CopilotKit/CopilotKit · warning

Thread name generation returned an empty or invalid title

Error message

Thread name generation returned an empty or invalid title

What it means

The runtime's automatic thread-naming feature asks the configured agent to generate a title for a new thread, retrying over multiple attempts. When an attempt completes but yields an empty/invalid title (agent returned empty string or the pipeline validated it away), the runtime logs this warning with agentId, attempt, and threadId and tries again; if all attempts fail, the thread simply keeps its default name.

Source

Thrown at packages/runtime/src/v2/runtime/handlers/intelligence/thread-names.ts:75

  }

  let generatedTitle: string | null = null;

  for (let attempt = 1; attempt <= MAX_TITLE_GENERATION_ATTEMPTS; attempt++) {
    try {
      generatedTitle = await runTitleGenerationAttempt({
        runtime,
        request,
        agentId,
        threadId: thread.id,
        prompt,
      });

      if (generatedTitle) {
        break;
      }

      logger.warn(
        { agentId, attempt, threadId: thread.id },
        "Thread name generation returned an empty or invalid title",
      );
    } catch (error) {
      logger.warn(
        { err: error, agentId, attempt, threadId: thread.id },
        "Thread name generation attempt failed",
      );
    }
  }

  await runtime.intelligence.updateThread({
    threadId: thread.id,
    userId,
    agentId,
    updates: { name: generatedTitle ?? FALLBACK_THREAD_TITLE },
  });
}

View on GitHub (pinned to 68fbe97d87)

Solutions

1. Check the configured naming agent/prompt — make it explicitly return a short non-empty title string
2. Use a stronger model for title generation if the current one returns empty responses
3. If thread naming is unimportant, disable the intelligence thread-naming feature to remove the noise
4. Reproduce with logging on the naming attempt to see the raw model output before validation

Example fix

// before: naming prompt that invites empty output
"Name this conversation."

// after
"Name this conversation. Respond with ONLY the title, 3-6 words, never empty. Example: 'Bug fix discussion'"

const isValidTitle = (t: unknown): t is string =>
  typeof t === "string" && t.trim().length > 0 && t.trim().length <= 100;

const isNonEmptyTitle = (t: unknown): t is string =>
  typeof t === "string" && t.trim().length > 0;

When it happens

Trigger: handleIntelligenceRun -> generateThreadNameForNewThread calls runTitleGenerationAttempt; the agent finishes generation but generatedTitle is falsy — e.g. the model replied with empty output, a title-extraction step stripped everything, or the agent's graph ends without producing a title state.

Common situations: Small/weak models returning empty completions for naming prompts; misconfigured title-generation prompts that get filtered out by output validators; token limits truncating the title to nothing; intelligence/thread-naming feature enabled with a model that doesn't handle the naming task.

Related errors

• Skipping thread naming because the agent could not be cloned
• channel "${opts.name ?? "(unnamed)"}" failed to start its ma
• channel_memory_unavailable
• A delegate is only created for Intelligence mode
• Intelligence mode requires runtimeUrl, agentId, and intellig

AI-assisted analysis of CopilotKit/CopilotKit@68fbe97d87 (2026-08-27). Data as JSON: /api/errors/c68865b68cce777d. Report an issue: GitHub.