---
title: "Package-level declarations"
sidebar_label: "Package-level declarations"
description: "Kotlin binding reference: Package-level declarations."
---
//[clika-runtime](../../index.md)/[io.clika.modelverse](index.md)
# Package-level declarations
## Types
| Name | Summary |
|---|---|
| [AutoConfig](AutoConfig/index.md) | [common]
object [AutoConfig](AutoConfig/index.md)
The identity of the checkpoint at source, from its configuration alone. |
| [AutomaticSpeechRecognitionPipeline](AutomaticSpeechRecognitionPipeline/index.md) | [common]
class [AutomaticSpeechRecognitionPipeline](AutomaticSpeechRecognitionPipeline/index.md) : [Pipeline](Pipeline/index.md)
Speech recognition over an [SttModel](SttModel/index.md): a clip's mono 16-bit samples answer its transcript. |
| [AutoModel](AutoModel/index.md) | [common]
object [AutoModel](AutoModel/index.md)
Load any registered model by name: the family's declared modalities pick the door, or task names it ([Task](Task/index.md)). The answer is the door's handle, a [PreTrainedModel](PreTrainedModel/index.md). A door this binding serves no handle for yet (image segmentation) refuses with `UNSUPPORTED`, naming the task. |
| [AutoModelForCausalLM](AutoModelForCausalLM/index.md) | [common]
object [AutoModelForCausalLM](AutoModelForCausalLM/index.md)
A text-generation model (chat included) by name. |
| [AutoModelForDepthEstimation](AutoModelForDepthEstimation/index.md) | [common]
object [AutoModelForDepthEstimation](AutoModelForDepthEstimation/index.md)
A monocular depth model by name. |
| [AutoModelForImageTextToText](AutoModelForImageTextToText/index.md) | [common]
object [AutoModelForImageTextToText](AutoModelForImageTextToText/index.md)
A multimodal chat model (an image, audio or video in the prompt) by name. |
| [AutoModelForObjectDetection](AutoModelForObjectDetection/index.md) | [common]
object [AutoModelForObjectDetection](AutoModelForObjectDetection/index.md)
A closed-set object detector by name. |
| [AutoModelForSeq2SeqLM](AutoModelForSeq2SeqLM/index.md) | [common]
object [AutoModelForSeq2SeqLM](AutoModelForSeq2SeqLM/index.md)
A translation (sequence-to-sequence) model by name. |
| [AutoModelForSpeakerDiarization](AutoModelForSpeakerDiarization/index.md) | [common]
object [AutoModelForSpeakerDiarization](AutoModelForSpeakerDiarization/index.md)
A speaker-diarization model (who spoke when, over a clip) by name. |
| [AutoModelForSpeechSeq2Seq](AutoModelForSpeechSeq2Seq/index.md) | [common]
object [AutoModelForSpeechSeq2Seq](AutoModelForSpeechSeq2Seq/index.md)
A speech-to-text model by name. |
| [AutoModelForTextEmbedding](AutoModelForTextEmbedding/index.md) | [common]
object [AutoModelForTextEmbedding](AutoModelForTextEmbedding/index.md)
A text-embedding model (one vector per text) by name. |
| [AutoModelForTextRanking](AutoModelForTextRanking/index.md) | [common]
object [AutoModelForTextRanking](AutoModelForTextRanking/index.md)
A reranking model (one relevance score per document against a query) by name. |
| [AutoModelForTextToImage](AutoModelForTextToImage/index.md) | [common]
object [AutoModelForTextToImage](AutoModelForTextToImage/index.md)
An image-generation model (pictures out of a prompt) by name. |
| [AutoModelForTextToVideo](AutoModelForTextToVideo/index.md) | [common]
object [AutoModelForTextToVideo](AutoModelForTextToVideo/index.md)
A video-generation model (a clip out of a prompt) by name. |
| [AutoModelForTextToWaveform](AutoModelForTextToWaveform/index.md) | [common]
object [AutoModelForTextToWaveform](AutoModelForTextToWaveform/index.md)
A text-to-speech model by name. |
| [AutoModelForZeroShotObjectDetection](AutoModelForZeroShotObjectDetection/index.md) | [common]
object [AutoModelForZeroShotObjectDetection](AutoModelForZeroShotObjectDetection/index.md)
An open-vocabulary detector (a text query) by name. |
| [AutoProcessor](AutoProcessor/index.md) | [common]
object [AutoProcessor](AutoProcessor/index.md)
The input processor of the checkpoint at source: the composite of its tokenizer and media processors, from the snapshot's processor files (a hub source fetches them without the weights). |
| [AutoTokenizer](AutoTokenizer/index.md) | [common]
object [AutoTokenizer](AutoTokenizer/index.md)
The tokenizer of the checkpoint at source (a local directory or a hub repo id; a hub source fetches the configuration and tokenizer files alone), the same tokenizer a loaded model of the checkpoint uses. |
| [BackendReport](BackendReport/index.md) | [common]
data class [BackendReport](BackendReport/index.md)(val api: [String](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-string/index.html), val compiled: [Boolean](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-boolean/index.html), val available: [Boolean](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-boolean/index.html), val deviceCount: [Int](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-int/index.html), val unavailableReason: [String](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-string/index.html), val devices: [List](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin.collections/-list/index.html)<[DeviceReport](DeviceReport/index.md)>)
One compute backend of the runtime, as the hardware inventory lists it. |
| [BenchCacheMode](BenchCacheMode/index.md) | [common]
enum [BenchCacheMode](BenchCacheMode/index.md) : [Enum](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-enum/index.html)<[BenchCacheMode](BenchCacheMode/index.md)>
The key-value cache the bench's decoders run on: the paged pool (the serving default, whose prefix cache the `admittedTok` and `retires` columns report) or one window per session. |
| [BenchCell](BenchCell/index.md) | [common]
data class [BenchCell](BenchCell/index.md)(val isl: [Long](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-long/index.html), val osl: [Long](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-long/index.html), val lateIsl: [Long](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-long/index.html) = 0)
One cell of the sweep ([Modelverse.bench](Modelverse/bench.md)): [isl](BenchCell/isl.md) prompt tokens in, [osl](BenchCell/osl.md) tokens decoded, both exact (the prompts are synthesized in the model's own vocabulary at that length and the decode is pinned to the count). A positive [lateIsl](BenchCell/lateIsl.md) adds one more session with a prompt of that many tokens that joins once every regular session has streamed its first token, the shape of a long prompt arriving while others decode; the row then reports what its prefill cost the running sessions (`itlMaxMs`, `itlP99Ms`) and what the prompt waited itself (`lateTtftMs`). The command line spells a cell `ISL:OSL` or `ISL:OSL+LATE`. |
| [BenchColumns](BenchColumns/index.md) | [common]
object [BenchColumns](BenchColumns/index.md)
The column names of a bench report's table, the library's own spellings: the columns of [BenchReport.table](BenchReport/table.md), the keys of [BenchRow.cells](BenchRow/cells.md) and the header of the CSV text. |
| [BenchConfigKeys](BenchConfigKeys/index.md) | [common]
object [BenchConfigKeys](BenchConfigKeys/index.md)
The keys of a bench report's configuration document ([BenchReport.config](BenchReport/config.md)), the library's own spellings. |
| [BenchOptions](BenchOptions/index.md) | [common]
data class [BenchOptions](BenchOptions/index.md)(val cells: [List](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin.collections/-list/index.html)<[BenchCell](BenchCell/index.md)>? = null, val concurrency: [List](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin.collections/-list/index.html)<[Int](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-int/index.html)>? = null, val warmup: [Int](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-int/index.html)? = null, val iters: [Int](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-int/index.html)? = null, val cacheMode: [BenchCacheMode](BenchCacheMode/index.md)? = null, val seed: [Long](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-long/index.html)? = null, val kvQuant: [String](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-string/index.html)? = null, val prefillChunkTokens: [Long](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-long/index.html)? = null, val stepTokenBudget: [Long](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-long/index.html)? = null, val sharedPrefixTokens: [Long](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-long/index.html)? = null, val promptLookupNumTokens: [Long](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-long/index.html)? = null, val uniquePrompts: [Boolean](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-boolean/index.html)? = null, val profile: [Boolean](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-boolean/index.html)? = null, val profilePipeline: [Boolean](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-boolean/index.html)? = null, val profileDir: [String](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-string/index.html)? = null, val reportPrefix: [String](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-string/index.html)? = null, val maxKvCacheBytes: [Long](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-long/index.html)? = null)
The knobs of [Modelverse.bench](Modelverse/bench.md), the command line's `bench` verb's and the Python `modelverse.bench.run`'s: every field null keeps the library's own default (the one the command line prints in its `--help`), so a `BenchOptions()` runs the library's single-cell sweep at concurrency 1. |
| [BenchReport](BenchReport/index.md) | [common]
class [BenchReport](BenchReport/index.md)
One bench's record ([Modelverse.bench](Modelverse/bench.md)), the library's own `BenchReport`: [table](BenchReport/table.md), the runtime's columnar frame with one row per (cell, concurrency) in the sweep's order under the columns of [BenchColumns](BenchColumns/index.md) (the same frame the command line's `bench` verb prints as CSV and writes as `PREFIX.csv`; [toCsvString](BenchReport/toCsvString.md) renders it); [config](BenchReport/config.md), the run's configuration document under the keys of [BenchConfigKeys](BenchConfigKeys/index.md) (the model, the device, the cache mode and the knobs, the machine's hardware inventory, the checkpoint's on-disk bytes and the loaded model's resident bytes; what the verb writes as `PREFIX.json`); [profileSummary](BenchReport/profileSummary.md), the per-op text of a profiled run (empty otherwise); and [rows](BenchReport/rows.md), the table read into typed [BenchRow](BenchRow/index.md)s. The table and the document are native objects: [close](BenchReport/close.md) releases them (`use {}`), and the typed rows stay readable after it. |
| [BenchRow](BenchRow/index.md) | [common]
