Plugin: apu_governor

[a0-bot[bot]]

A.P.U Automatic Polling Utility

Non-blocking API rate throttle for Agent Zero. Enforces configurable cooldown delays after every tool execution to prevent 429 errors and quota exhaustion. High-compute detection (api/fetch/download/scrape/crawl/stream patterns + runtime > 5s) applies a multiplier automatically. Interruptible wake button in the chat bar. Cooldown state persists across container restarts. Zero external dependencies.

Quick links

• Index entry: https://github.com/agent0ai/a0-plugins/tree/main/plugins/apu_governor
• Plugin repository: https://github.com/MrTrenchTrucker/apu-governor
• Author: @MrTrenchTrucker

About this discussion

This discussion is auto-created by a workflow when a plugin is added to the index.

[DonovanDeHart]

For a consumer PC dual rtx 5090s, 96gb ram, Amd 9950x3d cpu, what would be safe recommended variables?

[MrTrenchTrucker]

If you are running you're own models with only one AI Agent, then I suggest setting the Minimum Cooldown (seconds) closer to 0. No need to cool down unless you have multiple models for A0 that you are hot swapping or multiple AI Agents using the same GPU pool. I still reccomend the other 2 settings tho for tool calling but with a 9950x3d it should be fast enough to set the Base Wait (seconds) extremely low assuming you are not running anything else off that hardware at the same time or that you are not using A0 to download like 100gb worth of LLM Models at a single time. You can also point your Agent Zero at the official GitHub repository, and it should be able to understand the instructions and set itself using a tool call. It can do so dynamically as needed which I recommend for people who use their own hardware GPU pools.
Honestly the best advantage for you is the tool call waiting function will save you electricity and mechanical stress on your gpu pool. Hope it helps!