AI Agents / Tool contracts and execution
AI Agents: Dry-run contracts
Make preview return the exact proposed changes and target identifiers.
Environment and inputs: Model tool calling with a validated application dispatcher; Ollama is a local-runtime example.
When to use it
The model proposes a tool call; application code decides whether and how it executes. Test the contract at the dispatcher boundary so a fluent model response cannot conceal an invalid argument or an uncommitted operation.
Check the result
Compare an accepted preview with the eventual write and reject changed targets.
Save and compare results
Save the test input, configuration and result. Fix the failing boundary, repeat the same test and compare it with the saved working configuration.
Scope and limits
These exercises have not been tested on your systems. Check feature support and configuration names against the installed release and linked documentation.