An internal tool by NexiHealth
Did this change make our AI worse?
LedgeLM collects the AI evaluation results that NexiHealth's CI runs produce and compares every pull request with recent results from the main branch.
The dashboard is only available to members of the NexiHealth GitHub organization.
Collect eval results
CI jobs report results from the eval tools the team already uses: pytest, Vitest, promptfoo, notebooks or custom judges.
Compare with a baseline
Each pull request is compared with a rolling baseline from recent mainline runs, not only the previous commit.
Report on pull requests
LedgeLM publishes one GitHub Check per run, so regressions show up where code review happens.