SKaiNET-developers

@skainet-developers source repo

12 skills
0 followers
8 hours ago last updated

▌
Kmp · skainet-developers bundle
Use ONLY when configuring KMP targets, source-set hierarchies, or `expect`/`actual` placement in a module INSIDE the SKaiNET repository. Trigger tokens include `kotlin { ... }`, `iosArm64()`, `commonMain`, `jvmMain`, `expect fun`, `actual fun`, `androidNative`, `wasmJs`, source-set dependency edits in `SKaiNET/skainet-*/build.gradle.kts`. Do NOT fire on KMP setup in a CONSUMER project (an app that depends on SKaiNET) — those concerns are simpler and live in `skainet-consumer-setup`.

0
installs
▌
Kotlin · skainet-developers bundle
kotlin

0
installs
▌
Skainet Nn Dsl · skainet-developers bundle
Use when defining a neural network architecture with the SKaiNET public DSL — `sequential<T, V> { ... }` for linear stacks, `dag { ... }` for graphs with branching / skip connections. Applies in both consumer apps and inside the SKaiNET repo. Trigger tokens include `sequential<`, `dag {`, `dagModule`, layer names (`input`, `dense`, `flatten`, `activation`, `softmax`, `batchNorm`, `groupNorm`, `layerNorm`, `conv1d`, `conv2d`, `conv3d`, `maxPool2d`, `avgPool2d`, `upsample2d`), DAG nodes (`parameter`, `constant`, `matmul`, `relu`, `output`). Do NOT fire on tensor *creation* (that's `skainet-data-dsl`), on assertions (use the contributor `skainet-testing` skill in-repo), or on running a model end-to-end (that's `skainet-inference`).

0
installs
▌
Skainet Data Dsl · skainet-developers bundle
Use when constructing tensors, slicing them, or building transform pipelines using the SKaiNET public DSL — applies in both consumer apps and inside the SKaiNET repo. Trigger tokens include `tensor(`, `tensor {`, `data<`, `pipeline<`, `sliceView`, `segment {`, `rescale`, `normalize`, `unsqueeze` (in pipeline context), `FP32::class`, `FP16::class`, `Int8::class`, `Int32::class`, `Ternary::class`, `randn(`, `uniform(`. Do NOT fire on tensor *assertions* (in-repo tests go to the contributor `skainet-testing` skill) or on neural-network builders (`sequential` / `dag` go to `skainet-nn-dsl`).

0
installs
▌
Skainet Inference · skainet-developers bundle
Use when running a SKaiNET model end-to-end — picking the right `ExecutionContext`, calling `model.forward(x, ctx)`, batching inputs, configuring the eval/train phase, applying TurboQuant for KV-cache compression. Trigger tokens include `DirectCpuExecutionContext.create()`, `DefaultNeuralNetworkExecutionContext`, `model.forward(`, `module.forward(`, `Phase.EVAL`, `Phase.TRAIN`, `ctx.inTraining`, `TurboQuantPolar`, `TurboQuantConfig`, `TurboQuantCodec`. Do NOT fire on tensor construction (`skainet-data-dsl`), model architecture definition (`skainet-nn-dsl`), or model file loading (`skainet-model-loading`).

0
installs
▌
Skainet Testing · skainet-developers bundle
skainet-testing

0
installs
▌
Skainet Java Consumer · skainet-developers bundle
Use when calling SKaiNET from a pure-Java app (Spring Boot, Android Java, Quarkus, plain main()) — adding the JVM-only Maven coordinates, calling `SKaiNET.context()` and `TensorJavaOps.*`, using `StableHloConverterFactory` / `TokenizerFactory`. Trigger tokens include `SKaiNET.context()`, `SKaiNET.tensor(`, `TensorJavaOps.`, `StableHloConverterFactory`, `TokenizerFactory.fromGguf`, `package sk.ainet.java` import in a `.java` file, "from Java", "Java consumer". Do NOT fire when designing the Java facade INSIDE SKaiNET (that's the contributor `skainet-java-interop` skill) or when the consumer is in Kotlin (that's `skainet-inference`).

0
installs
▌
Skainet Model Loading · skainet-developers bundle
Use when loading a pre-trained model into a SKaiNET-based app — GGUF (`GGUFModelReader`), ONNX (`OnnxLoader`), SafeTensors (`SafeTensorsParametersLoader`), or JSON. Trigger tokens include `GGUFModelReader`, `OnnxLoader.fromModelSource`, `SafeTensorsParametersLoader`, `loadGGUF`, `loadOnnx`, `.safetensors`, `.gguf`, `.onnx`, `ParametersLoader`. Do NOT fire on `model.forward(x, ctx)` (that's `skainet-inference`), defining model architecture in code (`skainet-nn-dsl`), or constructing tensors from raw arrays (`skainet-data-dsl`).

0
installs
▌
Gradle Multimodule · skainet-developers bundle
Use ONLY when editing build scripts INSIDE the SKaiNET repository — `SKaiNET/build.gradle.kts`, `SKaiNET/settings.gradle.kts`, `SKaiNET/gradle/libs.versions.toml`, anything under `SKaiNET/build-logic/`, or adding/renaming/removing a `skainet-*` module within SKaiNET. Enforces version-catalog-only references, convention-plugin reuse, BOM registration, binary-compatibility-validator, vanniktech maven-publish, kover. Do NOT fire on a CONSUMER project's build script that just depends on `sk.ainet:skainet-bom` — that's the `skainet-consumer-setup` skill.

0
installs
▌
Skainet Consumer Setup · skainet-developers bundle
Use when adding SKaiNET as a library dependency to your Gradle project (KMP, JVM, or Android) — adding `sk.ainet:skainet-bom` to the version catalog, picking which `sk.ainet.core:skainet-*` artifacts the project needs (inference vs. training vs. format-specific I/O), or troubleshooting "I added SKaiNET but X doesn't resolve". Trigger tokens include `sk.ainet:skainet-bom`, `sk.ainet.core:skainet-`, `platform("sk.ainet:skainet-bom")`, `implementation(libs.skainet.*)`, "add SKaiNET to my project". Do NOT fire when editing build scripts INSIDE the SKaiNET repo itself (that's the contributor `gradle-multimodule` skill).

0
installs
▌
Skainet Java Interop · skainet-developers bundle
Use ONLY when designing or editing the Java-facing facade INSIDE the SKaiNET repository — files under `SKaiNET/skainet-*/src/jvmMain/kotlin/sk/ainet/java/`. Trigger tokens include `@file:JvmName`, `@JvmStatic`, `@JvmOverloads`, `package sk.ainet.java`, `object SKaiNET`, `object TensorJavaOps`. Encodes the contributor-side rules for the convention. Do NOT fire when the user is CALLING the Java facade from a Java app — that's the consumer-side `skainet-java-consumer` skill.

0
installs
▌
Skainet Android Integration · skainet-developers bundle
Use when integrating SKaiNET into an Android app — loading model files from `assets/`, picking the CPU backend ABI (ARM, ARM64, NDK vendor-native), wiring inference to `viewModelScope` / `lifecycleScope`, handling memory pressure, and managing the JVM/Native bridge. Trigger tokens include `AssetManager`, `Context.assets.open(`, `androidNativeArm64`, `viewModelScope`, `lifecycleScope`, `OnTrimMemory`, `Application` in a SKaiNET-using app. Do NOT fire when the user is editing the SKaiNET repo's own Android source sets (the contributor `kmp` skill covers that).

0
installs

12 published skills