data class [BenchRow](BenchRow/index.md)(val isl: [Long](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-long/index.html), val osl: [Long](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-long/index.html), val lateIsl: [Long](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-long/index.html), val concurrency: [Int](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-int/index.html), val status: [BenchRowStatus](BenchRowStatus/index.md), val reason: [String](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-string/index.html), val notes: [String](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-string/index.html), val firstCallMs: [Double](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-double/index.html), val ttftMs: [Double](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-double/index.html), val lateTtftMs: [Double](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-double/index.html), val prefillTokS: [Double](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-double/index.html), val prefillChunkMsP50: [Double](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-double/index.html), val prefillChunkMsMax: [Double](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-double/index.html), val itlP50Ms: [Double](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-double/index.html), val itlP99Ms: [Double](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-double/index.html), val itlMaxMs: [Double](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-double/index.html), val decodeTokS: [Double](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-double/index.html), val speedupVsC1: [Double](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-double/index.html), val batchingEfficiency: [Double](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-double/index.html), val effectiveIsl: [Long](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-long/index.html), val admittedTok: [Long](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-long/index.html), val retires: [Long](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-long/index.html), val peakActiveBytes: [Long](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-long/index.html), val numAllocs: [Long](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-long/index.html), val numInflightParkWaits: [Long](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-long/index.html), val acceptedPredictionTokens: [Long](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-long/index.html), val rejectedPredictionTokens: [Long](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-long/index.html), val speculativeSteps: [Long](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-long/index.html), val cells: [Map](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin.collections/-map/index.html)<[String](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-string/index.html), [Any](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-any/index.html)?>)
One row of a bench report, typed: one (cell, concurrency) of the sweep. [status](BenchRow/status.md) says what the row is; the metrics of a row that is not [BenchRowStatus.OK](BenchRowStatus/OK/index.md) are zeros, never measurements, and [reason](BenchRow/reason.md) says why in the library's words. The times are milliseconds, the rates tokens per second. [cells](BenchRow/cells.md) carries every column of the row as the table holds it, by the names of [BenchColumns](BenchColumns/index.md): a `Long` for an integer column, a `Double` for a float one, a `String` for a text one, null for an empty cell; a column a later library adds is there before it has a typed field. The rows are read off [BenchReport.table](BenchReport/table.md) once, so a row outlives the report's close. |
| [BenchRowStatus](BenchRowStatus/index.md) | [common]
enum [BenchRowStatus](BenchRowStatus/index.md) : [Enum](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-enum/index.html)<[BenchRowStatus](BenchRowStatus/index.md)>
What a report row is: a measurement, a cell the loaded window cannot fit, or a run whose pipeline failed. |
| [Box](Box/index.md) | [common]
data class [Box](Box/index.md)(val left: [Float](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-float/index.html), val top: [Float](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-float/index.html), val right: [Float](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-float/index.html), val bottom: [Float](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-float/index.html))
A rectangle in the input picture's pixels, top-left origin, [left](Box/left.md) and [top](Box/top.md) the inclusive edges. |
| [Capabilities](Capabilities/index.md) | [common]
data class [Capabilities](Capabilities/index.md)(val inputModalities: [List](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin.collections/-list/index.html)<[String](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-string/index.html)>, val outputModalities: [List](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin.collections/-list/index.html)<[String](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-string/index.html)>, val hasChatTemplate: [Boolean](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-boolean/index.html), val contextLength: [Int](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-int/index.html), val maxImagesPerRequest: [Int](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-int/index.html), val thinking: [ThinkingControl](ThinkingControl/index.md), val defaults: [GenerationConfig](GenerationConfig/index.md), val family: [String](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-string/index.html), val modelType: [String](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-string/index.html), val device: [String](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-string/index.html), val voices: [List](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin.collections/-list/index.html)<[String](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-string/index.html)> = emptyList(), val languages: [List](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin.collections/-list/index.html)<[String](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-string/index.html)> = emptyList(), val needsReference: [Boolean](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-boolean/index.html) = false, val speakKnobs: [List](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin.collections/-list/index.html)<[SpeakKnob](SpeakKnob/index.md)> = emptyList(), val visionTasks: [List](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin.collections/-list/index.html)<[String](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-string/index.html)> = emptyList(), val labels: [List](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin.collections/-list/index.html)<[String](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-string/index.html)> = emptyList(), val openVocabulary: [Boolean](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-boolean/index.html) = false, val embeddingDim: [Int](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-int/index.html) = 0, val defaultImageWidth: [Int](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-int/index.html) = 0, val defaultImageHeight: [Int](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-int/index.html) = 0, val defaultFrames: [Int](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-int/index.html) = 0, val defaultFps: [Double](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-double/index.html) = 0.0, val metricDepth: [Boolean](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-boolean/index.html) = false, val supportsRepetitionKnobs: [Boolean](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-boolean/index.html) = false, val supportsTools: [Boolean](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-boolean/index.html) = false, val toolCallFormat: [String](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-string/index.html) = "", val templateRendersTools: [Boolean](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-boolean/index.html) = false)
What a loaded model can take and give, read from the checkpoint's documents and the family's registration when the model opens. A consumer greys out the controls a model cannot serve from this and nothing else. |
| [ChatMessage](ChatMessage/index.md) | [common]
data class [ChatMessage](ChatMessage/index.md)(val role: [String](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-string/index.html), val content: [String](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-string/index.html), val images: [List](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin.collections/-list/index.html)<[ByteArray](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-byte-array/index.html)> = emptyList(), val toolCalls: [List](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin.collections/-list/index.html)<[ToolCall](ToolCall/index.md)> = emptyList(), val toolCallId: [String](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-string/index.html) = "", val name: [String](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-string/index.html) = "")
One turn of a conversation. [role](ChatMessage/role.md) is `system`, `user` or `assistant`; [content](ChatMessage/content.md) the text; [images](ChatMessage/images.md) the encoded image files (PNG or JPEG bytes) the turn carries, decoded by the library and placed in the turn where the model's template puts them. A model that takes no images refuses a turn with one, and a request carrying more than [Capabilities.maxImagesPerRequest](Capabilities/maxImagesPerRequest.md) across its turns is refused; either request ends in [GenerationListener.onError](GenerationListener/onError.md). An assistant turn that called tools carries them as [toolCalls](ChatMessage/toolCalls.md); a tool's result is a [ROLE_TOOL](ChatMessage/Companion/ROLE_TOOL.md) turn naming the call it answers ([toolCallId](ChatMessage/toolCallId.md)). |
| [ChatSession](ChatSession/index.md) | [common]
class [ChatSession](ChatSession/index.md)
A conversation over a [GenerativeModel](GenerativeModel/index.md): the turns so far ([messages](ChatSession/messages.md)), a [send](ChatSession/send.md) that appends the user turn and the reply, a [respond](ChatSession/respond.md) that replies to the turns as they stand, the tools channel ([tools](ChatSession/tools.md), [toolCalls](ChatSession/toolCalls.md), [addToolResult](ChatSession/addToolResult.md)) and the last reply's [report](ChatSession/report.md). The whole conversation goes to the model on every turn, rendered through the checkpoint's chat template; the model's cache serves the earlier turns ([GenerationReport.cachedPromptTokens](GenerationReport/cachedPromptTokens.md)). One session speaks at a time, on the model's own worker thread; a second session over the same model waits its turn. The session owns no weights: [close](ChatSession/close.md) ends it and leaves the model open. Built by [GenerativeModel.chatSession](GenerativeModel/chatSession.md). |
| [CheckOptions](CheckOptions/index.md) | [common]
data class [CheckOptions](CheckOptions/index.md)(val device: [String](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-string/index.html)? = null, val maxSeq: [Long](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-long/index.html)? = null, val imageWidth: [Int](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-int/index.html) = 0, val imageHeight: [Int](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-int/index.html) = 0, val devices: [List](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin.collections/-list/index.html)<[FitDevice](FitDevice/index.md)> = emptyList(), val revision: [String](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-string/index.html) = "main", val cacheDir: [String](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-string/index.html)? = null, val token: [String](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-string/index.html)? = null, val offline: [Boolean](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-boolean/index.html) = false, val weights: [String](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-string/index.html)? = null, val dtype: [String](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-string/index.html)? = null) : [SourceOptions](SourceOptions/index.md)
The knobs of [Modelverse.check](Modelverse/check.md), the command line's `check` verb's and the Python `modelverse.check`'s. [device](CheckOptions/device.md) is the device judged: a label the `devices` listing prints (`CPU:0`), the runtime's spelling (`cpu`, `vulkan:1`), an api name alone (`cuda`, every device of that api), or a label of [devices](CheckOptions/devices.md); null judges every device. [maxSeq](CheckOptions/maxSeq.md) is the sequence length judged, null the model's own context length. [imageWidth](CheckOptions/imageWidth.md) and [imageHeight](CheckOptions/imageHeight.md) are the image an image generation pipeline is judged at (0 = its weights alone, with the largest square image that fits reported). [devices](CheckOptions/devices.md) names devices to judge instead of this machine's (a phone's memory, a step's remaining budget); empty takes this machine's inventory. The hub knobs ([revision](CheckOptions/revision.md), [cacheDir](CheckOptions/cacheDir.md), [token](CheckOptions/token.md), [offline](CheckOptions/offline.md), [weights](CheckOptions/weights.md)) are the ones every load reads; nothing is fetched beyond the model's small documents and its file listing, and nothing runs. |
