[BUG] Rate limiting doesn't work if no fallback models is configured

[VictorSu000]

Quick Summary

I only set agents.defaults.model_name and don't set any fallback models. This model's rpm config doesn't work.

Environment & Tools

• PicoClaw Version: 0.3.1
• Go Version: go 1.26-alpine
• AI Model & Provider: gpt-5.5
• Operating System: inside docker container (debian bookworm), and host machine is Ubuntu Server 24.04.4 LTS
• Channels: WebUI, QQ

📸 Steps to Reproduce

1. Set a model with rpm config in model_list.
2. Set "agents" config to only use this model, that is, only set agents.defaults.model_name. Do not set agents.defaults.model.primary or agents.defaults.model.fallbacks
3. The actual API calling rpm can exceed the setting rpm. Rate limiting doesn't work.

💬 Additional Context

FYI:
I asked AI but was not sure if it is correct. AI said when len(exec.activeCandidates) > 1, callback chain and rate limiter will be used. Otherwise PicoClaw will directly use exec.activeProvider.Chat(...), so rate limiting is bypassed. Besides, side question and pipeline streaming are the same.

[github-actions]

This issue has had no activity for 7 days and has been marked as stale. If it is still relevant, please reply or update; otherwise it will be closed automatically in 7 days.

[github-actions]

This issue has been closed after 14 days of inactivity. If it is still needed, feel free to reopen it anytime.