Transforms
Auto-cache: OpenAI tool results
Insert explicit prompt-cache breakpoints on tool-result blocks for GPT models.
What it does
This transform inserts explicit OpenAI prompt-cache breakpoints on eligible tool-result content blocks. It targets GPT model families that support explicit cache breakpoints.
Requests without eligible tool-result blocks pass through unchanged.
Reference
| Property | Value |
|---|---|
| Type ID | cache_openai_tool_use |
| Phase | request |
| Scopes | Provider, Global, API key |
When to use
- Agent clients run tool loops against an OpenAI-compatible upstream with explicit-breakpoint support.
- You want the tool-loop prefix cached between iterations.
Configuration
This transform has no options. Use an empty config object.
Example rule
{
"transform": "cache_openai_tool_use",
"enabled": true,
"phase": "request",
"models": [
"gpt-*"
],
"config": {}
}