModelOptions	The wrapped model and provider id that cross a workflow boundary for DecisionLanguageModel.
DecisionModelCall	Arguments for doDecide.
DecisionModelEvent	Model-call telemetry event.
DecisionModelFilePart	A file part the model receives. Maps to LanguageModelV4FilePart.
DecisionModelJsonPart	A JSON part the model receives.
DecisionModelResult	Model response for doDecide.
DecisionModels	Resolves decision models, including string model ids. Maps to resolveDecisionModel.
DecisionModelStatePart	A part of decision state as the model receives it. Maps to DecisionModelV4StatePart.
DecisionModelTextPart	A text part the model receives.
DecisionQuestion	One decision question. Maps to DecisionModelV4Question.
DecisionResult	Decision result.
DecisionRounding	Declared rounding precision.
DecisionScoreCriteria	Criteria of a score question: at least two ordered levels, indexed from zero. A null description means none.
DecisionState	Shared state of a decision: text, a JSON object, or ordered parts. Maps to DecisionState. A JSON array is not valid state. Wrap it in an object or in a DecisionJsonPart. CancellationToken) maps to prepareDecisionState.
DecisionStatePart	A part of decision state as the caller passes it. Maps to DecisionStatePart.
DecisionTextPart	A text part of decision state.
DecisionUnsupportedQuestionTypeException	The decision model does not support a question type. It carries the ErrorMarker marker, so IsInstance(object) recognizes it.
DownloadedMedia	Downloaded media.
DownloadException	Downloading generated media failed.
Embed	Embeds one value. Maps to embed.
EmbeddingModelCall	Arguments for one doEmbed call.
EmbeddingModelCallEvent	One inner doEmbed telemetry event.
EmbeddingModelResponse	One embedding returned by a model call.
EmbedEndEvent	End event for embed and embedMany.
EmbedMany	Embeds many values, splitting on count and UTF-8 byte limits. Maps to embedMany.
EmbedManyRequest	Options for CancellationToken).
EmbedManyResult	Result of CancellationToken).
EmbedRequest	Options for CancellationToken).
EmbedResult	Result of CancellationToken).
EmbedStartEvent	Start event for embed and embedMany.
EmbedTelemetry	Telemetry callbacks for an embedding operation.
Evaluate	Deprecated alias of Decide.
EvaluateEvent	Deprecated alias of DecideEvent.
EvaluateModelEvent	Deprecated alias of DecisionModelEvent.
EvaluateRequest	Deprecated alias of DecideRequest.
EvaluateResult	Deprecated alias of DecisionResult.
EvaluateTelemetry	Deprecated alias of DecideTelemetry.
EvaluationAnswer	Deprecated alias of DecisionAnswer.
EvaluationModelCall	Deprecated alias of DecisionModelCall.
EvaluationModelResult	Deprecated alias of DecisionModelResult.
EvaluationQuestion	Deprecated alias of DecisionQuestion.
EvaluationRounding	Deprecated alias of DecisionRounding.
EvaluationUnsupportedQuestionTypeException	Deprecated alias of DecisionUnsupportedQuestionTypeException.
GeneratedAudio	Generated speech audio.
GeneratedImage	A generated image file.
GeneratedVideo	A generated video file.
GenerateImage	Generates images. Maps to generateImage.
GenerateImageRequest	Options for CancellationToken).
GenerateImageResult	Result of CancellationToken).
GenerateSpeech	Generates speech. Maps to generateSpeech.
GenerateSpeechRequest	Options for CancellationToken).
GenerateSpeechResult	Result of CancellationToken).
GenerateVideo	Generates videos. Maps to experimentalgenerateVideo.
GenerateVideoRequest	Options shared by generate and start.
GenerateVideoResult	Result of CancellationToken).
GetVideoStatus	Reads video operation status. Maps to experimentalgetVideoStatus.
GetVideoStatusRequest	Options for CancellationToken).
GoogleMultimodalEmbedding	Gemini embedding caller used by EmbedMany. Splits multimodal google.content to the chunk that is about to be sent and posts embedContent or batchEmbedContents.
ImageCallResult	One provider call preserved on the result.
ImageModelCall	Arguments for one image model call.
ImageModelFile	An image sent to doGenerate as a file or URL.
ImageModelResult	One doGenerate result.
ImagePrompt	A string prompt or an image edit prompt.
ImagePromptParts	Normalized image prompt fields.
