JevMade Sign in
← Back to experiments

Apps & data pipelines

Kartify Support

Source screenshot of Kartify Support
SOURCE SCREENSHOTFull screenshot ↗

What it does

A customer-support demo that uses GPT-4o to write replies and Jev to check them.

How you can use it

A developer can borrow this code for a customer service app. The app reviews drafted messages before sending them. Your developer can read the safety rules code to see how to ask several questions together. The project requires an access key that connects the software to TypeSafe.

The checker AI struggles with math. The app calculates order dates first. It then asks the checker to review the facts. If a check takes too long, the software waits for a person instead of guessing.

Maker-reported (not independently measured by JevMade): 28 of 29 held-out cases correct, against 25 for the original GPT-4o prompts (author-reported) · median about 170 ms per check against about 630 ms (author-reported) · thresholds HAZARD_BLOCK_AT 0.5, INTENT_MIN_CONFIDENCE 0.5 and EVAL_PASS_AT 0.75 (source defaults)

Primitives
choice, noul
Platform
Python
Added
Project created
GitHub stars
0 (snapshot, not live)

Keep this for later

Sign in to bookmark experiments, guides and videos, and keep notes only you can see.

Continue to sign in

We’ll bring you back to this listing.