can1357/oh-my-pi · error · Error

Checker discovery failed: ${result.result.error}

Error message

Checker discovery failed: ${result.result.error}

What it means

The cleanse checker-discovery step runs an LLM prompt with a DISCOVERY_SCHEMA and, if the agent result carries result.error (the sub-agent failed — API error, aborted, invalid response), wraps it in 'Checker discovery failed: ...'. It means discovery could not produce checker specs, not that checkers themselves are broken.

Source

Thrown at packages/coding-agent/src/cleanse/agent.ts:143

	return {
		model: modelDisplay,
		sessionFile,
		async discoverCheckers(request: string, signal?: AbortSignal): Promise<CustomCleanseCheckerSpec[]> {
			sessionManager.appendCustomEntry("cleanse_discovery", { request });
			const result = await runStructuredSubagent({
				session: toolSession,
				invocationKind: "task",
				assignment: prompt.render(discoveryPrompt, { request }),
				agent: "task",
				model: modelSelector,
				outputSchema: DISCOVERY_SCHEMA,
				identity: { label: "CleanseDiscovery" },
				enableLsp: true,
				enableIrc: false,
				signal,
			});
			if (result.result.error) throw new Error(`Checker discovery failed: ${result.result.error}`);
			return parseDiscoverySpecs(result.result.structuredOutput?.data);
		},
		async dispatchWorker(
			assignment: CleanseAssignment,
			context: {
				worker: number;
				peers: readonly CleanseAssignment[];
				checkers: readonly CleanseCheckerDescriptor[];
			},
			signal?: AbortSignal,
		): Promise<CleanseAgentOutcome> {
			sessionManager.appendCustomEntry("cleanse_dispatch", {
				worker: context.worker,
				weight: assignment.weight,
				files: assignment.groups.map(group => group.file ?? "<project>"),
			});
			const name = `CleanseA${context.worker}`;
			options.hooks?.onStart?.(name, assignment);

View on GitHub (pinned to 9690622007)

Solutions

  1. Read the embedded result.error text to identify the root cause (auth vs rate limit vs abort)
  2. Re-run cleanse once transient errors/rate limits clear; check provider status if persistent
  3. Verify the configured cleanse model's provider auth is valid and the model is available
  4. Increase the discovery timeout / avoid cancelling so the signal doesn't abort mid-run
Defensive patterns

Strategy: retry

Validate before calling

// preflight: verify provider auth before discovery
const auth = await discoverAuthStorage();
if (!auth.get(model.provider)) throw new Error(`provider ${model.provider} not authenticated; cleanse discovery would fail`);

Try / catch

try {
  const specs = await discoverCheckers(signal);
  return specs;
} catch (err) {
  if (err instanceof Error && err.message.startsWith('Checker discovery failed')) {
    if (!signal.aborted) {
      await Bun.sleep(2000);
      return discoverCheckers(signal); // one retry for transient API errors
    }
  }
  throw err;
}

Prevention

When it happens

Trigger: The discovery agent run returns { error } — provider API failure, rate limit, auth failure, signal abort (timeout/cancel), or the model failed to produce a valid structured output.

Common situations: Expired/missing API credentials for the cleanse model; rate limits during parallel discovery; network outage; discovery model returning malformed structured output that the runtime reports as an error.

Related errors

AI-assisted analysis of can1357/oh-my-pi@9690622007 (2026-08-31). Data as JSON: /api/errors/ce2d69b297b908cc. Report an issue: GitHub.