XingAI Eval Registry
An Every-Eval-Ever-compatible evaluation registry—store, validate, and diff AI eval results with a fail-on-regression CI gate. Code open source; your evaluation data stays private.
Can do
Register eval runs as EEE-shaped JSON records, compare any two runs metric-by-metric, and fail CI when a metric regresses.
Best for
Teams shipping AI features who want regressions caught by a gate, not by users.
Screenshots
Features
- ✓Every Eval Ever record schema
- ✓One JSON file per record, git-diffable
- ✓Metric direction semantics (higher/lower is better)
- ✓diff --fail-on-regression CI gate
- ✓Private data directory override
- ✓MIT licensed
Versions & pricing
Free
$0
- ✓Every Eval Ever record schema
- ✓One JSON file per record, git-diffable
- ✓Metric direction semantics (higher/lower is better)
- ✓diff --fail-on-regression CI gate
- ✓Private data directory override
- ✓MIT licensed
Pro
Contact us
- ✓Every Eval Ever record schema
- ✓One JSON file per record, git-diffable
- ✓Metric direction semantics (higher/lower is better)
- ✓diff --fail-on-regression CI gate
- ✓Private data directory override
- ✓MIT licensed
Enterprise
Contact us
- ✓Every Eval Ever record schema
- ✓One JSON file per record, git-diffable
- ✓Metric direction semantics (higher/lower is better)
- ✓diff --fail-on-regression CI gate
- ✓Private data directory override
- ✓MIT licensed
Want the full source code?
Fork the open-source Next.js app — inventory scan, meal recommendations, and cooking steps. Deploy your own instance on Vercel.
View source on GitHub →Need a custom version?
We build tailored versions for teams and businesses. Tell us what you need and we’ll scope it together.
Contact usRelease roadmap
What’s coming next for this product.
- ShippedEEE-shaped records + CLI (add/list/show/diff)
- ShippedFirst consumer: Evidence Engine
- In progressCI regression gate in consumer repos
- PlannedEvaluation card export
- PlannedSecond consumer: SAT AI / claims verification