Jev directory
CommunityJev Integrations & Adapters
@effect-uai/typesafe-ai
Effect provider package that serves the DecisionModel capability with Jev, the System One model answering typed questions with probabilities instead of text. Defines classify, rate, and probability decisions answered in one call, with typed AiError mapping.
- Reading time
- 2 min
- Published by
- Community
- Added
- Oct 2026
Highlights
- Serves the DecisionModel capability with Jev, answering typed questions about an input with probabilities instead of generating text.
- One layer exposes two tags: Jev for its own reported confidence, DecisionModel for provider-portable code.
- Models jev-latest (default), jev-preview, and jev-1.13.0; 64k tokens per request, text only.
- Every decision in a definition is answered in one call and the encoded input is billed once, so decisions should be batched.
- HTTP statuses map to AiError variants including AuthFailed, Timeout, ContextLengthExceeded, RateLimited, and Unavailable.
Questions about this entry
Short answers drawn from the linked project or page.
Run pnpm add @effect-uai/core @effect-uai/typesafe-ai effect. The package is experimental: the class of model is new and its shape may change.
Read it with Config.Redacted("TYPESAFE_AI_API_KEY") and pass it into the jevLayer call, then provide FetchHttpClient.layer to the resulting layer.
jev-latest is the default, with jev-preview and jev-1.13.0 also available. Each request allows 64k tokens, of which roughly 32k covers the encoded input plus the longest single decision, and the model is text only.
No. Its formula is undocumented and it is absent on probability decisions. Decision.confidence and Decision.margin are computed from the distribution and defined for every kind, but normalised entropy shrinks as the label count grows, so gate on top probability and margin.
It encodes the input on the way out, so a schema that omits a field keeps that field from reaching the model. This is the supported way to hold irrelevant or sensitive context out of a request.
Statuses map to AiError variants: 401, 402, and 403 to AuthFailed subtypes auth, billing, and permission; 408 and 504 to Timeout; 413 to ContextLengthExceeded; 422 to InvalidRequest; 429 to RateLimited; and 5xx to Unavailable. Shape mismatches are GenerationFailed.
Known weak spots include counting and arithmetic, date ordering, multi-hop reasoning, and rubric scores, so expectedLevel should not be read as a magnitude. Input is not treated as hostile, so an injected instruction can steer an answer.
Reactions & coverage
Posts, threads, and videos about this entry from around the web.
X: Agent Handoff Gate: agents verify what they hand off
X: Jev as an LLM guardrail inside Agentgateway
X: Vercel's fx safety reviewer, measured at 18× faster
Reddit: Early experiments wiring Jev into Pi as a safety layer and model router.
More Jev resources like this
LangChain.js integration for TypeSafe System One models. It wraps the jev classifier as a Runnable that returns typed choice, score, and noul answers with calibrated probabilities, plus agent middleware for model routing and tool gating.
Jev Integrations & Adapters#typescript#sdk#routing
Community
TypeScript service and library that places Jev between a PBX such as Asterisk or FreePBX and call destinations, classifying caller intent into predefined routes. Each route maps to an explicit PBX action, with low-confidence or unhandled calls sent to a fallback route.
Jev Integrations & Adapters#typescript#routing#github
Communityrayanweragala/jev-call-router
n8n community node that sends a state and a map of Noul, Choice, and Score questions to the System One endpoint in one request, returning calibrated probabilities and confidence for workflow branching.
Jev Integrations & Adapters#github#typescript#guardrails
Communityzampierid4p/n8n-nodes-typesafe
An n8n community node for the TypeSafe AI System One (Jev) API. It adds Ask Yes/No, Ask Choice, Ask Score, Evaluate Questions and Get Many Models operations, so workflows can branch on calibrated probabilities instead of parsed free text.
Jev Integrations & Adapters#n8n#classification#routing
Community
Community n8n node that sends text or JSON items to Jev decision models and routes them by the answer. Supports Choice, Score, and Noul questions, backed by either a local Ollaya daemon or the hosted TypeSafe AI API.
Jev Integrations & Adapters#routing#local#typescript
Community
n8n community node that adds a TypeSafe resource to workflows, exposing Get Many Models and Ask Questions operations against the TypeSafe API. Generated from OpenAPI 0.2.0 and authenticated with a bearer token credential.
Jev Integrations & Adapters#typescript#classification#sdk
Community
From the Jev guides
Original write-ups on what this entry is for.
From the Jev AI model community
Posts from builders shipping with Jev right now.