@pr0gramm/fluester/dist/script/download.js

Node.js bindings for OpenAI's Whisper. Optimized for CPU.

#!/usr/bin/env node
// Javascript layer for using the whisper.cpp built-in model downloader scripts
// npx @pr0gramm/fluester download
import { createInterface } from "node:readline/promises";
import { canExecute, execute } from "../execute.js";
import { nodeModulesModelPath } from "../interop.js";
import { defaultModel, modelList, modelStats, } from "../model.js";
async function determineModel() {
    // ["/usr/bin/node", "../.bin/download", <model name>]
    const parameterModel = process.argv[2];
    if (parameterModel) {
        if (modelList.includes(parameterModel)) {
            return parameterModel;
        }
        console.error(`Invalid model name: "${parameterModel}"`);
        console.log(`Valid model names are: ${modelList.join(", ")}`);
        process.exit(-1);
    }
    const envModel = process.env.WHISPER_MODEL;
    if (!!envModel && modelList.includes(envModel)) {
        return envModel;
    }
    const rl = createInterface({
        input: process.stdin,
        output: process.stdout,
    });
    console.log("Which model should be downloaded?");
    console.log("You can skip this question by:");
    console.log("- passing the model as an env var: WHISPER_MODEL=tiny or");
    console.log("- as a parameter: `npx @pr0gramm/fluester download tiny`");
    console.table(modelStats);
    const answer = await rl.question(`\nEnter model name (e.g. 'base.en') or "cancel" to exit\n(default: "${defaultModel}"): `);
    const answerNormalized = answer.trim().toLowerCase();
    if (answerNormalized === "") {
        return defaultModel;
    }
    if (answerNormalized === "cancel") {
        process.exit(-1);
    }
    if (modelList.includes(answer)) {
        return answer;
    }
    console.error(`Invalid model name: "${answer}"`);
    console.log(`Valid model names are: ${modelList.join(", ")}`);
    process.exit(-1);
}
try {
    process.chdir(nodeModulesModelPath);
    // ensure running in correct path
    if (!(await canExecute("./download-ggml-model.sh"))) {
        throw new Error("Cannot run downloader. Maybe the path is incorrect or the current working directory is not correct.");
    }
    const modelName = await determineModel();
    const scriptPath = process.platform === "win32"
        ? "download-ggml-model.cmd"
        : "./download-ggml-model.sh";
    await execute(scriptPath, [modelName], true);
    console.log(`Model "${modelName}" downloaded successfully.`);
    process.exit(0);
}
catch (error) {
    console.log("Error while downloading model:");
    console.log(error);
    throw error;
}