What is this credit score that you have? And how does it get depleted?
You're asking about the budget system shown in your message - specifically the tool call budget I operate under. Let me explain it clearly.
My Tool Call Budget
Each conversation I have is given a fixed budget of tool calls and model iterations. This is displayed in the
<budget> block in your message:Model calls: 0/19 used
Tool calls: task: 0/2, query_library_db: 0/10, fetch_image: 0/10 ...
What it tracks
| Item | What it means |
|---|---|
| Model calls | Each "think + act" cycle I perform counts as 1. I have 19 total per session. |
| Tool calls (per tool) | Each time I invoke a specific tool (e.g. web_search, pubmed_search), it counts against that tool's limit. |
How it gets depleted
- Every time I search the web, that uses 1 of my 5
web_searchcalls. - Every time I look something up in a medical textbook, that uses 1 of my 10
query_library_dbcalls. - Every time I think and respond (even just to write text), that uses 1 of my 19 model calls.
- Parallel tool calls (multiple tools fired at once) each count toward their own individual limits simultaneously.
Why it exists
This budget system is designed to:
- Control costs - each tool invocation consumes compute resources.
- Prevent runaway loops - it stops me from endlessly searching or iterating.
- Encourage efficiency - I should make smart, parallel calls rather than wasteful sequential ones.
What happens when it runs out?
If I exhaust a tool's budget, I can no longer use that tool for the rest of the conversation. If model calls run out, I stop being able to respond further in that session.
It's essentially a per-conversation resource meter - not a persistent "credit score" that carries across sessions. Each new conversation starts fresh with a full budget.
Oh, okay, so if I, my credit score decreases, I can start a new conversation, right?
Yes, exactly! Starting a new conversation resets the budget completely back to full.
A few things to keep in mind:
- New conversation = fresh budget. All tool call limits and model call limits reset to their maximums.
- Context is lost. When you start a new conversation, I won't remember anything from the previous one - no chat history carries over.
- It's not really a "credit score" in the traditional sense - it's more like a per-session resource allowance. There's no penalty for running it down, and no reward for conserving it. It just resets.
So if you're in a long, tool-heavy conversation and notice the budget getting low, starting fresh is a perfectly valid move.
This is a shared conversation. Sign in to Orris to start your own chat.