Reduce LLM reply length | Friends of the Crustacean 🦞🤝 | Page 1

modest olive Apr 23, 2026, 2:48 PM

#

The reply from LLM somehow too verbose. Is it posible to limit the respond length to shorter, more focus?

queen monolithBOT Apr 23, 2026, 2:48 PM

#

dire quartzBOT Apr 23, 2026, 2:48 PM

#

To help others find answers, you can mark your question as solved via Right click solution message -> Apps -> ✅ Mark Solution

queen monolithBOT Apr 23, 2026, 2:48 PM

#

modest olive Apr 23, 2026, 2:56 PM

#

How to make LLM respond shorter, for example 1 ~ 3 sentences at most?

soft caveBOT Apr 23, 2026, 2:57 PM

#

Yep.

Best fix: add a short style rule in SOUL.md or AGENTS.md, for example:

Keep replies short and focused.
Default to 1 to 3 sentences max.
No bullet lists unless the user asks.

If you want a harder limit, also set model params in config, for example:

{
  agents: {
    defaults: {

#

      params: {
        maxTokens: 120,
        textVerbosity: "low"
      }
    }
  }
}

textVerbosity: "low" is mainly useful on OpenAI/GPT-5 style models, while the prompt rule works across models. Docs: https://docs.openclaw.ai/concepts/soul and https://docs.openclaw.ai/gateway/configuration-reference

modest olive Apr 23, 2026, 2:59 PM

#

Will textVerbosity work for other model like Gemini or Deepseek?

soft caveBOT Apr 23, 2026, 2:59 PM

#

Not reliably, no.

textVerbosity is an OpenAI-style knob, so for Gemini or DeepSeek you should assume it does not help. For cross-model control, put the rule in SOUL.md / AGENTS.md instead, for example: Default to 1-3 sentences, concise, no bullet list unless asked.

If you use maxTokens, it can help a bit, but it is a blunt limit and may cut answers off mid-sentence.

#Reduce LLM reply length