ModelPackage
ModelPackage =
object
Provider-neutral descriptor for a developer-supplied/custom local model made available to an IndeRun local-model provider family (for example, the ONNX Runtime family). It describes model identity, format, task support, source, files, integrity, licensing, and resource expectations. It is bootstrap/configuration metadata resolved before execution; it is not part of the public TaskRequest/TaskResult surface, and it must not carry raw secrets.
Indexable
Section titled “Indexable”[
property:string]:unknown
Properties
Section titled “Properties”files?
Section titled “files?”
optionalfiles?:Files
Files that make up the model package, expressed as source-relative names/paths. The provider adapter and model source resolve these to concrete bytes per platform.
format
Section titled “format”format:
Format
Model packaging format the target runtime family must understand. ‘onnx’ is a plain ONNX graph, ‘ort’ is an ONNX Runtime optimized/mobile format, ‘genai’ is an ONNX Runtime GenAI model package, ‘coreai’ is an Apple Core AI .aimodel asset (iOS/macOS 27+).
id:
string
Stable application-scoped identifier for the model package.
integrity?
Section titled “integrity?”
optionalintegrity?:Integrity
Optional integrity metadata used to validate resolved files before load.
license?
Section titled “license?”
optionallicense?:License
Optional license/source metadata for the model, for developer transparency. Free-form.
limits?
Section titled “limits?”
optionallimits?:Limits
Optional known resource expectations, used by capability checks to reject on constrained devices before load.
runtime?
Section titled “runtime?”
optionalruntime?:Runtime
Optional runtime compatibility expectations. Fields are advisory hints for capability checks; the provider adapter owns exact enforcement.
source?
Section titled “source?”
optionalsource?:Source
Where the model files are obtained from. Availability of each source type is platform-dependent; see the ONNX Runtime provider-family specification for the per-platform support matrix.
speech?
Section titled “speech?”
optionalspeech?:SpeechModelPackage
What a text_to_speech model package adds: the pipeline that turns text into PCM, the graphs it runs, its text frontend and the voices it offers. Required when ‘tasks’ contains ‘text_to_speech’.
tasks?
Section titled “tasks?”
optionaltasks?:string[]
IndeRun task kinds this model package can serve (for example ‘text_to_text’). Used by dynamic capability checks and route matching.
variant?
Section titled “variant?”
optionalvariant?:string
Optional concrete artifact variant of this logical model (for example ‘onnx-int8’, ‘coreai-fp16’). ‘id’ names the model, ‘variant’ the bytes.
version?
Section titled “version?”
optionalversion?:string
Optional application-defined version for the model package, used for cache invalidation and compatibility checks.
IndeRun — open-source AI execution for cross-platform apps. Built and maintained by Independo GmbH (opens in new tab), supported by netidee (opens in new tab).
MIT-licensed SDK · No SDK usage telemetry.
© 2025–2026 Independo GmbH