What causes "An unknown error occurred" in can1357/oh-my-pi?

Log the full output object at the callsite to capture stopReason and any partial content before the throw

can1357/oh-my-pi · error · AIError.ProviderResponseError

An unknown error occurred

Error message

An unknown error occurred

What it means

streamGoogleGeminiCli finished streaming a response whose output.stopReason was "aborted" or "error", and no provider error message was attached, so this generic fallback text is thrown as AIError.ProviderResponseError. It means the Gemini CLI endpoint signalled a failed/terminated response without explaining why.

Source

Thrown at packages/ai/src/providers/google-gemini-cli.ts:1104

					break;
				} catch (error) {
					const status = extractHttpStatusFromError(error);
					if (
						!isLastEndpoint &&
						!started &&
						(AIError.isTransientStatus(status) ||
							(status === undefined &&
								!(error instanceof AIError.ProviderResponseError && error.kind === "output") &&
								AIError.retriable(AIError.classify(error))))
					) {
						continue;
					}
					throw error;
				}
			}

			if (output.stopReason === "aborted" || output.stopReason === "error") {
				throw new AIError.ProviderResponseError(output.errorMessage ?? "An unknown error occurred", {
					provider: model.provider,
					kind: "output",
				});
			}

			output.duration = performance.now() - startTime;
			if (firstTokenTime) output.ttft = firstTokenTime - startTime;
			stream.push({ type: "done", reason: output.stopReason, message: output });
			stream.end();
		} catch (error) {
			const result = await AIError.finalize(error, { api: model.api, signal: options?.signal, rawRequestDump });
			output.stopReason = result.stopReason;
			output.errorStatus = result.status;
			output.errorId = result.id;
			output.errorMessage = result.message;
			output.duration = performance.now() - startTime;
			if (firstTokenTime) output.ttft = firstTokenTime - startTime;
			stream.push({ type: "error", reason: output.stopReason, error: output });

View on GitHub (pinned to 9690622007)

Solutions

Log the full output object at the callsite to capture stopReason and any partial content before the throw
Check Gemini CLI / Code Assist service status and auth token validity
Retry the request — transient backend failures often produce empty error messages
If aborts are user-driven, check options.signal before treating this as a provider failure

Example fix

// before
throw new AIError.ProviderResponseError(output.errorMessage ?? "An unknown error occurred", { provider: model.provider, kind: "output" });
// after
if (!output.errorMessage) {
  logger.error("gemini-cli opaque error", { stopReason: output.stopReason, provider: model.provider });
}
throw new AIError.ProviderResponseError(output.errorMessage ?? `Gemini CLI stream ended with stopReason=${output.stopReason}`, { provider: model.provider, kind: "output" });

Defensive patterns

Strategy: try-catch

Validate before calling

null

Type guard

null

Try / catch

try {
  await streamGoogleGeminiCli(model, params, { signal });
} catch (err) {
  if (err instanceof AIError.AbortError) throw err; // rethrow user aborts
  if (err instanceof AIError.ProviderResponseError && err.context?.kind === "output") {
    logger.warn("gemini-cli ended without error message", { message: err.message });
    return retryWithBackoff();
  }
  throw err;
}

Prevention

Always pass an AbortSignal and check signal.aborted before interpreting failures as provider errors
Attach an onError/message handler that captures error chunks into output.errorMessage so the fallback text is less likely
Add retry with backoff around Gemini CLI streams
Monitor Google Code Assist service status

When it happens

Trigger: The aggregated stream output has stopReason "error" or "aborted" but output.errorMessage is null/undefined — e.g. the Gemini CLI backend emitted a terminal chunk with an error stop reason and no message.

Common situations: Gemini Code Assist / CLI backend outages returning opaque error terminations; quota or safety terminations that omit a message; aborted generations surfaced as stop reasons rather than exceptions.

Related errors

AI-assisted analysis of can1357/oh-my-pi@9690622007 (2026-08-31). Data as JSON: /api/errors/dab1acc8e8028f2d. Report an issue: GitHub.