| [DepthEstimationPipeline](DepthEstimationPipeline/index.md) | [common]
class [DepthEstimationPipeline](DepthEstimationPipeline/index.md) : [Pipeline](Pipeline/index.md)
Depth estimation over a [VisionModel](VisionModel/index.md): a picture answers its depth map. |
| [DepthMap](DepthMap/index.md) | [common]
data class [DepthMap](DepthMap/index.md)(val width: [Int](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-int/index.html), val height: [Int](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-int/index.html), val values: [FloatArray](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-float-array/index.html), val metric: [Boolean](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-boolean/index.html), val min: [Float](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-float/index.html), val max: [Float](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-float/index.html))
A depth value per pixel of the input picture, row by row ([width](DepthMap/width.md) × [height](DepthMap/height.md) values): meters when [metric](DepthMap/metric.md), else relative depth where a larger value is closer; [min](DepthMap/min.md) and [max](DepthMap/max.md) are the map's own range. |
| [Detection](Detection/index.md) | [common]
data class [Detection](Detection/index.md)(val label: [String](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-string/index.html), val labelId: [Long](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-long/index.html), val score: [Float](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-float/index.html), val box: [Box](Box/index.md))
One detected object: the label (empty when the model's table names none for the id), its id, the score in 0 to 1, and the box in the input picture's pixels. |
| [DetectOptions](DetectOptions/index.md) | [common]
data class [DetectOptions](DetectOptions/index.md)(val maxResults: [Int](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-int/index.html) = DEFAULT_MAX_RESULTS, val scoreThreshold: [Float](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-float/index.html) = DEFAULT_SCORE_THRESHOLD, val labels: [List](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin.collections/-list/index.html)<[String](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-string/index.html)> = emptyList())
How a detection runs. The engine floors the scores at [scoreThreshold](DetectOptions/scoreThreshold.md), keeps the strongest [maxResults](DetectOptions/maxResults.md) and answers them score-descending. |
| [DeviceMemory](DeviceMemory/index.md) | [common]
data class [DeviceMemory](DeviceMemory/index.md)(val device: [String](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-string/index.html), val totalBytes: [Long](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-long/index.html), val freeBytes: [Long](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-long/index.html), val runtimeBytes: [Long](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-long/index.html), val activeBytes: [Long](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-long/index.html))
A device's memory as the runtime reads it ([Modelverse.deviceMemory](Modelverse/deviceMemory.md)): the physical total, what is free, the runtime's own share and the pool bytes in use. |
| [DeviceReport](DeviceReport/index.md) | [common]
data class [DeviceReport](DeviceReport/index.md)(val index: [Int](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-int/index.html), val name: [String](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-string/index.html), val totalMemoryBytes: [Long](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-long/index.html))
One compute device of a backend, as the hardware inventory lists it. |
| [DiarizationModel](DiarizationModel/index.md) | [common]
class [DiarizationModel](DiarizationModel/index.md) : [PreTrainedModel](PreTrainedModel/index.md)
A speaker-diarization model: a clip in, who spoke when out, as [SpeakerSegment](SpeakerSegment/index.md)s in time order (the `voice-activity-detection` kind). |
| [DiarizeHandle](DiarizeHandle/index.md) | [common]
interface [DiarizeHandle](DiarizeHandle/index.md)
A speaker-diarization request in flight. |
| [DiarizeListener](DiarizeListener/index.md) | [common]
interface [DiarizeListener](DiarizeListener/index.md)
The calls a speaker-diarization request makes on the model's worker thread, in order: [onSegments](DiarizeListener/onSegments.md) once with who spoke when, in time order, then [onDone](DiarizeListener/onDone.md); or [onError](DiarizeListener/onError.md) alone when the request fails. |
| [DiarizeOptions](DiarizeOptions/index.md) | [common]
data class [DiarizeOptions](DiarizeOptions/index.md)(val streaming: [Boolean](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-boolean/index.html) = false, val streamingMode: [String](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-string/index.html)? = null, val threshold: [Float](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-float/index.html) = 0.5f)
How a diarization runs. [streaming](DiarizeOptions/streaming.md) runs the clip chunk by chunk with the speaker cache, as a live stream does, [streamingMode](DiarizeOptions/streamingMode.md) naming the latency mode (the checkpoint's default when null); [threshold](DiarizeOptions/threshold.md) in (0, 1) is the activity a frame needs to count for a speaker. |
| [DiarizeReport](DiarizeReport/index.md) | [common]
data class [DiarizeReport](DiarizeReport/index.md)(val latencyMs: [Float](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-float/index.html), val accelerator: [String](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-string/index.html), val canceled: [Boolean](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-boolean/index.html))
The facts of one speaker-diarization request. |
| [DownloadEvent](DownloadEvent/index.md) | [common]
data class [DownloadEvent](DownloadEvent/index.md)(val file: [String](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-string/index.html), val index: [Int](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-int/index.html), val count: [Int](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-int/index.html), val done: [Long](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-long/index.html), val total: [Long](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-long/index.html), val planBytesDone: [Long](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-long/index.html), val planBytesTotal: [Long](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-long/index.html), val isWeight: [Boolean](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-boolean/index.html))
One step of a download ([DownloadListener](DownloadListener/index.md)): the file transferring (repo-relative), its position in the plan so far ([index](DownloadEvent/index-property.md) of [count](DownloadEvent/count.md); the plan grows as the resolve names its missing files), its bytes, and the plan's bytes done and total ([planBytesTotal](DownloadEvent/planBytesTotal.md) is 0 while a planned file's size is unknown). A cached file never joins the plan. |
| [DownloadListener](DownloadListener/index.md) | [common]
fun interface [DownloadListener](DownloadListener/index.md)
Hears a download's progress ([Modelverse.snapshot](Modelverse/snapshot.md)): one call per transfer step of each file the resolve fetches, on the calling thread, with the file's bytes and the plan's. Return `false` to stop the download: the call then refuses with the hub's own cancel status; an exception the listener lets escape stops it the same way. A snapshot served from the cache reports nothing, since a cached file never joins the plan. |
| [EmbeddingHandle](EmbeddingHandle/index.md) | [common]
interface [EmbeddingHandle](EmbeddingHandle/index.md)
A text-embedding request in flight. |
| [EmbeddingListener](EmbeddingListener/index.md) | [common]
interface [EmbeddingListener](EmbeddingListener/index.md)
The calls a text-embedding request makes on the model's worker thread, in order: [onEmbeddings](EmbeddingListener/onEmbeddings.md) once with one vector per text (in the texts' order), then [onDone](EmbeddingListener/onDone.md); or [onError](EmbeddingListener/onError.md) alone when the request fails. |
| [EmbeddingModel](EmbeddingModel/index.md) | [common]
class [EmbeddingModel](EmbeddingModel/index.md) : [PreTrainedModel](PreTrainedModel/index.md)
A text-embedding model: one vector per text, the same length for every text ([Capabilities.embeddingDim](Capabilities/embeddingDim.md)), so two texts compare by the cosine of their vectors. |
| [EmbeddingReport](EmbeddingReport/index.md) | [common]
data class [EmbeddingReport](EmbeddingReport/index.md)(val latencyMs: [Float](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-float/index.html), val accelerator: [String](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-string/index.html), val canceled: [Boolean](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-boolean/index.html))
The facts of one text-embedding request. |
| [FeatureExtractionPipeline](FeatureExtractionPipeline/index.md) | [common]
class [FeatureExtractionPipeline](FeatureExtractionPipeline/index.md) : [Pipeline](Pipeline/index.md)
Text embeddings over an [EmbeddingModel](EmbeddingModel/index.md): a text answers its vector, a list of texts one vector each, in order (`feature-extraction` and its alias `sentence-similarity`). |
| [FinishReason](FinishReason/index.md) | [common]
enum [FinishReason](FinishReason/index.md) : [Enum](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-enum/index.html)<[FinishReason](FinishReason/index.md)>
How a generation ended. |
| [FitColumns](FitColumns/index.md) | [common]
object [FitColumns](FitColumns/index.md)
The column names of [FitReport.table](FitReport/table.md), the library's own spellings. |
| [FitDevice](FitDevice/index.md) | [common]
data class [FitDevice](FitDevice/index.md)(val label: [String](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-string/index.html), val totalBytes: [Long](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-long/index.html), val cores: [Int](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-int/index.html) = 0)
One device to judge a model against: [label](FitDevice/label.md) as the `devices` listing prints it (`CPU:0`, `Vulkan:1`), or any name the caller chooses for a device that is not this machine (a phone, or the memory left for one step of a pipeline after the earlier steps took theirs); [totalBytes](FitDevice/totalBytes.md) its memory; [cores](FitDevice/cores.md) the CPU cores it decodes with (0 = unknown, no many-cores bonus in the speed estimate). |
| [FitGrade](FitGrade/index.md) | [common]
enum [FitGrade](FitGrade/index.md) : [Enum](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-enum/index.html)<[FitGrade](FitGrade/index.md)>
How comfortably a model's need sits in a device's memory, read from the memory share at the admitted context: at most 60 percent is [PERFECT](FitGrade/PERFECT/index.md), at most 85 [GOOD](FitGrade/GOOD/index.md), at most 98 [MARGINAL](FitGrade/MARGINAL/index.md), above it [TOO_TIGHT](FitGrade/TOO_TIGHT/index.md) (no room left for the allocator), and a need no context of the device holds is [DOES_NOT_FIT](FitGrade/DOES_NOT_FIT/index.md). [word](FitGrade/word.md) is the library's own spelling, the one the command line's `check` verb prints. |
| [FitReport](FitReport/index.md) | [common]
class [FitReport](FitReport/index.md)
The answer of [Modelverse.check](Modelverse/check.md): [verdicts](FitReport/verdicts.md), one per device judged, in the request's order; and [table](FitReport/table.md), the runtime's `Table` the command line's `check` verb prints (one row per verdict, sorted by grade then memory share, the columns of [FitColumns](FitColumns/index.md); [toCsvString](FitReport/toCsvString.md) renders it). The table is a native object: [close](FitReport/close.md) releases it (`use {}`); the verdicts stay. |
