Skip to content

v1.12 — Interactive Tool Permissions & Self-Correction

Choose a tag to compare

View all tags
@mdaiWorks mdaiWorks released this 01 Oct 19:29
· 28 commits to main since this release
v1.12
b3467bc

🛡️ Interactive Tool Permissions & Self-Correction (v1.12)

1. User Permission Gate for Unselected Tool Requests

When the AI requests a tool not pre-selected by the Router, Yengi now prompts for explicit user permission:

  • [ ✅ Allow & Add Tool ] — Dynamically adds the tool to the catalog and lets AI continue safely.
  • [ ❌ Deny Permission ] — Politely notifies the AI to proceed without using that tool.

2. Self-Correction Retry Loop for Missing Tool Arguments

If a tool invocation fails due to missing JSON arguments, Yengi automatically sends a self-correction hint back to the model:

  • The AI auto-corrects its parameters on the next turn without stopping the workflow or confusing the user.

3. Loop Continuation Fix

Resolved an issue where approving "Continue 10 More Steps" did not re-enter the loop body.

Built with .NET 10 (LTS) | WPF | Full TR/EN/ZH Localization

Assets 3
Loading