---
name: jevproxy
license: MIT
description: >
  Autonomous agent decision gateway for any domain (Coding, E-Commerce, Finance, Healthcare, DevOps, Legal).
  Intercepts intermediate agent micro-decisions, tool routing, classification, and safety guardrails in sub-25ms
  at $0.0001 instead of waiting 1.4s+ and burning $0.015+ on frontier autoregressive LLMs (Claude 3.5, GPT-4o).
  Built on TypeSafe AI System One architecture and calibrated non-autoregressive decision models.
---

# Universal JevProxy Agent Gateway Skill

Use **JevProxy** to intercept repetitive, deterministic, and control-plane agent turns in **sub-25ms at $0.0001** before escalating complex reasoning or creative code generation to frontier models (such as Claude 3.5 Sonnet or GPT-4o).

---

## 1. Gateway Configuration & Authentication

- **Production Endpoint**: `https://api.jevproxy.com/v1` (fallback: `http://194.163.131.175/v1`)
- **Authentication**: `Authorization: Bearer jev_live_...`
- **Supported Headers**:
  - `x-jev-route: force` — Forces sub-25ms Jev System One execution regardless of prompt length.
  - `x-jev-route: bypass` — Passes directly through to upstream LLM.
  - `x-openai-api-key: sk-...` — Optional: customer's own key for creative fallback text.

---

## 2. Drop-in OpenAI SDK Integration

```typescript
import OpenAI from "openai";

const openai = new OpenAI({
  baseURL: "https://api.jevproxy.com/v1",
  apiKey: process.env.JEVPROXY_API_KEY || "jev_live_...",
});

// Classification, intent routing & tool picking execute in <25ms
const completion = await openai.chat.completions.create({
  model: "gpt-4o", // Automatically intercepted by Jev System One
  messages: [
    { role: "system", content: "Classify incoming ticket priority." },
    { role: "user", content: "Refund request: charged twice on order #412" }
  ],
});
```