Jev in production › Evaluation and testing

grounded-ai (CascadeEvaluator)

Routes agent-turn judgments to Jev before a larger model runs inside an eval harness, benchmarked 51% faster on 200 real turns (vendor).

Open on GitHub ↗

51% fastermeasured against a baseline, as published by the source
Use
Evaluation and testing
Industry
Developer tools
Form
Open-source tool
Stage
In production
Listed
2026-10-09
Found via
github
Repository
grounded-ai/grounded_ai
Stars
71
Forks
7
Last push
2026-10-09
Language
Python
License
MIT

The README opens with

One Python interface for evaluating what LLM apps and agents say and do. Use a frontier model as the judge, a small local model, or a decision model, and swap between them without rewriting your evals.

CascadeEvaluator asks Jev, a fast decision model, first. Jev reports how confident it is in each answer. Confident answers are kept; only the rest go to the LLM judge of your choice.

On 200 real agent turns (should the agent have called a tool here, or replied?), against Claude Sonnet 4.6 judging everything:

Badge

For the project's own README, linking back here:

Listed in Jev in production

Also used for evaluation and testing