| [FitVerdict](FitVerdict/index.md) | [common]
data class [FitVerdict](FitVerdict/index.md)(val source: [String](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-string/index.html), val device: [String](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-string/index.html), val deviceBytes: [Long](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-long/index.html), val weightBytes: [Long](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-long/index.html), val kvBytesFixed: [Long](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-long/index.html), val kvBytesPerToken: [Long](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-long/index.html), val kvDtype: [String](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-string/index.html), val contextLength: [Long](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-long/index.html), val maxSeq: [Long](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-long/index.html), val activationBytes: [Long](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-long/index.html), val needBytes: [Long](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-long/index.html), val fits: [Boolean](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-boolean/index.html), val maxSeqFits: [Long](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-long/index.html), val maxImageFits: [Int](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-int/index.html), val reason: [String](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-string/index.html), val judged: [Boolean](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-boolean/index.html), val grade: [FitGrade](FitGrade/index.md)?, val memorySharePercent: [Int](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-int/index.html), val parameters: [Long](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-long/index.html), val weightEncoding: [String](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-string/index.html), val tokSEstimate: [Double](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-double/index.html), val tokSMeasured: [Double](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-double/index.html))
Does a model fit a device, judged from its metadata alone, the library's own `FitVerdict`. The law, every term a field: |
| [GeneratedImage](GeneratedImage/index.md) | [common]
data class [GeneratedImage](GeneratedImage/index.md)(val width: [Int](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-int/index.html), val height: [Int](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-int/index.html), val rgba8888: [ByteArray](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-byte-array/index.html))
One generated picture: [width](GeneratedImage/width.md) by [height](GeneratedImage/height.md) pixels, four bytes each in the order red, green, blue, alpha (the alpha opaque). |
| [GeneratedImages](GeneratedImages/index.md) | [common]
data class [GeneratedImages](GeneratedImages/index.md)(val images: [List](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin.collections/-list/index.html)<[GeneratedImage](GeneratedImage/index.md)>, val seed: [Long](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-long/index.html))
The pictures of one request and the [seed](GeneratedImages/seed.md) they were drawn from. |
| [GeneratedVideo](GeneratedVideo/index.md) | [common]
data class [GeneratedVideo](GeneratedVideo/index.md)(val frames: [List](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin.collections/-list/index.html)<[GeneratedImage](GeneratedImage/index.md)>, val fps: [Double](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-double/index.html), val seed: [Long](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-long/index.html), val audioPcm16: [ShortArray](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-short-array/index.html)?, val audioSampleRate: [Int](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-int/index.html))
One generated clip: its [frames](GeneratedVideo/frames.md) in order (each a [GeneratedImage](GeneratedImage/index.md)), the [fps](GeneratedVideo/fps.md) they play at, the [seed](GeneratedVideo/seed.md) they were drawn from, and the sound as 16-bit mono samples at [audioSampleRate](GeneratedVideo/audioSampleRate.md) hertz where the family answers one ([audioPcm16](GeneratedVideo/audioPcm16.md) null and the rate 0 otherwise). |
| [GenerationConfig](GenerationConfig/index.md) | [common]
data class [GenerationConfig](GenerationConfig/index.md)(val maxTokens: [Int](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-int/index.html) = 0, val temperature: [Float](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-float/index.html) = 0.0f, val topK: [Int](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-int/index.html) = 0, val topP: [Float](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-float/index.html) = 1.0f, val repetitionPenalty: [Float](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-float/index.html) = 1.0f, val thinking: [Boolean](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-boolean/index.html)? = null, val noRepeatNgramSize: [Int](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-int/index.html) = 0, val noRepeatNgramWindow: [Int](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-int/index.html) = 0)
The decode policy of one request. A zero [maxTokens](GenerationConfig/maxTokens.md) or [topK](GenerationConfig/topK.md) keeps the model's own default for that knob ([Capabilities.defaults](Capabilities/defaults.md)); [temperature](GenerationConfig/temperature.md) and [topP](GenerationConfig/topP.md) apply as given, so a caller that wants the model's defaults for them copies them from [Capabilities.defaults](Capabilities/defaults.md). |
| [GenerationHandle](GenerationHandle/index.md) | [common]
interface [GenerationHandle](GenerationHandle/index.md)
A running generation. [cancel](GenerationHandle/cancel.md) may be called from any thread: the decode stops at its next step, the listener receives the text streamed so far, and [GenerationListener.onDone](GenerationListener/onDone.md) arrives with [FinishReason.CANCELED](FinishReason/CANCELED/index.md). Calling it after the generation ended is a no-op. |
| [GenerationListener](GenerationListener/index.md) | [common]
interface [GenerationListener](GenerationListener/index.md)
The stream of one generation. The calls arrive in order on the model's own worker thread, never on the thread that called [GenerativeModel.chat](GenerativeModel/chat.md): [onToken](GenerationListener/onToken.md) and [onThinking](GenerationListener/onThinking.md) as text is produced, then exactly one of [onDone](GenerationListener/onDone.md) or [onError](GenerationListener/onError.md). A listener that needs the main thread posts to it. |
| [GenerationReport](GenerationReport/index.md) | [common]
data class [GenerationReport](GenerationReport/index.md)(val finish: [FinishReason](FinishReason/index.md), val promptTokens: [Int](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-int/index.html), val completionTokens: [Int](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-int/index.html), val timeToFirstTokenMs: [Float](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-float/index.html), val prefillTokensPerSecond: [Float](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-float/index.html), val decodeTokensPerSecond: [Float](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-float/index.html), val latencyMs: [Float](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-float/index.html), val accelerator: [String](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-string/index.html), val cachedPromptTokens: [Int](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-int/index.html) = 0, val text: [String](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-string/index.html) = "", val reasoning: [String](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-string/index.html) = "", val visibleText: [String](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-string/index.html) = "", val toolCalls: [List](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin.collections/-list/index.html)<[ToolCall](ToolCall/index.md)> = emptyList(), val toolsCalled: [Boolean](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-boolean/index.html) = false)
The counts and timings of one generation, delivered once through [GenerationListener.onDone](GenerationListener/onDone.md). |
| [GenerativeModel](GenerativeModel/index.md) | [common]
class [GenerativeModel](GenerativeModel/index.md) : [PreTrainedModel](PreTrainedModel/index.md)
A text model loaded from its files on disk, on the CPU. |
| [HardwareReport](HardwareReport/index.md) | [common]
data class [HardwareReport](HardwareReport/index.md)(val runtimeVersion: [String](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-string/index.html), val physicalCores: [Int](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-int/index.html), val logicalCores: [Int](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-int/index.html), val backends: [List](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin.collections/-list/index.html)<[BackendReport](BackendReport/index.md)>)
The inventory of this machine's compute ([Modelverse.hardware](Modelverse/hardware.md)): the runtime's version, the CPU's cores and every backend with its devices. |
| [ImageFeatureExtractionPipeline](ImageFeatureExtractionPipeline/index.md) | [common]
class [ImageFeatureExtractionPipeline](ImageFeatureExtractionPipeline/index.md) : [Pipeline](Pipeline/index.md)
Image embeddings over a [VisionModel](VisionModel/index.md): a picture answers its vector. |
| [ImageGenerationHandle](ImageGenerationHandle/index.md) | [common]
interface [ImageGenerationHandle](ImageGenerationHandle/index.md)
An image-generation request in flight. |
| [ImageGenerationListener](ImageGenerationListener/index.md) | [common]
interface [ImageGenerationListener](ImageGenerationListener/index.md)
The calls an image-generation request makes on the model's worker thread, in order: [onImages](ImageGenerationListener/onImages.md) once with the pictures and the seed the model drew, then [onDone](ImageGenerationListener/onDone.md); or [onError](ImageGenerationListener/onError.md) alone when the request fails. |
| [ImageGenerationModel](ImageGenerationModel/index.md) | [common]
class [ImageGenerationModel](ImageGenerationModel/index.md) : [PreTrainedModel](PreTrainedModel/index.md)
An image-generation model: a prompt in, pictures out (the `text-to-image` kind), each as the RGBA pixels an Android bitmap takes. |
| [ImageGenerationOptions](ImageGenerationOptions/index.md) | [common]
data class [ImageGenerationOptions](ImageGenerationOptions/index.md)(val negativePrompt: [String](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-string/index.html)? = null, val width: [Int](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-int/index.html) = 0, val height: [Int](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-int/index.html) = 0, val steps: [Int](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-int/index.html) = 0, val guidanceScale: [Float](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-float/index.html)? = null, val seed: [Long](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-long/index.html) = -1L, val count: [Int](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-int/index.html) = 1)
How pictures are generated. A knob left at its zero keeps the model's own default: [width](ImageGenerationOptions/width.md) and [height](ImageGenerationOptions/height.md) in pixels (the family's size law applies; see [Capabilities.defaultImageWidth](Capabilities/defaultImageWidth.md) and [Capabilities.defaultImageHeight](Capabilities/defaultImageHeight.md)), [steps](ImageGenerationOptions/steps.md) the denoising steps, [guidanceScale](ImageGenerationOptions/guidanceScale.md) the prompt's pull (null keeps the model's), [seed](ImageGenerationOptions/seed.md) the draw (-1 lets the model pick one; the answer names it), [count](ImageGenerationOptions/count.md) the number of pictures, [negativePrompt](ImageGenerationOptions/negativePrompt.md) what to steer away from (null for none). |
