tp-typesafe-jev
Browser studio and templates for trying Jev as a System 1 reflex gate in front of frontier LLMs, with code exporters.
This repository is mostly documentation or write-up rather than runnable code.
Overview
tp-typesafe-jev is a repository of integration templates, blueprints and an interactive web studio for TypeSafe's Jev. It frames Jev as the fast System 1 in a dual-process architecture: a single forward pass returns typed outputs with calibrated confidence, and only cases below a confidence gate (0.85 in its diagram) go to a slower frontier LLM.
The hosted studio lets you simulate four scenarios: an agent reflex router, SRE outage and log triage, a prompt-injection guardrail and a cloud cost anomaly check. It exports code for Pydantic, Zod, LangGraph and Kubernetes sidecars. The README makes large savings claims (90 to 98 percent fewer LLM calls, specific per-decision costs) that are the author's projections rather than reported measurements, so treat them with care.
What the decision model decides
| Type | Question | Answers |
|---|---|---|
| Yes / no | Is calibrated confidence at least 0.85 so the fast path can act? | yes · no |
| Choice | How should this alert, ticket or payload be routed?(inferred) | — |
How it works
- Unstructured state (log, alert, query) enters.
- Jev returns a typed answer and confidence.
- At or above the threshold, a deterministic tool runs.
- Below it, the case escalates to a frontier LLM.
Strengths
- Interactive studio for experimenting without writing code.
- Code exporters for several stacks.
Caveats
- Savings and latency figures are projections, not reported benchmarks.
- Mostly templates and marketing-style documentation.
- Hosted studio is on the author's personal site.
Getting started
Open the interactive studio at https://talaripradeep.info/tools/jev-system1/At a glance
- Type
- Project
- Status
- prototype
- Runtime
- browser studio, templates
Use cases and integrations
How curated lists describe it
Developer Studio repository for TypeSafe Jev: System 1 AI Studio. Category: ai.
wh000wh000/awesome-jev-live
Frequently asked questions
What is this repository?
Templates, blueprints and a browser studio showing Jev as a fast gate in front of slower LLMs.
Are the cost numbers measured?
They are presented as design projections; no benchmark method is given in the README excerpt.
How do I try it?
Open the hosted studio and pick one of its four scenarios.
Related projects
Korean course code for attaching Jev to Claude Code, Codex and OpenClaw: skills, hooks, routers and open-model comparisons.
ProjectproductionJavaScriptCompanion code for the ZazenCodes course on decision models: primitives notebook, use-case examples and a Pi model-router build.
ProjectproductionPythonFree 12-notebook course building 13 agent tools with Jev as the fast decider and an LLM as the slow writer, for under $0.20.
ProjectproductionPythonSix side-by-side demos where Jev makes the decision, Claude or Kimi writes the words, and code owns control flow.
AppprototypeTypeScriptBeginner manual, video script and three small examples for building an agent harness with Jev via langchain-typesafe.
ProjectprototypePythonGuide and small prototype for using Jev with Herdr to route tasks to coding agents like Claude, Codex and Hermes under code-owned policy.
Projectprototype