GPT-5.4 API：Responses 与界面截图分析

curl --fail-with-body --max-time 180 https://api.apizt.com/v1/responses \
  -H "Authorization: Bearer $tokenba_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model":"gpt-5.4","input":"核对这段 CSV 的订单收入口径：id,status,amount_minor,currency；A,paid,1200,CNY；B,refunded,800,CNY；A,paid,1200,CNY；C,paid,,CNY；D,paid,500,USD。规则：收入只统计 paid，每个 id 一次，不混加币种，缺金额需报错。指出具体问题行，按币种给出可确定金额和不能确定的部分，并建议程序校验规则。","reasoning":{"effort":"medium"},"max_output_tokens":4096,"store":false}' \
  -o response.json &&
python3 - <<'PY'
import json
with open("response.json", encoding="utf-8") as file:
    result = json.load(file)
print("status:", result.get("status"), "usage:", result.get("usage"))
if result.get("status") != "completed":
    raise SystemExit(result.get("incomplete_details") or result.get("error") or "请求未完成")
texts = [part["text"] for item in result.get("output", []) if item.get("type") == "message" for part in item.get("content", []) if part.get("type") == "output_text"]
if not texts:
    raise SystemExit("响应没有可见文本，请检查 output 中的拒绝或工具调用")
print("\n".join(texts))
PY

import base64
import os
from pathlib import Path
from openai import OpenAI

image = base64.b64encode(Path("checkout.png").read_bytes()).decode("ascii")
client = OpenAI(api_key=os.environ["tokenba_API_KEY"], base_url="https://api.apizt.com/v1")
response = client.responses.create(
    model="gpt-5.4",
    input=[{"role": "user", "content": [
        {"type": "input_text", "text": "审查这张结算页面截图。列出可见的文本层级、按钮辨识度和表单提示问题，按影响排序并标注截图位置。不要推断隐藏交互，也不要声称测出了 WCAG 对比度数值。"},
        {"type": "input_image", "image_url": "data:image/png;base64," + image, "detail": "high"},
    ]}],
    reasoning={"effort": "medium"},
    max_output_tokens=4096,
    store=False,
)
print("status:", response.status, "usage:", response.usage)
if response.status != "completed" or not response.output_text:
    raise RuntimeError("分析未完成，检查 incomplete_details、error 或拒绝")
print(response.output_text)