| [ImageGenerationReport](ImageGenerationReport/index.md) | [common]
data class [ImageGenerationReport](ImageGenerationReport/index.md)(val latencyMs: [Float](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-float/index.html), val accelerator: [String](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-string/index.html), val canceled: [Boolean](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-boolean/index.html))
The facts of one image-generation request. |
| [ImageInput](ImageInput/index.md) | [common]
sealed interface [ImageInput](ImageInput/index.md)
A picture handed to a vision model: the bitmap's pixels, or the encoded file. |
| [ImageToTextPipeline](ImageToTextPipeline/index.md) | [common]
class [ImageToTextPipeline](ImageToTextPipeline/index.md) : [Pipeline](Pipeline/index.md)
Text recognition over a [VisionModel](VisionModel/index.md): a picture answers the text read off it. |
| [LabelScore](LabelScore/index.md) | [common]
data class [LabelScore](LabelScore/index.md)(val label: [String](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-string/index.html), val score: [Float](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-float/index.html))
One label with its score against the picture (the cosine of the two embeddings), descending in a result. |
| [LoadOptions](LoadOptions/index.md) | [common]
data class [LoadOptions](LoadOptions/index.md)(val contextLength: [Int](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-int/index.html) = 0, val maxActive: [Int](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-int/index.html) = 0, val thinking: [Boolean](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-boolean/index.html)? = null, val revision: [String](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-string/index.html) = "main", val cacheDir: [String](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-string/index.html)? = null, val token: [String](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-string/index.html)? = null, val offline: [Boolean](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-boolean/index.html) = false, val device: [String](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-string/index.html)? = null, val weights: [String](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-string/index.html)? = null, val dtype: [String](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-string/index.html)? = null) : [SourceOptions](SourceOptions/index.md)
The load-time choices a caller makes for a generative model. Every field has the library's own default; a zero or a null keeps it. |
| [Modelverse](Modelverse/index.md) | [common]
object [Modelverse](Modelverse/index.md)
The entry of the Modelverse binding: loads the three native libraries in the one order that works, then answers the library's version. |
| [ModelverseException](ModelverseException/index.md) | [common]
class [ModelverseException](ModelverseException/index.md)(message: [String](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-string/index.html), val codeName: [String](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-string/index.html) = "", cause: [Throwable](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-throwable/index.html)? = null)
A Modelverse failure, in words a person can act on: a model directory that holds no checkpoint the library serves, a weight file the runtime cannot decode, a request past the context window, a library that did not load. |
| [ObjectDetectionPipeline](ObjectDetectionPipeline/index.md) | [common]
class [ObjectDetectionPipeline](ObjectDetectionPipeline/index.md) : [Pipeline](Pipeline/index.md)
Detection over a [VisionModel](VisionModel/index.md): a picture answers its detections; the candidateLabels name the phrases an open-vocabulary detector looks for. |
| [OcrOptions](OcrOptions/index.md) | [common]
data class [OcrOptions](OcrOptions/index.md)(val maxNewTokens: [Int](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-int/index.html) = 0)
How text recognition runs. |
| [Pipeline](Pipeline/index.md) | [common]
sealed class [Pipeline](Pipeline/index.md)
A task pipeline: one call that runs a loaded model's door end to end and answers the result, under the task's own name ([Modelverse.pipeline](Modelverse/pipeline.md) with a task string: `text-generation`, `automatic-speech-recognition`, `text-to-speech`, `translation`, `image-to-text`, `object-detection`, `depth-estimation`, `image-feature-extraction`, `zero-shot-image-classification`, `feature-extraction`, `text-ranking`, `voice-activity-detection`, `text-to-image`, `text-to-video`, and their aliases). Each pipeline's `invoke` blocks on the caller's thread for the model's reply; a failure throws [ModelverseException](ModelverseException/index.md) with the model's message. The pipeline owns its [model](Pipeline/model.md): [close](Pipeline/close.md) closes it. The handle's own doors stay reachable through [model](Pipeline/model.md) for a caller that streams. |
| [PretrainedConfig](PretrainedConfig/index.md) | [common]
data class [PretrainedConfig](PretrainedConfig/index.md)(val family: [String](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-string/index.html), val modelType: [String](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-string/index.html), val architecture: [String](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-string/index.html), val vendor: [String](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-string/index.html), val description: [String](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-string/index.html), val licenseUrl: [String](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-string/index.html), val pipelineTags: [List](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin.collections/-list/index.html)<[String](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-string/index.html)>, val inputModalities: [List](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin.collections/-list/index.html)<[String](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-string/index.html)>, val outputModalities: [List](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin.collections/-list/index.html)<[String](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-string/index.html)>, val isMoe: [Boolean](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-boolean/index.html), val isMultimodal: [Boolean](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-boolean/index.html), val hasTokenizer: [Boolean](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-boolean/index.html), val hasImageProcessor: [Boolean](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-boolean/index.html), val hasAudioProcessor: [Boolean](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-boolean/index.html), val hasVideoProcessor: [Boolean](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-boolean/index.html), val numHiddenLayers: [Int](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-int/index.html), val hiddenSize: [Int](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-int/index.html), val numAttentionHeads: [Int](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-int/index.html), val numKeyValueHeads: [Int](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-int/index.html), val slidingWindow: [Long](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-long/index.html), val numExperts: [Int](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-int/index.html), val runnable: [Boolean](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-boolean/index.html), val identityOnlyPurpose: [String](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-string/index.html))
A checkpoint's identity as the registry resolves it from its configuration alone, before any weight is read: the family that serves it, its model type and architecture, the family's description, the modalities it takes and gives, the variant's facts, what its processor carries, and whether the family runs in this binding ([runnable](PretrainedConfig/runnable.md); [identityOnlyPurpose](PretrainedConfig/identityOnlyPurpose.md) says what an identity-only family is for). |
| [PreTrainedModel](PreTrainedModel/index.md) | [common]
sealed interface [PreTrainedModel](PreTrainedModel/index.md)
A loaded model of any door: its [config](PreTrainedModel/config.md) (the identity the registry resolved for its source) and close. The handles are [GenerativeModel](GenerativeModel/index.md), [SttModel](SttModel/index.md), [TtsModel](TtsModel/index.md), [VisionModel](VisionModel/index.md) and [TranslateModel](TranslateModel/index.md). |
| [RegisteredFamily](RegisteredFamily/index.md) | [common]
data class [RegisteredFamily](RegisteredFamily/index.md)(val family: [String](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-string/index.html), val vendor: [String](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-string/index.html), val runnable: [Boolean](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-boolean/index.html), val modelTypes: [List](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin.collections/-list/index.html)<[String](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-string/index.html)>, val architectures: [List](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin.collections/-list/index.html)<[String](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-string/index.html)>, val ggufArchitectures: [List](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin.collections/-list/index.html)<[String](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-string/index.html)>, val inputModalities: [List](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin.collections/-list/index.html)<[String](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-string/index.html)>, val outputModalities: [List](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin.collections/-list/index.html)<[String](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-string/index.html)>, val description: [String](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-string/index.html) = "", val licenseUrl: [String](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-string/index.html) = "", val pipelineTags: [List](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin.collections/-list/index.html)<[String](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-string/index.html)> = emptyList(), val identityOnlyPurpose: [String](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-string/index.html) = "")
One model family the model library registers, as [Modelverse.registry](Modelverse/registry.md) lists it: the identity keys a checkpoint is matched by, whether the family carries a runnable path, and its modalities. A checkpoint whose `model_type` or `architectures[0]` (a `config.json`) or `general.architecture` (a GGUF header) is in no runnable family's keys has no template in this library and will be refused at load; a key that is listed names a template, and the load may still decline a checkpoint on its own facts (a weight format the runtime does not decode, a model type the family registers only to refuse). The per-release supported-models document is the measured verdict. |
| [RerankerModel](RerankerModel/index.md) | [common]
class [RerankerModel](RerankerModel/index.md) : [PreTrainedModel](PreTrainedModel/index.md)
A reranking model: a query and documents in, one relevance score per document out, in the documents' order (a higher score is a better match; the caller sorts). |
| [RerankHandle](RerankHandle/index.md) | [common]
interface [RerankHandle](RerankHandle/index.md)
A reranking request in flight. |
| [RerankListener](RerankListener/index.md) | [common]
interface [RerankListener](RerankListener/index.md)
