How is this different from prompting an LLM guardrail for JSON?

ReflexGate fans seven questions out in one Jev call and lets TypeScript code own the verdict, instead of chaining prompts and parsing free text with retry loops for routing.

What happens when the Jev API is unreachable or slow?

The policy router still runs, but verdicts degrade: unverifiable requests surface as review-unavailable style outcomes, and the audit store records what was evaluated so gaps are visible.

Can I retune blocking thresholds without redeploying?

Yes, through the REST API or dashboard sliders for injection thresholds, toxicity limits, PII sensitivity, and escalation triggers, with no server restart required.

Where do past triage decisions and audit trails live?

Every evaluated payload lands in a local SQLite store via node:sqlite with WAL and prepared statements, searchable and filterable with one-click CSV export.

What are the seven dimensions judged in the single call?

Prompt injection, toxicity, PII or secret leaks, department routing, incident severity, customer churn risk and actionability. Jev fans them out in one request, and the TypeScript router turns the scores into SECURITY_BLOCK, ESCALATE_CRITICAL, AUTO_DISPATCH or HUMAN_REVIEW.

What are the runtime requirements for ReflexGate?

Node.js 22.0.0 or higher and a TypeSafe API key. The audit store uses Node 22's native node:sqlite with WAL and prepared statements, so there is no external database to run, and multi-stage Docker and compose files ship for container deployment.

Can I try it without wiring up my own payloads?

npm run demo runs the core engine against four built-in scenarios: a prompt-injection jailbreak, a production database outage, a billing receipt request and an ambiguous payload. npm run dev then serves the interactive dashboard at localhost:3000.

CommunityJev Benchmarks & Evaluations

Reflexgate

TypeScript gateway that fans one payload out to seven Jev risk and routing questions in one call, then enforces deterministic policy verdicts.

See highlights

Open github.com

Stars: 0
Language: TypeScript
License: MIT
Published by: Community
Author: intelliDean
Last verified: Oct 2026

intelliDean/reflexgate

User repository on GitHub

View on GitHub

Ultra-fast, sub-100ms API & webhook guardrail and triage gateway powered by TypeSafe AI System One (Jev). Parallel 7-dimension speculative evaluation, deterministic policy router, zero-dependency SQLite audit trail, and automated outbound dispatch.

Forks: 0
Last pushed: Updated Sep 2026

Repo stats from the GitHub API, cached Oct 2026.

About this project

Intelligent, Sub-100ms API & Webhook Guardrail and Triage Gateway Evaluating 7 Parallel Risk & Operational Dimensions with TypeSafe AI System One (Jev) Features • Architecture • The System One Advantage • Quick Start • REST API • Dashboard • Deployment.

Quoted from the project's README. Read the full README.

Tagscommunity github

Works forJev for Routing Jev for Evaluation Jev for Classification

Used inJev for Software & DevTools

Languages: TypeScript

Highlights

One Jev call scores prompt injection, toxicity, PII leak, department, severity, churn risk, and actionability.
Deterministic router emits SECURITY_BLOCK, ESCALATE_CRITICAL, AUTO_DISPATCH, or HUMAN_REVIEW from scores.
SQLite audit store with search, filters, and CSV export plus Slack, Discord, and downstream dispatch.
Dashboard with live workbench presets, policy sliders, outbound monitor, plus Docker deployment files.

Watch out

Behavior depends on your Jev key, thresholds, and payloads; tune gates before trusting blocks or dispatches. Copy reviewed against the repository README on 2026-10-04, not from running it.

Questions about this entry

Short answers drawn from the linked project or page.

More Jev resources like this

LangWatch

4.9k

LangWatch Instant Evals ships a TypeSafe Jev classifier that batches Noul, Score, and Choice judgments on production traces in one POST per text.

Jev Benchmarks & Evaluations#community#typescript#evaluation

Community

langwatch/langwatchTypeScript

convex-evals

130

Convex's own coding evals, plus a separate multiple-choice decision benchmark in the same repo. Jev is wired into the decision side through OpenRouter's alpha decisions endpoint with a pinned version, while the coding evals still route as chat completions.

Jev Benchmarks & Evaluations#community#jevbest

Community

get-convex/convex-evalsTypeScript

jev-evals

Rubric-based eval harness that judges LLM and agent outputs with Jev through the Vercel AI Gateway, answering all rubrics for a case in one round trip.

Jev Benchmarks & Evaluations#community#github

Community

NicolasMontone/jev-evalsTypeScript

jev-eval

Third-party TypeScript check that compares Jev (typesafe-ai/jev) against openai/gpt-4o-mini and anthropic/claude-sonnet-4.5 under identical conditions, routing 60 synthetic booking inquiries in four languages for a photo-shoot service, all through one Vercel AI Gateway key.

Jev Benchmarks & Evaluations#community#benchmark#comparison

Community