Need clarification on API usage behavior – high token usage & 429 errors
Hi @Krill 🦐
I recently started testing Moltbot with the Google Gemini API (Gemini 2.5 Flash) and connected it using my own API key. I wanted to understand how Moltbot actually consumes tokens and whether my current behavior is normal.
After only a small amount of testing, I noticed the following in my Gemini dashboard:
• Around 60 total API requests
• A sudden spike of about 1.5M+ input tokens
• Multiple 429 TooManyRequests error
• A sharp drop in success rate on the same day
From what I understand, Moltbot makes multiple LLM calls per task (reasoning, tool selection, execution, etc.), but I was surprised to see such high token usage from relatively few commands.
I have a few questions:
1. How exactly does Moltbot structure its prompts when using external APIs like Gemini or GPT?
• Does it resend full memory, tool schemas, and history on every request?
2. What are the best practices to avoid:
• Large context size
• Rapid token exhaustion
• Rate-limit errors (429)
3. Are there recommended settings for:
• Context window limits
• Request throttling
• Disabling autonomous loops
• Reducing memory per call
4. Is it normal for agent workflows to reach millions of input tokens so quickly, or does this indicate a configuration issue on my side?
My goal is to use Moltbot safely with my own API key without accidentally exhausting quota or hitting limits.
Any guidance on optimal configuration for Gemini (or GPT) would be greatly appreciated!
Thanks in advance
I recently started testing Moltbot with the Google Gemini API (Gemini 2.5 Flash) and connected it using my own API key. I wanted to understand how Moltbot actually consumes tokens and whether my current behavior is normal.
After only a small amount of testing, I noticed the following in my Gemini dashboard:
• Around 60 total API requests
• A sudden spike of about 1.5M+ input tokens
• Multiple 429 TooManyRequests error
• A sharp drop in success rate on the same day
From what I understand, Moltbot makes multiple LLM calls per task (reasoning, tool selection, execution, etc.), but I was surprised to see such high token usage from relatively few commands.
I have a few questions:
1. How exactly does Moltbot structure its prompts when using external APIs like Gemini or GPT?
• Does it resend full memory, tool schemas, and history on every request?
2. What are the best practices to avoid:
• Large context size
• Rapid token exhaustion
• Rate-limit errors (429)
3. Are there recommended settings for:
• Context window limits
• Request throttling
• Disabling autonomous loops
• Reducing memory per call
4. Is it normal for agent workflows to reach millions of input tokens so quickly, or does this indicate a configuration issue on my side?
My goal is to use Moltbot safely with my own API key without accidentally exhausting quota or hitting limits.
Any guidance on optimal configuration for Gemini (or GPT) would be greatly appreciated!
Thanks in advance
The AI that actually does things. Emails, calendar, home automation - all from your favorite chat app. New shell, ready to help. The lobster way.
85,605Members
Resources
Similar Threads
Was this page helpful?