The calls a reranking request makes on the model's worker thread, in order: [onScores](RerankListener/onScores.md) once with one relevance score per document (in the documents' order), then [onDone](RerankListener/onDone.md); or [onError](RerankListener/onError.md) alone when the request fails. |
| [RerankReport](RerankReport/index.md) | [common]
data class [RerankReport](RerankReport/index.md)(val latencyMs: [Float](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-float/index.html), val accelerator: [String](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-string/index.html), val canceled: [Boolean](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-boolean/index.html))
The facts of one reranking request. |
| [ResidencyPart](ResidencyPart/index.md) | [common]
data class [ResidencyPart](ResidencyPart/index.md)(val name: [String](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-string/index.html), val device: [String](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-string/index.html), val weightsBytes: [Long](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-long/index.html), val kvCacheBytes: [Long](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-long/index.html), val otherBytes: [Long](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-long/index.html), val totalBytes: [Long](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-long/index.html))
What one part of a loaded model holds on one device ([ResidencyReport](ResidencyReport/index.md)). |
| [ResidencyReport](ResidencyReport/index.md) | [common]
data class [ResidencyReport](ResidencyReport/index.md)(val parts: [List](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin.collections/-list/index.html)<[ResidencyPart](ResidencyPart/index.md)>, val totalBytes: [Long](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-long/index.html))
What a loaded model holds on each device, as the runtime's pools report it ([PreTrainedModel.residency](PreTrainedModel/residency.md)). The weights are the pool growth the load left on the device; two loads running at once on one device are not told apart by that read, so load one at a time where the number matters. |
| [ServeOptions](ServeOptions/index.md) | [common]
data class [ServeOptions](ServeOptions/index.md)(val host: [String](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-string/index.html) = "127.0.0.1", val port: [Int](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-int/index.html) = 8000, val modelId: [String](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-string/index.html) = "", val maxActive: [Int](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-int/index.html) = 4, val maxQueued: [Int](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-int/index.html) = 256, val defaultBudget: [Int](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-int/index.html) = 256, val enableWebUi: [Boolean](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-boolean/index.html) = true, val enableCors: [Boolean](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-boolean/index.html) = false, val logRequestTiming: [Boolean](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-boolean/index.html) = false, val maxUploadBytes: [Long](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-long/index.html) = 0, val readTimeoutSeconds: [Int](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-int/index.html) = 0)
The server's knobs ([Modelverse.serve](Modelverse/serve.md)): the bind address, the advertised model id (the family when empty), the engine's width (requests decoding at once), its queue and its default decode budget, the web page, CORS, the request timing log, and the upload and read limits (0 keeps the library's). |
| [Server](Server/index.md) | [common]
class [Server](Server/index.md)
The server over a loaded [GenerativeModel](GenerativeModel/index.md) ([Modelverse.serve](Modelverse/serve.md)): the chat API (`/v1/chat/completions`, the models listing, `/health`, the web page when enabled) on its own threads. [start](Server/start.md) listens and returns at once; [waitUntilReady](Server/waitUntilReady.md) probes `/health`; [stop](Server/stop.md) ends the listening and joins the server's threads; [close](Server/close.md) stops and releases the served engine. The model stays open, the caller's, and refuses its own `close()` while a server holds it: close the server first. Several servers may serve one model. [http](Server/http.md) is the server's own http server, borrowed, for routes and middleware of the app's beside the API's, registered before [start](Server/start.md). |
| [Snapshot](Snapshot/index.md) | [common]
data class [Snapshot](Snapshot/index.md)(val localDir: [String](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-string/index.html), val files: [List](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin.collections/-list/index.html)<[String](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-string/index.html)>, val weightsPrimary: [String](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-string/index.html)?, val licenseLine: [String](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-string/index.html)?, val companions: [List](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin.collections/-list/index.html)<[SnapshotCompanion](SnapshotCompanion/index.md)>)
What [Modelverse.snapshot](Modelverse/snapshot.md) answers: the directory the model files rest in and the file names in it; the selected weight file on the GGUF lane (null elsewhere); the checkpoint's license line as the command line logs it (null when the facts were unavailable); and the companions, each with its outcome. |
| [SnapshotCompanion](SnapshotCompanion/index.md) | [common]
data class [SnapshotCompanion](SnapshotCompanion/index.md)(val repo: [String](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-string/index.html), val localDir: [String](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-string/index.html), val files: [Int](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-int/index.html), val failure: [String](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-string/index.html), val failureCode: [String](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-string/index.html))
One companion of a snapshot: a hub repository the model's family needs beside the source (a codec, say), where it landed, and the hub's refusal when it did not ([failure](SnapshotCompanion/failure.md) empty means it landed, or was cached). |
| [SourceOptions](SourceOptions/index.md) | [common]
interface [SourceOptions](SourceOptions/index.md)
The source knobs every load and resolve reads, the hub ecosystem's own: a source is a local directory, a hub repo id or a `.gguf` path; [revision](SourceOptions/revision.md) is the hub revision (a branch, a tag or a commit), [cacheDir](SourceOptions/cacheDir.md) the hub cache (null for the hub's default), [token](SourceOptions/token.md) a hub access token passed through to the download (null for the hub's own token chain), [offline](SourceOptions/offline.md) answers from the cache alone, and [device](SourceOptions/device.md) is the runtime's device spelling ("cpu", "vulkan:0"; null keeps the door's default). A local source ignores the hub knobs. |
| [SpeakerSegment](SpeakerSegment/index.md) | [common]
data class [SpeakerSegment](SpeakerSegment/index.md)(val speaker: [Int](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-int/index.html), val startMs: [Long](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-long/index.html), val endMs: [Long](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-long/index.html))
One speaker's turn: who ([speaker](SpeakerSegment/speaker.md), a 0-based index the model assigns) spoke from [startMs](SpeakerSegment/startMs.md) to [endMs](SpeakerSegment/endMs.md). |
| [SpeakHandle](SpeakHandle/index.md) | [common]
interface [SpeakHandle](SpeakHandle/index.md)
A running synthesis. [cancel](SpeakHandle/cancel.md) may be called from any thread. The model library runs an utterance to its end, so a cancel does not shorten the wait: the clip is dropped when the synthesis returns, [SpeakListener.onClip](SpeakListener/onClip.md) is not called, and [SpeakListener.onDone](SpeakListener/onDone.md) arrives with [SpeakReport.canceled](SpeakReport/canceled.md) true. |
| [SpeakKnob](SpeakKnob/index.md) | [common]
data class [SpeakKnob](SpeakKnob/index.md)(val name: [String](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-string/index.html), val min: [Float](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-float/index.html), val max: [Float](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-float/index.html)?, val keepsDefault: [Float](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-float/index.html)?)
One synthesis knob a text-to-speech model reads ([Capabilities.speakKnobs](Capabilities/speakKnobs.md)), named as the constants here spell it, with the range the binding accepts. |
| [SpeakListener](SpeakListener/index.md) | [common]
interface [SpeakListener](SpeakListener/index.md)
The outcome of one synthesis. The calls arrive on the model's own worker thread: [onClip](SpeakListener/onClip.md) with the utterance, then exactly one of [onDone](SpeakListener/onDone.md) or [onError](SpeakListener/onError.md). This release of the model library answers the whole utterance in one call, so [onClip](SpeakListener/onClip.md) is called once; a library that produces the audio in chunks calls it once per chunk, in order, on the same listener. |
| [SpeakOptions](SpeakOptions/index.md) | [common]
data class [SpeakOptions](SpeakOptions/index.md)(val exaggeration: [Float](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-float/index.html)? = null, val temperature: [Float](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-float/index.html)? = null, val cfgScale: [Float](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-float/index.html)? = null, val steps: [Int](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-int/index.html) = 0, val seed: [Long](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-long/index.html) = -1, val maxSpeechTokens: [Int](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-int/index.html) = 0, val language: [String](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-string/index.html)? = null)
The synthesis knobs of one request. A null knob, or the value [SpeakKnob.keepsDefault](SpeakKnob/keepsDefault.md) names for it, keeps the family's own trained default (the chatterbox family: exaggeration 0.5, temperature 0.6, guidance 1.0, ten refinement steps, no seed, the text-length budget). A value under a knob's [SpeakKnob.min](SpeakKnob/min.md) is refused before the model runs. |
| [SpeakReport](SpeakReport/index.md) | [common]
data class [SpeakReport](SpeakReport/index.md)(val latencyMs: [Float](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-float/index.html), val accelerator: [String](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-string/index.html), val canceled: [Boolean](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-boolean/index.html))
The outcome of one synthesis, delivered once through [SpeakListener.onDone](SpeakListener/onDone.md). |
| [SpeechClip](SpeechClip/index.md) | [common]
data class [SpeechClip](SpeechClip/index.md)(val pcm16: [ShortArray](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-short-array/index.html), val sampleRate: [Int](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-int/index.html), val durationMs: [Long](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-long/index.html))
A synthesized utterance: mono 16-bit samples at [sampleRate](SpeechClip/sampleRate.md) hertz (the chatterbox family speaks at 24 kHz), [durationMs](SpeechClip/durationMs.md) long. |
| [StageEvent](StageEvent/index.md) | [common]
