JevFlow
TypeScript library composing Jev judgments into deterministic threshold workflows that return explainable action sets.
Overview
JevFlow treats probabilistic AI decisions as a programming primitive for backend developers. It keeps two things strictly separate: probabilistic intelligence, where the model answers a question about some state (is this transaction suspicious? 0.93), and deterministic policy, where your application decides what that number means. You define typed judgments with decision.noul, decision.score and decision.choice, then build a workflow that evaluates them and attaches rules such as when suspicious probability is greater than 0.9 then human-review, or when payment risk is at least a level then enhanced-verification.
The judgments in a workflow are batched into a single systemOne call, and running a workflow returns an ordered, explainable action set rather than executing side effects; matched rules record the actual value behind each action. The model never sees your business rules. Jev is the default provider, but the provider is a replaceable seam, and a mock provider lets policy tests run without an API key.
What the decision model decides
| Type | Question | Answers |
|---|---|---|
| Score | How risky is this payment? | Very low · Low · Medium · High · Very high |
| Yes / no | Is this payment suspicious? | — |
How it works
- Define typed judgments
- Compose a workflow with threshold rules
- Run: judgments batched into one call
- Receive ordered actions with matched-rule explanations
Strengths
- Strict separation of model output and business policy
- One batched call per workflow
- Explainable action sets; no hidden side effects
- Mock provider for tests
Caveats
- You must define and tune thresholds
- No benchmarks reported
At a glance
- Type
- Library
- Status
- beta
- Language
- TypeScript
- Runtime
- Node library
- Has code
- yes
- Has tests
- yes
Use cases and integrations
How curated lists describe it
TypeScript ecosystem: composes Jev `Noul`, `Score`, and `Choice` decisions into deterministic threshold workflows that batch into a single `systemOne` call and return an ordered, explainable action set instead of side effects, with matched rules recording the actual value behind each action and a mock provider so policy tests run without an API key.
yibie/awesome-jevProbabilistic AI decisions as composable backend primitives — typed judgments…
wh000wh000/awesome-jev-liveProbabilistic AI decisions as composable backend primitives — typed judgments (noul/score/choice), deterministic thresholds, and explainable workflows. Powered by TypeSafe's Jev, provider-agnostic.
hellogumbo/awesome-jev
Frequently asked questions
Does a workflow execute actions?
No, it returns an ordered action set for your code to execute.
Can I swap providers?
Yes, Jev is the default but the provider is replaceable.
Can I test without an API key?
Yes, a mock provider supports policy tests.
Related projects
Rust CLI and library that asks Jev typed questions via OpenRouter and combines the probabilities with fuzzy-logic rules into decisions.
CLIbetaRustUnofficial standard-library Go client for Jev with generic typed Choice values and local confidence thresholds.
ProjectbetaGoRust CLI for Jev: ask yes/no, choice and rubric questions about text and branch on exit codes, JSON or MCP tools.
MCP serverbetaRustTypeScript library for typed semantic decisions via Zod schemas or native batches, defaulting to Jev, with confidence-gated policies.
LibrarybetaTypeScriptMonitoring proxy and client that records Jev answer probabilities and alerts on version changes, drift, edge crowding or accuracy drops.
ProxybetaPythonComposable Python toolkit for Jev: question builders, confidence gates, batch evaluator, calibration, CLI, MCP server and agent skill.
MCP serverbetaPython