InvalidArgumentException	Invalid caller argument. Maps to InvalidArgumentError.
InvalidResponseDataException	The provider response did not match the expected shape.
ListedBatch	One listed batch.
MediaTypeDetector	Detects media types from file signatures.
ModelPartStream	Model stream that stays open until Cancel() when Completes is false.
ModelStreamPart	A part emitted by a transcription or translation model stream.
NoImageGeneratedException	No image was produced after retries.
NormalizedBatchRequest	Normalized request sent to doStartBatch.
NoSpeechGeneratedException	No speech audio was produced.
NoTranscriptGeneratedException	No transcript text was produced.
NoTranslationGeneratedException	No translation text or audio was produced.
NoVideoGeneratedException	No video was produced.
OperationCallbacks	Invokes callbacks and swallows their exceptions.
OperationFullStream<T>	A single-consumer stream with an explicit Cancel.
OperationHeaders	Header and retry helpers shared by the operation types.
OperationJson	JSON helpers used while normalizing provider options.
OperationRequest	Shared request fields for an operation call.
OperationRetry	Retries a provider call the way prepareRetries does.
OperationRetryException	Retries were exhausted. LastError is the final attempt.
OperationUsage	Token counts for an image or evaluation call.
OperationWarning	A warning returned by a model call.
PartQueue<T>	Async queue used to hand stream parts to one consumer.
PendingPromise<T>	A promise that can be resolved once.
ProviderFileReference	A provider file reference.
ProviderResponse	Headers and body returned by one provider call.
RankedDocument	One document in the reranked order.
Rerank	Reranks documents. Maps to rerank.
RerankEndEvent	End event for rerank.
RerankModelCall	Arguments for doRerank.
RerankModelCallEvent	Inner rerank telemetry event.
RerankModelDocuments	Documents sent to a reranking model.
RerankModelRank	One ranked index from the model.
RerankModelResponse	Model response for doRerank.
RerankRequest	Options for CancellationToken).
RerankResponseInfo	Response metadata for a rerank call.
RerankResult	Result of CancellationToken).
RerankStartEvent	Start event for rerank.
RerankTelemetry	Telemetry callbacks for rerank.
RetryableCallException	A provider HTTP failure that may be retried.
SearchableTool	A tool the search step can hide until it is discovered.
SkillFile	One file inside a skill upload.
SkillFilePayload	Normalized skill file passed to the provider.
SpeechAudioFormat	Audio format of a speech stream, such as audio/pcm at 24000 Hz or audio/mpeg.
SpeechModelCall	Arguments for doGenerate on a speech model.
SpeechModelResult	Speech model response.
SpeechModels	Resolves string speech model ids.
SpeechModelStreamCall	Arguments for doStream on a speech model. Maps to SpeechModelV4StreamOptions.
SpeechModelStreamPart	A part emitted by a streaming speech model. Maps to SpeechModelV4StreamPart.
SpeechModelStreamResponse	Response data returned by doStream.
SpeechModelStreamResult	The result of a speech model doStream call. Maps to SpeechModelV4StreamResult.
SpeechStreamPart	A part of FullStream.
SpeechStreamUsage	Usage of a speech stream call. Every field is optional because providers bill speech in different units.
SpeechTextPart	One piece of streamed input text. Consecutive parts are synthesized in order as one continuous script. Maps to SpeechTextPart and SpeechModelV4StreamTextPart.
SpeechTextStream	The text a speech model reads. The provider owns it once doStream is called: it must enumerate it to the end or dispose its enumerator on every terminal path.
SpeechTranslationModels	Resolves string speech-translation model ids.
StartVideo	Starts a video generation. Maps to experimentalstartVideo.
StartVideoResult	Result of CancellationToken).
StreamSpeech	Streams speech audio. Maps to experimentalstreamSpeech.
StreamSpeechRequest	Options for CancellationToken).
StreamSpeechResult	Result of CancellationToken). Maps to StreamSpeechResult.
StreamTranscribe	Streams a transcript. Maps to streamTranscribe.
StreamTranscribeRequest	Options for CancellationToken).
StreamTranscribeResult	Streaming transcription result.
StreamTranslate	Streams a speech translation. Maps to streamTranslate.
StreamTranslateRequest	Options for CancellationToken).
StreamTranslateResult	Streaming translation result.
ToolCallerBinding	How a tool may be called.
