Self Correcting

Improve answer quality with an independent verifier. It reviews the agent’s output, grades it, and can trigger an automatic reprompt when needed.

How it works

• The primary agent answers the user.
• A verifier model evaluates the response for accuracy, completeness, and adherence to policy.
• The verifier assigns a grade (A/B/C/F) and optional notes.
• If enabled, the system auto‑reprompts below a threshold (e.g., C or lower) to improve the result.

Why it matters

• Catches obvious mistakes before users see them.
• Provides transparent grades in traces for debugging and tuning.
• Lets you be aggressive for internal users and conservative for customers.