Jev architecture patterns
TypeSafe documents application patterns: speculative fan-out, confidence-gated routing, composite scoring, intent routing. They do not publish a full model-architecture paper in the pages we used. This site will not invent one.
Independent. Unknowns stay labeled. docs.typesafe.ai/patterns.
Independent angle (cover, do not clone)
Win delta (architecture-known-unknowns): explicit known/unknown boundary; no speculation as fact; cite TypeSafe only. Not a clone of SOM’s architecture guide.
Known (documented)
- One evaluation endpoint, questions in parallel, state ingested once.
- Three primitives; mix in one call.
- Confidence on Choice/Score; Noul is a single probability.
- Official cookbooks for RAG filter, citation check, hierarchical Choice, etc.
- Jaggedness list for
jev-1.13.
Unknown (do not fabricate)
- Layer counts, sampler internals, exact RLCD recipe beyond their primer.
- Unpublished latency distributions on your mix.
- Whether a future version keeps the same jaggedness.
If a rival site asserts internals without a TypeSafe citation, treat it as rumor.
Patterns we use on jev.pro
Decide-then-draft; regex-then-Jev; central decision service; job-queued evaluation; cache policy at the edge, call Jev only when allowed.
How we compose (application level)
These are our names, not TypeSafe’s nav:
- Regex-then-Jev — identifiers in code, paraphrase in Jev.
- Decide-then-draft — Jev enum, LLM prose.
- Shortlist-then-score — retrieve K, Jev keep/drop.
- Central service — apps see enums; one adapter to a door.
- Queued evaluation — sync POST inside a worker.
Unknown internals stay unknown. If a handbook draws a non-autoregressive sampler as fact without a TypeSafe citation, ignore it here.
What this page does not claim
- No architecture fiction.
FAQ
Is Jev non-autoregressive? Some third-party glossaries say so. We will not promote it as fact without TypeSafe docs.
Where is the official pattern list? Patterns.
Disclaimer
This is an independent unofficial site and is not affiliated with TypeSafe AI; official documentation is available at https://docs.typesafe.ai. Never treat jev.pro as TypeSafe official documentation. We do not sell, issue, or proxy API keys.
Open-cluster pages are independent field-guide notes. Replicas and third-party interfaces mentioned anywhere on jev.pro are not Jev and not endorsed. Hub: Open. Siblings: reference architecture, model agnostic design, vs centralized decisions. Canonical: https://docs.typesafe.ai.
Sources
Public TypeSafe or adjacent documentation only. No private claims.