can1357/oh-my-pi · error · Error

Unable to resolve ${SHERPA_PACKAGE} in compiled runtime at $

Error message

Unable to resolve ${SHERPA_PACKAGE} in compiled runtime at ${nodeModules}

What it means

In a compiled binary, the native sherpa-onnx addon is installed into a side runtime directory (node_modules under the STT runtime dir) and resolved via resolveRuntimeModule before being loaded with createRequire. If, after the ensureRuntimeInstalled step, the package entry still cannot be found at that node_modules path, the worker throws — the runtime install did not produce a loadable package.

Source

Thrown at packages/coding-agent/src/stt/asr-worker.ts:164

 * Memoized so the runtime loads once per process.
 */
function loadSherpaRuntime(transport: SttTransport, requestId: string, modelKey: SttModelKey): Promise<SherpaRuntime> {
	return sherpaRuntime.load(async () => {
		if (!isCompiledBinary()) return loadSourceSherpaRuntime(import.meta.url);
		const runtimeDir = await ensureRuntimeInstalled({
			runtimeDir: getSherpaRuntimeDir(),
			install: { dependencies: { [SHERPA_PACKAGE]: getSherpaVersionSpec() } },
			probePackage: SHERPA_PACKAGE,
			onPhase: phase =>
				transport.send({
					type: "progress",
					id: requestId,
					event: { modelKey, status: phase, name: `${SHERPA_PACKAGE}@${getSherpaVersionSpec()}` },
				}),
		});
		const nodeModules = path.join(runtimeDir, "node_modules");
		const entry = resolveRuntimeModule(nodeModules, SHERPA_PACKAGE);
		if (!entry) throw new Error(`Unable to resolve ${SHERPA_PACKAGE} in compiled runtime at ${nodeModules}`);
		return createRequire(entry)(entry) as SherpaRuntime;
	});
}

async function loadPipelineOnDevice(
	transformers: TransformersRuntime,
	spec: TransformersSttModelSpec,
	modelKey: SttModelKey,
	transport: SttTransport,
	requestId: string,
	device: TinyModelDevice,
): Promise<AutomaticSpeechRecognitionPipeline> {
	return transformers.pipeline(ASR_TASK, spec.repo, {
		device,
		dtype: sttModelDtypeOverride ?? spec.dtype,
		progress_callback: info => sendProgress(transport, requestId, modelKey, info),
	});
}

View on GitHub (pinned to 9690622007)

Solutions

1. Delete the stale STT runtime directory under the tiny-models cache (…/stt-runtime/sherpa-*) and retry so ensureRuntimeInstalled reinstalls cleanly.
2. Check disk space and directory permissions for the cache dir.
3. Ensure your antivirus/security software is not removing .node native modules.
4. If it persists, verify the installed sherpa-onnx version actually ships a resolvable entry for your platform.

Example fix

// before
rm -rf ~/.cache/omp/tiny-models/stt-runtime/sherpa-*   # then retry the model load
// after
ensureRuntimeInstalled(...) succeeds and resolveRuntimeModule finds the entry

const entry = resolveRuntimeModule(path.join(runtimeDir, "node_modules"), "sherpa-onnx-node");
if (!entry) await ensureRuntimeInstalled({ runtimeDir, install: { dependencies: { "sherpa-onnx-node": spec } }, probePackage: "sherpa-onnx-node" });

try { const runtime = await loadSherpaRuntime(transport, id, modelKey); } catch (err) { await rm(runtimeDir, { recursive: true, force: true }); /* retry once to reinstall the side runtime */ }

When it happens

Trigger: Loading any sherpa STT model in a compiled binary when resolveRuntimeModule(runtimeDir/node_modules, "sherpa-onnx-node") returns null — install failed silently, was interrupted, or the runtime dir is corrupt/partial.

Common situations: Disk-full or permission errors during the side runtime install, an interrupted first run leaving a partial runtime dir, antivirus quarantining native .node binaries, or a version-spec/install mismatch pointing at an incomplete tree.

Related errors

• No stt model devices configured
• Unknown stt model: ${modelKey}
• Failed to download speech model (${spec.repo})${detail}
• Speech model download finished without required files (${spe
• Cannot find module onnxruntime-node beside ${fastembedEntry}

AI-assisted analysis of can1357/oh-my-pi@9690622007 (2026-08-31). Data as JSON: /api/errors/5255a96ed8c88923. Report an issue: GitHub.