data class [StageEvent](StageEvent/index.md)(val kind: [StageEventKind](StageEventKind/index.md), val stage: [String](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-string/index.html), val slot: [Int](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-int/index.html), val tStartNs: [Long](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-long/index.html), val tEndNs: [Long](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-long/index.html), val text: [String](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-string/index.html) = "", val isFinal: [Boolean](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-boolean/index.html) = false, val language: [String](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-string/index.html) = "", val mediaCount: [Int](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-int/index.html) = 0, val samples: [FloatArray](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-float-array/index.html)? = null, val sampleRate: [Int](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-int/index.html) = 0)
One event of a model's pipeline, reported to a [StageObserver](StageObserver/index.md) on the stage's own thread. [stage](StageEvent/stage.md) names the stage (`tokenize`, `prefill`, `decode` and `detokenize` on a text model, `vision` when a tower runs); [slot](StageEvent/slot.md) is the request's slot when the stage serves several at once, -1 otherwise; the times are the runtime's monotonic clock in nanoseconds. A payload field is empty, or null, when the kind carries none. |
| [StageEventKind](StageEventKind/index.md) | [common]
enum [StageEventKind](StageEventKind/index.md) : [Enum](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-enum/index.html)<[StageEventKind](StageEventKind/index.md)>
What one [StageEvent](StageEvent/index.md) reports. |
| [StageObserver](StageObserver/index.md) | [common]
fun interface [StageObserver](StageObserver/index.md)
Hears a model's pipeline stages ([GenerativeModel.setStageObserver](GenerativeModel/setStageObserver.md), or one call's through [GenerativeModel.generate](GenerativeModel/generate.md)). [onStage](StageObserver/onStage.md) runs on the stage's own thread, never the caller's, and must return at once: a slow observer holds that stage and nothing else. An exception it lets escape goes to its thread's uncaught-exception handler. |
| [SttModel](SttModel/index.md) | [common]
class [SttModel](SttModel/index.md) : [PreTrainedModel](PreTrainedModel/index.md)
A speech recognition model loaded from its files on disk, on the CPU. |
| [Task](Task/index.md) | [common]
object [Task](Task/index.md)
The hub task vocabulary: what [AutoModel.fromPretrained](AutoModel/fromPretrained.md) takes as a `task` to name the door explicitly, and what kindFor answers. |
| [TextGenerationPipeline](TextGenerationPipeline/index.md) | [common]
class [TextGenerationPipeline](TextGenerationPipeline/index.md) : [Pipeline](Pipeline/index.md)
Text generation over a [GenerativeModel](GenerativeModel/index.md): a prompt answers the generated text (the prompt and the reply together when returnFullText, the reply alone otherwise), a conversation answers its turns plus the assistant's. |
| [TextRankingPipeline](TextRankingPipeline/index.md) | [common]
class [TextRankingPipeline](TextRankingPipeline/index.md) : [Pipeline](Pipeline/index.md)
Reranking over a [RerankerModel](RerankerModel/index.md): a query and documents answer one relevance score per document, in the documents' order (`text-ranking` and its alias `reranking`); a higher score is a better match, and the caller sorts. |
| [TextToImagePipeline](TextToImagePipeline/index.md) | [common]
class [TextToImagePipeline](TextToImagePipeline/index.md) : [Pipeline](Pipeline/index.md)
Image generation over an [ImageGenerationModel](ImageGenerationModel/index.md): a prompt answers its pictures and the seed they were drawn from (`text-to-image`). |
| [TextToSpeechPipeline](TextToSpeechPipeline/index.md) | [common]
class [TextToSpeechPipeline](TextToSpeechPipeline/index.md) : [Pipeline](Pipeline/index.md)
Speech synthesis over a [TtsModel](TtsModel/index.md): a text answers one clip, the utterance's chunks joined. |
| [TextToVideoPipeline](TextToVideoPipeline/index.md) | [common]
class [TextToVideoPipeline](TextToVideoPipeline/index.md) : [Pipeline](Pipeline/index.md)
Video generation over a [VideoGenerationModel](VideoGenerationModel/index.md): a prompt answers its clip and the seed it was drawn from (`text-to-video`). |
| [ThinkingControl](ThinkingControl/index.md) | [common]
enum [ThinkingControl](ThinkingControl/index.md) : [Enum](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-enum/index.html)<[ThinkingControl](ThinkingControl/index.md)>
The reasoning controls a model admits, as its chat template declares them. |
| [ToolCall](ToolCall/index.md) | [common]
data class [ToolCall](ToolCall/index.md)(val id: [String](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-string/index.html), val name: [String](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-string/index.html), val argumentsJson: [String](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-string/index.html))
One call a reply made: the function's [name](ToolCall/name.md) and its arguments as JSON text ([argumentsJson](ToolCall/argumentsJson.md)); [id](ToolCall/id.md) names the call for its result ([ChatSession.addToolResult](ChatSession/addToolResult.md)), empty when the model's format wrote none (a [ChatSession](ChatSession/index.md) mints `call_`). |
| [ToolChoice](ToolChoice/index.md) | [common]
sealed class [ToolChoice](ToolChoice/index.md)
How the model may use the tools a request offers. |
| [ToolDef](ToolDef/index.md) | [common]
data class [ToolDef](ToolDef/index.md)(val name: [String](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-string/index.html), val description: [String](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-string/index.html) = "", val parametersJson: [String](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-string/index.html) = "{}")
One tool a request offers: the function's [name](ToolDef/name.md), its [description](ToolDef/description.md) and its parameters as a JSON schema document ([parametersJson](ToolDef/parametersJson.md), an object; `{}` takes no arguments). The chat template renders the offer, and the reply's calls are read against the names. |
| [TranscribeHandle](TranscribeHandle/index.md) | [common]
interface [TranscribeHandle](TranscribeHandle/index.md)
A running transcription. [cancel](TranscribeHandle/cancel.md) may be called from any thread: a clip transcribed in several windows stops at the next window boundary, and [TranscribeListener.onDone](TranscribeListener/onDone.md) arrives with [TranscribeReport.canceled](TranscribeReport/canceled.md) true. |
| [TranscribeListener](TranscribeListener/index.md) | [common]
interface [TranscribeListener](TranscribeListener/index.md)
The stream of one transcription. The calls arrive in order on the model's own worker thread: [onSegment](TranscribeListener/onSegment.md) for each transcribed span, then exactly one of [onDone](TranscribeListener/onDone.md) or [onError](TranscribeListener/onError.md). |
| [TranscribeOptions](TranscribeOptions/index.md) | [common]
data class [TranscribeOptions](TranscribeOptions/index.md)(val language: [String](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-string/index.html)? = null, val translate: [Boolean](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-boolean/index.html) = false, val targetLanguage: [String](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-string/index.html)? = null)
How a clip is transcribed: [language](TranscribeOptions/language.md) names the spoken language as an ISO code (`en`), null lets the model detect it; [targetLanguage](TranscribeOptions/targetLanguage.md) names the language of the text, null keeps the spoken one (a transcription) and a code asks a model that translates as it transcribes for the text in it; a model that cannot translate refuses a target that differs from the spoken language, naming what it lacks, and the request ends in [TranscribeListener.onError](TranscribeListener/onError.md). [translate](TranscribeOptions/translate.md) is the English shorthand: true reads as `targetLanguage = "en"` where no target is named. |
| [TranscribeReport](TranscribeReport/index.md) | [common]
data class [TranscribeReport](TranscribeReport/index.md)(val latencyMs: [Float](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-float/index.html), val accelerator: [String](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-string/index.html), val canceled: [Boolean](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-boolean/index.html), val language: [String](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-string/index.html)? = null)
The outcome of one transcription, delivered once through [TranscribeListener.onDone](TranscribeListener/onDone.md). |
| [Transcript](Transcript/index.md) | [common]
data class [Transcript](Transcript/index.md)(val text: [String](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-string/index.html), val report: [TranscribeReport](TranscribeReport/index.md))
A clip's transcript with the report of the request that made it ([AutomaticSpeechRecognitionPipeline.transcribe](AutomaticSpeechRecognitionPipeline/transcribe.md)). |
| [TranscriptSegment](TranscriptSegment/index.md) | [common]
data class [TranscriptSegment](TranscriptSegment/index.md)(val startMs: [Long](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-long/index.html), val endMs: [Long](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-long/index.html), val text: [String](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-string/index.html))
One transcribed span, with its position in the clip in milliseconds. |
| [TranslateHandle](TranslateHandle/index.md) | [common]
interface [TranslateHandle](TranslateHandle/index.md)
A running translation. [cancel](TranslateHandle/cancel.md) may be called from any thread; the model library runs the batch to its end, the translations are dropped when it returns, and [TranslateListener.onDone](TranslateListener/onDone.md) arrives with [TranslateReport.canceled](TranslateReport/canceled.md) true. |
| [TranslateListener](TranslateListener/index.md) | [common]
interface [TranslateListener](TranslateListener/index.md)
The outcome of one translation request. The calls arrive on the model's own worker thread: [onTranslations](TranslateListener/onTranslations.md) with one translation per input text, in order, then exactly one of [onDone](TranslateListener/onDone.md) or [onError](TranslateListener/onError.md). |
| [TranslateLoadOptions](TranslateLoadOptions/index.md) | [common]
data class [TranslateLoadOptions](TranslateLoadOptions/index.md)(val contextLength: [Int](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-int/index.html) = 0, val modelArgs: [Map](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin.collections/-map/index.html)<[String](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-string/index.html), [String](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-string/index.html)> = emptyMap(), val revision: [String](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-string/index.html) = "main", val cacheDir: [String](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-string/index.html)? = null, val token: [String](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-string/index.html)? = null, val offline: [Boolean](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-boolean/index.html) = false, val device: [String](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-string/index.html)? = null, val weights: [String](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-string/index.html)? = null, val dtype: [String](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-string/index.html)? = null) : [SourceOptions](SourceOptions/index.md)
The load-time choices a caller makes for a translation model. |
| [TranslateModel](TranslateModel/index.md) | [common]
class [TranslateModel](TranslateModel/index.md) : [PreTrainedModel](PreTrainedModel/index.md)
A translation model loaded from its files on disk, on the CPU: an encoder-decoder model that translates between the languages it names ([Capabilities.languages](Capabilities/languages.md)). A chat model that translates by instruction is a [GenerativeModel](GenerativeModel/index.md), not this. |