ToolCallers	Direct tool invocation. An omitted routing entry permits this caller.
ToolSearch	Creates the search tool. Maps to toolSearch.
ToolSearchCall	A planned model tool call.
ToolSearchGeneration	Runs the tool-search step loop used by generateText, streamText, and the tool-loop agent. The model tool list, execution timing, and catalog messages match across those modes.
ToolSearchMatch	One discovered tool. Schemas are not included.
ToolSearchMatchList	Tools returned by one search.
ToolSearchRun	The outcome of a tool-search generation.
ToolSearchState	Per-generation discovery state. Maps to createToolSearchState.
ToolSearchStepItem	A tool result or tool error from one step.
ToolSearchStepRecord	One model step produced by ToolSearchGeneration.
Transcribe	Transcribes audio. Maps to transcribe.
TranscribeRequest	Options for CancellationToken).
TranscribeResult	Transcription result.
TranscriptionModelCall	Arguments for a transcription doGenerate.
TranscriptionModelResult	Transcription model response.
TranscriptionModels	Resolves string transcription model ids.
TranscriptionStreamCall	Arguments for streaming transcription.
TranscriptionStreamStart	A started transcription stream.
TranscriptSegment	One timed transcript segment.
TranslationStreamCall	Arguments for a speech-translation stream.
TranslationStreamStart	A started translation stream.
UnsupportedFunctionalityException	The provider does not implement the requested operation.
UnsupportedModelVersionException	The model specification version is not supported.
UploadData	Tagged file bytes, text, or a stream.
UploadFile	Uploads a file. Maps to uploadFile.
UploadFileCall	Arguments forwarded to a files API.
UploadFileModelResult	Provider upload result.
UploadFileRequest	Options for CancellationToken).
UploadFileResult	Result of CancellationToken).
UploadSkill	Uploads a skill. Maps to uploadSkill.
UploadSkillCall	Arguments for a skill upload.
UploadSkillRequest	Options for CancellationToken).
UploadSkillResult	Skill upload result.
UploadStream	A stream the provider may take ownership of.
VideoCallInputs	Normalized prompt, frames, and references.
VideoFrameFile	A normalized frame image.
VideoFrameImage	A role-tagged frame image.
VideoModelCall	Arguments shared by generate, start, and status.
VideoModelFile	A file or URL passed to a video model.
VideoModelResult	Synchronous doGenerate result or a completed status payload.
VideoOutput	One generated video payload from the model.
VideoPrompt	Text prompt or a prompt with a start image.
VideoReference	Reference media, either raw content or an object with a media type.
VideoStartResult	Result of doStart.
VideoStatusResult	Result of doStatus.
VideoWebhook	Webhook notification.
WarningLog	Warning logger. Observer sees every call, including empty warning lists. Logger maps to globalThis.AISDKLOGWARNINGS.
WarningLogContext	Arguments passed to the warning logger.

Interfaces

Type	Summary
IBatchApi	Batch API.
IBatchProvider	Provider that exposes experimentalbatch().
IDecisionCaller	Decision model used by Decide.
IDecisionProvider	A provider that returns decision models. ProviderRegistry and CustomProvider resolve decision models through it. Maps to the optional decisionModel factory of a provider.
IEmbeddingCaller	Embedding model used by Embed and EmbedMany.
IEvaluationCaller	Deprecated decision model contract. Implement IDecisionCaller instead.
IEvaluationDecisionCaller	A decision model that also has the deprecated CancellationToken). Returned by the deprecated EvaluationModel accessors of ProviderRegistry and CustomProvider. Both methods decide the same way.
IEvaluationProvider	Deprecated provider factory for decision models. Implement IDecisionProvider instead. A provider that implements both is asked through IDecisionProvider only.
IFileStore	Files API.
IFileStoreProvider	Provider that can expose a files API.
IImageCaller	Image model used by GenerateImage.
IRerankCaller	Reranking model used by Rerank.
ISkillStore	Skills API.
ISkillStoreProvider	Provider that can expose a skills API.
ISpeechCaller	Speech model used by GenerateSpeech.
ISpeechTranslationCaller	Speech translation model used by StreamTranslate.
IStreamingSpeechCaller	A speech model that can also stream speech audio for streamed text input. Maps to ExperimentalStreamingSpeechModelV4.
ITranscriptionCaller	Transcription model used by Transcribe.
IVideoCaller	Video model used by generate, start, and status.