Skip to content
Metatext Decide — typed decisions for your agents, with a probability on every answerGet early access
Metatext
Sign Up

tinyjevclient

Typed Rust client for TypeSafe's System One API that sends Choice, Score and Noul questions and validates responses.

LibrarybetaGPL-3.0-only
View on GitHub ↗

Overview

tinyjevclient is a Rust library for calling Jev through TypeSafe's System One API, or through OpenRouter and a Tiny Humans endpoint that accept the same format. It sends shared state with independent Choice, Score and Noul questions, checks the provider's response against the original request, and returns latency, attempt count, usage and request metadata alongside the typed answers.

The design keeps policy out of the model: a Choice can only pick caller-supplied values, and callers own thresholds, escalation and side effects. The API key is read from TYPESAFE_API_KEY, redacted from Debug output and never placed in errors. Remote endpoints must use HTTPS. Failed calls keep their classified error and timing so reliability measurements stay complete. The repository is GPL-3.0-only and consumed by pinned git revision.

What the decision model decides

TypeQuestionAnswers
ChoiceWhich team should handle this ticket?billing · technical

How it works

  1. Build an EvaluationRequest with state and named questions.
  2. Client sends it to TypeSafe, OpenRouter or Tiny Humans.
  3. The response is validated against the request.
  4. Caller applies its own thresholds and actions to the typed answers.

Strengths

  • Response validation catches provider mismatches.
  • Keys are redacted from logs and errors.
  • Records latency, attempts and usage, including for failures.
  • Supports several providers with the same request shape.

Caveats

  • GPL-3.0-only licence may limit use in proprietary code.
  • Not published to crates.io; used by git revision.
  • Requires a hosted provider key.

Getting started

TYPESAFE_API_KEY='<key>' cargo run -p tinyjevclient --example basic

At a glance

Type
Library
Status
beta
Language
Rust
Runtime
library
Has code
yes
License
GPL-3.0-only

Use cases and integrations

How curated lists describe it

Frequently asked questions

Which providers does it support?

TypeSafe's System One API, OpenRouter, and a Tiny Humans OpenRouter endpoint.

Does it make decisions itself?

No, it returns typed answers; thresholds and actions stay in the caller's code.

How is the API key handled?

Read from TYPESAFE_API_KEY or config, redacted from Debug output and never included in errors.

Related projects

  • ai

    Hack Club's lightweight LLM proxy, listed for adding a Jev forwarding endpoint behind its auth, limits and usage logging.

    MCP serverproductionTypeScript
  • Ax

    Ax, a DSPy-style multi-language LLM framework, listed with a native TypeSafe provider for Boolean, Choice, Score and raw Jev questions.

    PluginproductionTypeScript
  • Effect

    Effect TypeScript library whose @effect/ai-typesafe package maps classify, probability and rating operations to Jev questions.

    ProjectproductionTypeScript
  • instructor-php

    PHP monorepo whose Polyglot module adds a TypeSafe driver for typed Noul, Choice and Score decisions alongside LLM inference.

    ProjectproductionPHP
  • jev-mcp (freepik-company)

    Go MCP server giving any client typed Jev decisions via OpenRouter or TypeSafe, with validated answers and no paid retries.

    MCP serverproductionGo
  • llm-typesafe

    Simon Willison's LLM plugin that adds Jev as a model for yes/no, choice and score questions from the command line.

    PluginproductionPython