# Migration and coexistence

> Local Laya as a pre-filter plus hosted Jev for escalation, with schema mapping, calibration, and ops checks when you leave API-only System One.

## Answer

Migration is a schema, calibration, and ops project. It is not a find-replace on endpoint URLs.

Map Jev-shaped gates to Laya checkpoints without breaking production. And know when not to migrate.

## Start from your Jev contract

Inventory parallel questions, state shape, thresholds, and escalation rules in your existing TypeSafe client (/learn/jev-typesafe). Note option counts and languages. That map tells you whether laya-typed-decisions, multilingual Router, or staying on Jev for specific steps is realistic.

## Schema mapping

Laya's Python SDK expects question dicts with type choice, score, or noul and criteria blocks. Jev's POST /v1/systemone JSON is similar in spirit but not byte-identical. Plan an adapter layer in your service rather than duplicating business logic in two formats.

## Hybrid pattern: local pre-filter

Run Laya on-device or in a sidecar for cheap parallel questions (language detection, coarse routing, game ticks). Send only uncertain or high-stakes rows to hosted Jev. This preserves data minimization while keeping an authoritative cloud path for audit.

## Hybrid pattern: environment split

Developers on Apple Silicon use Laya-MLX for fast iteration. Staging and production stay on Jev until you have GPU capacity and MLOps to host Convai checkpoints with monitoring. /learn/where-to-run-jev lists hosted options.

## Calibration migration

If you threshold Jev probabilities today, re-fit calibration on Laya outputs before copying numeric cutoffs. Convai documents temperature scaling per question type. Your legacy 0.92 gate may mean something different on open weights.

## When to stay on Jev only

Skip migration when you rely on gateway billing, need more than 50 options without tuning, lack GPU ops, or require TypeSafe's supported SLA. Laya is not a moral upgrade. It is a deployment choice.

## In the directory

- [Laya vs Jev hub](https://jev.aitools.fyi/learn/laya-vs-jev)
- [Laya vs Jev compared](https://jev.aitools.fyi/learn/laya-vs-jev/compare)
- [System One primer](https://jev.aitools.fyi/learn/system-one)
- [What Jev is](https://jev.aitools.fyi/learn/jev-typesafe)
- [Benchmarks](https://jev.aitools.fyi/categories/benchmarks)
- [SDKs and clients](https://jev.aitools.fyi/categories/sdks)
- [@receptron/laya](https://jev.aitools.fyi/tools/receptron-laya)
- [ollaya](https://jev.aitools.fyi/tools/ollaya-dev-ollaya)
- [Laya-CoreML](https://jev.aitools.fyi/tools/mizorewww-laya-coreml)

## FAQ

### Can I run the same questions on both?

Yes for experiments. Log disagreements and calibration drift. Do not dual-write production decisions without a clear source of truth.

### Does Vercel AI Gateway support Laya?

Gateway documentation centers on hosted Jev. Local Laya sits outside that path unless you wrap it in your own HTTP service.

## Sources

Directory copy checked 2026-09-30. Primary sources:

- [Hugging Face: convaiinnovations/laya (model card and benchmarks)](https://huggingface.co/convaiinnovations/laya?ref=jev.aitools.fyi)
- [GitHub: NandhaKishorM/laya (upstream SDK)](https://github.com/NandhaKishorM/laya?ref=jev.aitools.fyi)
- [PyPI: laya package](https://pypi.org/project/laya/?ref=jev.aitools.fyi)
- [TypeSafe docs: System One / Jev API](https://docs.typesafe.ai/?ref=jev.aitools.fyi)
- [Laya-MLX: BENCHMARKS.md (M3 Max, independent port)](https://github.com/mizorewww/laya-mlx/blob/main/BENCHMARKS.md?ref=jev.aitools.fyi)

Canonical: https://jev.aitools.fyi/learn/laya-vs-jev/migration-and-coexistence
Markdown: https://jev.aitools.fyi/learn/laya-vs-jev/migration-and-coexistence.md