Starlings.Models.Kev
0.5.0
dotnet add package Starlings.Models.Kev --version 0.5.0
NuGet\Install-Package Starlings.Models.Kev -Version 0.5.0
<PackageReference Include="Starlings.Models.Kev" Version="0.5.0" />
<PackageVersion Include="Starlings.Models.Kev" Version="0.5.0" />
<PackageReference Include="Starlings.Models.Kev" />
paket add Starlings.Models.Kev --version 0.5.0
#r "nuget: Starlings.Models.Kev, 0.5.0"
#:package Starlings.Models.Kev@0.5.0
#addin nuget:?package=Starlings.Models.Kev&version=0.5.0
#tool nuget:?package=Starlings.Models.Kev&version=0.5.0
Starlings.Models.Kev
The model definitions of the Kev 1.0 decision models (jaredpalmer/kev-0.8b and jaredpalmer/kev-4b), without a runtime: requests, Kev's request rendering and escaping, the encoder that turns a request into token-id rows, the checkpoint identities, and the result types. No ONNX Runtime dependency (only the owned tokenizer Starlings.Tokenizers, and Starlings.Acquisition for the checkpoint locks). To run the models on ONNX Runtime, reference Starlings.Models.Kev.Onnx, which brings this package. Experimental; the API may still change.
KevRequest(a state and namedKevQuestions:Noul,Choice,Score;ParseSystemOnefor a System One request body) andKevContent(text, or JSON rendered as Kev does).KevText: Kev's rendering of JSON values (Pythonstr()of numbers and booleans), option text, andEscapeSpecialTokens, which rewrites<|…|>in user text so it can never become a delimiter.KevRequest.ToRecord(): the rendered record Kev's encoder reads (KevRecord,KevRecordQuestion).KevEncoder:FromFile(tokenizerPath)checks Kev's delimiter ids in theqwen2-bpe-kevtokenizer;Encode(record)returns the state ids and one row per question (KevEncoding,KevRow: ids, positions continuing after the state, the<decide>index and each option's</opt>index), or throwsKevInputTooLongException(state or row over 8,192 tokens; withmaxBranchTokens, a branch over that budget). Every over-long question is found in one pass: the exception lists them all (OverLongQuestions, each aKevOverLongQuestionwithIndex,Id,Limit,TokenCount,MaxTokens), or, withdecline: true, the encoding leaves them out and lists them inKevEncoding.Declined.EncodeStateandEncodeRowsare the two halves, for runtimes that reuse one state across question sets.KevModelProfile(Kev08B,Kev4B): repositories, revisions, base checkpoints, head and adapter digests, dimensions and calibrated temperature.CheckpointLockandBaseCheckpointLockare the asset locks of the Hub checkpoints themselves (Kev's adapter, head and tokenizer files; the frozen base's configuration, safetensors weights and tokenizer files), the inputs of the merge that builds the ONNX bundle. The package also carries them underlocks/(kev-0.8b,kev-4b,qwen3.5-0.8b-base,qwen3.5-4b-base, each.checkpoint.lock.json; JSON lock schema version 2, read byModelAssetLock.Load). Nothing downloads unless you pass a lock toModelAssetStore.DownloadAsync. The locks leave out each repository'sREADME.mdandLICENSE(Kev's repositories have noLICENSE; the Qwen3.5 base checkpoints are Apache-2.0): a directory filled fromBaseCheckpointLockholds the weights without the license text, so fetchLICENSEfrom the base repository at the same revision if you redistribute them.KevDecision(withDeclined) andKevAnswer(withTemperatureandIsCalibrated): the result types the runtime returns.
using Starlings.Models.Kev;
var request = new KevRequest("The order arrived two weeks late and the box was crushed.",
[new("sentiment", KevQuestion.Choice("How does the customer feel?", "positive", "neutral", "negative"))]);
KevEncoding rows = KevEncoder.FromFile(tokenizerPath).Encode(request.ToRecord());
// rows.StateIds is the prefill; each rows.Rows[k] runs on the state's cache and ends at its Decide index.
Maturity
Maturity is per profile, generated from the registry manifests by model-pipeline catalog (do not edit this table by hand). Gates, evidence and capabilities per profile: the model catalog.
| Profile | Maturity | Platforms |
|---|---|---|
kev-0.8b/onnx-fp32-cpu |
experimental | linux-x64 cpu: experimental; linux-x64 cuda: experimental; windows-x64 cpu: untested |
kev-0.8b/onnx-fp32-cuda |
experimental | linux-x64 cuda: experimental; windows-x64 cuda: untested |
kev-4b/onnx-fp32-cpu |
experimental | linux-x64 cpu: experimental; linux-x64 cuda: experimental; windows-x64 cpu: untested |
kev-4b/onnx-fp32-cuda |
experimental | linux-x64 cuda: experimental; windows-x64 cuda: untested |
Attribution and license
The rendering, escaping and encoding reimplement Kev's kev/api.py and kev/model.py at 6b719c3c3f367295f6ef336f4f751cf5ff970abc (tag kev-1.0) by Jared Palmer, Apache-2.0; no upstream file is copied. This package's code is MIT OR Apache-2.0.
| Product | Versions Compatible and additional computed target framework versions. |
|---|---|
| .NET | net10.0 is compatible. net10.0-android was computed. net10.0-browser was computed. net10.0-ios was computed. net10.0-maccatalyst was computed. net10.0-macos was computed. net10.0-tvos was computed. net10.0-windows was computed. |
-
net10.0
- Starlings.Acquisition (= 0.5.0)
- Starlings.Tokenizers (= 0.3.4)
NuGet packages (1)
Showing the top 1 NuGet packages that depend on Starlings.Models.Kev:
| Package | Downloads |
|---|---|
|
Starlings.Models.Kev.Onnx
Kev decision models (jaredpalmer/kev-0.8b and jaredpalmer/kev-4b, Kev 1.0) on ONNX Runtime: one state and typed questions (noul, choice, score) in, a calibrated probability per option out, in one prefill pass without generating text. KevDecisionModel over the model definitions Starlings.Models.Kev (request rendering, delimiter escaping and token ids from the owned tokenizer Starlings.Tokenizers profile qwen2-bpe-kev), the Qwen3.5 backbone as an fp32 ONNX graph (the LoRA merged; ONNX Runtime GenAI builder without the LM head), the state run once and every question as its own row on the cached state, and Kev's pointer head. Ships the asset locks, not the weights. Profiles kev-0.8b/onnx-fp32-cpu and kev-4b/onnx-fp32-cpu, and onnx-fp32-cuda on the same bundles (TF32 off). The application adds the ONNX Runtime native package. Maturity is per model profile (Stable, Preview or Experimental): see the README, generated from the model registry. |
GitHub repositories
This package is not used by any popular GitHub repositories.