Idea: Optional follow-up suggestions after each LLM response

[ZhouJ-sh]

Idea

After an LLM finishes its final response, nanobot could optionally show a small set of suggested follow-up questions or next steps.

For example:

The repository analysis is complete.

• Review the highest-priority issue
• Compare the related pull requests
• Draft an implementation plan

A user could select one suggestion to send it as the next user message, or ignore the suggestions and continue typing normally.

Proposed behavior

• Generate 2–4 concise, context-aware suggestions after a completed assistant response.
• Suggestions may be follow-up questions or clear next actions.
• Selecting a suggestion sends it as a normal user message. It must not execute an action automatically.
• Hide suggestions when they would add little value, such as simple greetings or short factual answers.
• Keep the feature optional through configuration so users can enable or disable it.

The exact configuration shape is open for discussion. For example:

{
  "followUpSuggestions": {
    "enabled": true
  }
}

Why this may help

• Makes it easier to continue exploring a topic.
• Helps users discover useful next actions without having to formulate another prompt.
• Could reuse the existing structured choice/button capabilities already used by WebUI and Telegram.

Questions for the community

1. Would you use this feature?
2. Should it be enabled or disabled by default?
3. Should suggestions be generated by the same LLM response, or by a separate lightweight call?
4. How many suggestions would feel useful without adding noise?
5. Should this initially target WebUI only, or use a channel-neutral response format?
6. Are there privacy, latency, token-cost, or UX concerns that should shape the design?

Related existing work includes #708, #3398, #3454, and #975, which provide interactive choice or structured reply foundations but do not currently add automatic post-response follow-up suggestions.