Jev vs JSON Mode：类型化决策与受限生成的对比

import json
import time
from openai import OpenAI

client = OpenAI()
VALID_QUEUES = {"billing", "technical", "sales", "abuse"}
MAX_RETRIES = 3

def route_ticket(ticket_text: str) -> dict:
    for attempt in range(MAX_RETRIES):
        resp = client.chat.completions.create(
            model="gpt-4o-mini",
            response_format={"type": "json_object"},
            messages=[
                {"role": "system", "content": 'Reply as JSON: {"queue": "...", "urgent": true|false}'},
                {"role": "user", "content": ticket_text},
            ],
        )
        payload = json.loads(resp.choices[0].message.content)  # json_mode：一定能解析
        if payload.get("queue") in VALID_QUEUES:  # 值校验：json_mode 对此只字未提
            return payload
        time.sleep(2 ** attempt)  # 幻觉枚举 -> 整个生成重新计费
    raise ValueError("queue never converged")

# 典型路径：每次尝试 500-1500ms，每轮重试都按 20-60 个输出 token 计费

import requests

JEV_ENDPOINT = "https://api.typesafe.ai/v1/jev/evaluate"
AUTO_ROUTE_CONFIDENCE = 0.85

QUESTIONS = {
    "intent": {
        "type": "choice",
        "instructions": "判断这条进站消息属于哪类",
        "criteria": {
            "billing": "付款、发票或退款问题",
            "technical": "产品故障或 bug",
            "sales": "购买、升级或报价咨询",
            "abuse": "安全或滥用举报",
        },
    },
    "is_actionable": {
        "type": "noul",
        "instructions": "这条消息需要团队采取行动吗？",
    },
    "priority": {
        "type": "score",
        "instructions": "为处理优先级打分，1（有空再说）到 5（放下手头一切）",
    },
}

resp = requests.post(
    JEV_ENDPOINT,
    json={"state": {"message": "..."}, "questions": QUESTIONS},
    timeout=5,
)
resp.raise_for_status()
data = resp.json()

if data["intent"]["confidence"] >= AUTO_ROUTE_CONFIDENCE:
    lane = f"auto:{data['intent']['answer']}"  # ~70-100ms，只计输入 token，零重试
else:
    lane = "review"

import OpenAI from "openai";
import { z } from "zod";
import { zodResponseFormat } from "openai/helpers/zod";

const RouteDecision = z.object({
  queue: z.enum(["billing", "technical", "sales", "abuse"]),
  urgent: z.boolean(),
});

const client = new OpenAI();

const resp = await client.chat.completions.create({
  model: "gpt-4o-mini",
  response_format: zodResponseFormat(RouteDecision, "route_decision"),
  messages: [{ role: "user", content: ticketText }],
});

const decision = RouteDecision.parse(
  JSON.parse(resp.choices[0].message.content!),
);
// 形状有保证。"queue" 是不是正确的队列——以及模型有多确定——
// 都无法表达：你拿到的是一个猜测，不是校准过的决策。

const JEV_ENDPOINT = "https://api.typesafe.ai/v1/jev/evaluate";
const AUTO_ROUTE_CONFIDENCE = 0.85;

const QUESTIONS = {
  intent: {
    type: "choice",
    instructions: "判断这条进站消息属于哪类",
    criteria: {
      billing: "付款、发票或退款问题",
      technical: "产品故障或 bug",
      sales: "购买、升级或报价咨询",
      abuse: "安全或滥用举报",
    },
  },
  is_actionable: {
    type: "noul",
    instructions: "这条消息需要团队采取行动吗？",
  },
  priority: {
    type: "score",
    instructions: "为处理优先级打分，1（有空再说）到 5（放下手头一切）",
  },
} as const;

const resp = await fetch(JEV_ENDPOINT, {
  method: "POST",
  headers: { "Content-Type": "application/json" },
  body: JSON.stringify({
    state: { message: ticketText },
    questions: QUESTIONS,
  }),
});
const data = await resp.json();

const lane =
  data.intent.confidence >= AUTO_ROUTE_CONFIDENCE
    ? `auto:${data.intent.answer}`
    : "review"; // 低置信带或安全旗标 -> 同一条复核队列