tool-caching

Overview ¶

@sk-task tool-caching#T4.2: comprehensive example with agent + Ollama

This example demonstrates tool result caching at two levels:

1. Direct tool level: a slow tool wrapped with cache — compare miss vs hit latency
2. Agent level: an agent with Cache:true uses the same tools; cache counters show savings

Requires Ollama running on localhost:11434 with a model pulled (e.g., gemma3:4b or llama3.2). The cache works with any LLM provider — Ollama is just the example.

Usage:

ollama pull gemma3:4b
go run ./examples/features/tool-caching/