| [TranslateOptions](TranslateOptions/index.md) | [common]
data class [TranslateOptions](TranslateOptions/index.md)(val maxNewTokens: [Int](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-int/index.html) = 0)
How a translation runs. |
| [TranslateReport](TranslateReport/index.md) | [common]
data class [TranslateReport](TranslateReport/index.md)(val latencyMs: [Float](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-float/index.html), val accelerator: [String](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-string/index.html), val canceled: [Boolean](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-boolean/index.html))
The outcome of one translation request, delivered once through [TranslateListener.onDone](TranslateListener/onDone.md). |
| [Translation](Translation/index.md) | [common]
data class [Translation](Translation/index.md)(val text: [String](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-string/index.html), val sourceTokens: [Int](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-int/index.html), val generatedTokens: [Int](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-int/index.html), val stoppedAtBudget: [Boolean](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-boolean/index.html))
One translated text. [stoppedAtBudget](Translation/stoppedAtBudget.md) is true when the translation stopped at its token budget instead of the model's end token, so it may be incomplete. |
| [TranslationPipeline](TranslationPipeline/index.md) | [common]
class [TranslationPipeline](TranslationPipeline/index.md) : [Pipeline](Pipeline/index.md)
Translation over a [TranslateModel](TranslateModel/index.md): texts answer their translations, in order. |
| [TtsLoadOptions](TtsLoadOptions/index.md) | [common]
data class [TtsLoadOptions](TtsLoadOptions/index.md)(val contextLength: [Int](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-int/index.html) = 0, val scratchDir: <Error class: unknown class>? = null, val modelArgs: [Map](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin.collections/-map/index.html)<[String](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-string/index.html), [String](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-string/index.html)> = emptyMap(), val revision: [String](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-string/index.html) = "main", val cacheDir: [String](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-string/index.html)? = null, val token: [String](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-string/index.html)? = null, val offline: [Boolean](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-boolean/index.html) = false, val device: [String](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-string/index.html)? = null, val weights: [String](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-string/index.html)? = null, val dtype: [String](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-string/index.html)? = null) : [SourceOptions](SourceOptions/index.md)
The load-time choices a caller makes for a text-to-speech model. |
| [TtsModel](TtsModel/index.md) | [common]
class [TtsModel](TtsModel/index.md) : [PreTrainedModel](PreTrainedModel/index.md)
A text-to-speech model loaded from its files on disk, on the CPU. |
| [VideoGenerationHandle](VideoGenerationHandle/index.md) | [common]
interface [VideoGenerationHandle](VideoGenerationHandle/index.md)
A video-generation request in flight. |
| [VideoGenerationListener](VideoGenerationListener/index.md) | [common]
interface [VideoGenerationListener](VideoGenerationListener/index.md)
The calls a video-generation request makes on the model's worker thread, in order: [onVideo](VideoGenerationListener/onVideo.md) once with the clip and the seed the model drew, then [onDone](VideoGenerationListener/onDone.md); or [onError](VideoGenerationListener/onError.md) alone when the request fails. |
| [VideoGenerationModel](VideoGenerationModel/index.md) | [common]
class [VideoGenerationModel](VideoGenerationModel/index.md) : [PreTrainedModel](PreTrainedModel/index.md)
A video-generation model: a prompt in, a clip out (the `text-to-video` kind), its frames each as the RGBA pixels an Android bitmap takes, with the clip's sound where the family answers one. |
| [VideoGenerationOptions](VideoGenerationOptions/index.md) | [common]
data class [VideoGenerationOptions](VideoGenerationOptions/index.md)(val negativePrompt: [String](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-string/index.html)? = null, val width: [Int](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-int/index.html) = 0, val height: [Int](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-int/index.html) = 0, val frames: [Int](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-int/index.html) = 0, val fps: [Float](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-float/index.html) = 0.0f, val steps: [Int](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-int/index.html) = 0, val guidanceScale: [Float](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-float/index.html)? = null, val seed: [Long](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-long/index.html) = -1L)
How a clip is generated. A knob left at its zero keeps the model's own default: [width](VideoGenerationOptions/width.md) and [height](VideoGenerationOptions/height.md) in pixels and [frames](VideoGenerationOptions/frames.md) the frame count (the family's size and frame laws apply; see [Capabilities.defaultImageWidth](Capabilities/defaultImageWidth.md), [Capabilities.defaultImageHeight](Capabilities/defaultImageHeight.md) and [Capabilities.defaultFrames](Capabilities/defaultFrames.md)), [fps](VideoGenerationOptions/fps.md) the frame rate ([Capabilities.defaultFps](Capabilities/defaultFps.md)), [steps](VideoGenerationOptions/steps.md) the denoising steps, [guidanceScale](VideoGenerationOptions/guidanceScale.md) the prompt's pull (null keeps the model's), [seed](VideoGenerationOptions/seed.md) the draw (-1 lets the model pick one; the answer names it), [negativePrompt](VideoGenerationOptions/negativePrompt.md) what to steer away from (null for the family's own). |
| [VideoGenerationReport](VideoGenerationReport/index.md) | [common]
data class [VideoGenerationReport](VideoGenerationReport/index.md)(val latencyMs: [Float](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-float/index.html), val accelerator: [String](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-string/index.html), val canceled: [Boolean](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-boolean/index.html))
The facts of one video-generation request. |
| [VisionHandle](VisionHandle/index.md) | [common]
interface [VisionHandle](VisionHandle/index.md)
A running vision request. [cancel](VisionHandle/cancel.md) may be called from any thread. The model library runs a request to its end, so a cancel does not shorten the wait: the result is dropped when the model returns, no result callback is made, and [VisionListener.onDone](VisionListener/onDone.md) arrives with [VisionReport.canceled](VisionReport/canceled.md) true. |
| [VisionListener](VisionListener/index.md) | [common]
interface [VisionListener](VisionListener/index.md)
The outcome of one vision request. The calls arrive on the model's own worker thread: the one result callback the request's task answers with (a detection request calls [onDetections](VisionListener/onDetections.md), a depth request [onDepth](VisionListener/onDepth.md), a text reading [onText](VisionListener/onText.md), an embedding [onEmbedding](VisionListener/onEmbedding.md), a classification [onScores](VisionListener/onScores.md)), then exactly one of [onDone](VisionListener/onDone.md) or [onError](VisionListener/onError.md). The result callbacks default to nothing, so a listener implements the one its request needs. |
| [VisionLoadOptions](VisionLoadOptions/index.md) | [common]
data class [VisionLoadOptions](VisionLoadOptions/index.md)(val modelArgs: [Map](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin.collections/-map/index.html)<[String](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-string/index.html), [String](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-string/index.html)> = emptyMap(), val revision: [String](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-string/index.html) = "main", val cacheDir: [String](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-string/index.html)? = null, val token: [String](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-string/index.html)? = null, val offline: [Boolean](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-boolean/index.html) = false, val device: [String](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-string/index.html)? = null, val weights: [String](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-string/index.html)? = null, val dtype: [String](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-string/index.html)? = null) : [SourceOptions](SourceOptions/index.md)
The load-time choices a caller makes for a vision model. |
| [VisionModel](VisionModel/index.md) | [common]
class [VisionModel](VisionModel/index.md) : [PreTrainedModel](PreTrainedModel/index.md)
A vision model loaded from its files on disk, on the CPU: a detector, a depth estimator, a text reader, or an embedding model (a model that embeds text and pictures in one space also classifies against labels). |
| [VisionReport](VisionReport/index.md) | [common]
data class [VisionReport](VisionReport/index.md)(val latencyMs: [Float](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-float/index.html), val accelerator: [String](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-string/index.html), val canceled: [Boolean](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-boolean/index.html))
The outcome of one vision request, delivered once through [VisionListener.onDone](VisionListener/onDone.md). |
| [VisionTask](VisionTask/index.md) | [common]
object [VisionTask](VisionTask/index.md)
The tasks a vision model may serve, the spellings [Capabilities.visionTasks](Capabilities/visionTasks.md) carries. |
| [VoiceActivityDetectionPipeline](VoiceActivityDetectionPipeline/index.md) | [common]
class [VoiceActivityDetectionPipeline](VoiceActivityDetectionPipeline/index.md) : [Pipeline](Pipeline/index.md)
Speaker diarization over a [DiarizationModel](DiarizationModel/index.md): a clip answers who spoke when, in time order (`voice-activity-detection`, the hub's name for the kind). |
| [VoiceReference](VoiceReference/index.md) | [common]
sealed interface [VoiceReference](VoiceReference/index.md)
The voice a text-to-speech request clones: a clip of the speaker, mono, a few seconds long (a voice-cloning family reads about the first ten seconds for the generator and the whole clip, silence trimmed, for the speaker embedding). Any sample rate; the library resamples. |
| [ZeroShotImageClassificationPipeline](ZeroShotImageClassificationPipeline/index.md) | [common]
class [ZeroShotImageClassificationPipeline](ZeroShotImageClassificationPipeline/index.md) : [Pipeline](Pipeline/index.md)
Zero-shot image classification over a dual-tower [VisionModel](VisionModel/index.md): a picture and the candidate labels answer each label's score. |
## Functions
| Name | Summary |
|---|---|
| [load](load.md) | [android]
@[JvmOverloads](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin.jvm/-jvm-overloads/index.html)
fun [Modelverse](Modelverse/index.md).[load](load.md)(context: Context, license: [String](https://kotlinlang.org/api/core/kotlin-stdlib/kotlin/-string/index.html)? = null)
The Android entry: pin the runtime's cache root under the app's cache directory and place [license](load.md) through [ClikaRtAndroid.load](../io.clika.runtime/ClikaRtAndroid/load.md), then [Modelverse.load](Modelverse/load.md). One call readies the binding on Android; a second call is a no-op. [license](load.md) is the runtime's license credential (the `CLIKA1-...` text, or the path of a file holding it); `null` leaves the process environment as it is. |