@exodus/qwen3-model-js
v0.1.0
Published
Pinned Qwen3 1.7B Q8_0 GGUF: artifact metadata, SHA256 verification, and streaming downloader. Declares node-llama-cpp as a peer so consumers pick a compatible runtime.
Maintainers
Keywords
Readme
@exodus/qwen3-model-js
Pinned Qwen3 1.7B Q8_0 GGUF: artifact metadata, SHA256 verification, and streaming downloader.
Example
import { ensurePinnedModel, getPinnedModel } from '@exodus/qwen3-model-js'
const { filename } = getPinnedModel()
await ensurePinnedModel(`./models/${filename}`, {
onProgress: ({ fraction }) => console.log(`${(fraction * 100).toFixed(1)}%`),
})Exports
getPinnedModel()— returns the frozen artifact record (repo,filename,sha256,sizeBytes,downloadUrl,license).verifyPinnedModelFile(path)— streams the file and throws on size or SHA256 mismatch.ensurePinnedModel(path, opts?)— verifies if present, otherwise downloads to a temp file, verifies, and atomically renames into place. Returns{ downloaded }. Opts:onProgress,signal,timeoutMs(default 30 min,0disables).ModelMissingError— thrown when the file isn't on disk.ModelHashMismatchError— thrown on size or SHA256 mismatch; carriespath,expected,actual.
