从第一步，到你的下一步。From your first request to what's next.

介绍Introduction

在一个工作空间管理账户、余额、密钥和用量。工具与应用通过对应协议访问服务；选择的接入分组决定可用模型、权限和计费规则。Manage your account, balance, keys and usage in one workspace. Tools and applications connect using their supported protocols. Your access group determines available models, permissions and billing rules.

新手可以先读快速开始，已有客户端的用户可直接进入Start with the quick start guide, or connect an existing client in 教程中心Tutorials。接入线路在. Access routes are available in 工作空间Workspace中查看。.

快速开始Quick start

1. 进入工作空间Open workspace。新用户可以在本站开放注册时创建账户。. New users can create an account when registration is open.
2. 在In API Keys 中创建密钥，选择与你的工具匹配的接入分组。, create a key and choose an access group that matches your tool.
3. 查看Read the 工具接入指南tool setup guide，填写接入地址和 API Key。 and enter your endpoint and API key.
4. 完成首次调用后，在After your first request, open 用量明细Usage中查看记录和实际消费。 to review the request and its actual cost.

密钥与权限Keys & permissions

需要轮换密钥时，先创建新密钥并更新工具配置，再删除旧密钥。To rotate a key, create a new one and update your tool before deleting the old key.

接入地址Endpoints

查看地区线路与配置教程View regional routes and setup guides →

工具与 CLI 快速上手Tools & CLI quick start

原有 9 类工具指南都在教程中心，包含 Claude Code、Codex、Gemini CLI、CC Switch、Cursor、Continue.dev、Aider、LobeChat 和 Chatbox。All nine tool guides are in Tutorials: Claude Code, Codex, Gemini CLI, CC Switch, Cursor, Continue.dev, Aider, LobeChat and Chatbox.

调用示例API examples

先将 API Key 保存到当前终端的环境变量。将示例中的占位符替换为你的真实密钥和模型 ID。Save your API key as an environment variable in the current terminal. Replace placeholders with your key and model ID.

macOS / Linux

export PORTAL_API_KEY="YOUR_API_KEY"

Windows PowerShell

$env:PORTAL_API_KEY = "YOUR_API_KEY"

以下 cURL 命令使用 Bash / Zsh 语法。Windows 可用 WSL / Git Bash，或运行后面的 Python 示例；PowerShell 设置的环境变量可供同一终端启动的 Python 使用。The cURL examples use Bash / Zsh syntax. On Windows, use WSL / Git Bash or the Python example. Python launched from the same PowerShell session inherits its environment variables.

OpenAI Responses · cURL

curl "https://newapi.world/v1/responses" \
  -H "Authorization: Bearer $PORTAL_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model":"YOUR_MODEL_ID","input":"Hello!"}'

Python SDK

使用Run pip install openai 安装 SDK 后运行： to install the SDK, then run:

import os
from openai import OpenAI

client = OpenAI(
    api_key=os.environ["PORTAL_API_KEY"],
    base_url="https://newapi.world/v1",
)
response = client.responses.create(
    model="YOUR_MODEL_ID",
    input="Hello!",
)
print(response.output_text)

Anthropic Messages · cURL

curl "https://newapi.world/v1/messages" \
  -H "Authorization: Bearer $PORTAL_API_KEY" \
  -H "anthropic-version: 2023-06-01" \
  -H "Content-Type: application/json" \
  -d '{
    "model":"YOUR_MODEL_ID",
    "max_tokens":256,
    "messages":[{"role":"user","content":"Hello!"}]
  }'

基础对话 · Chat CompletionsBasic chat · Chat Completions

以下示例适用于已开放 Chat Completions 协议的分组。先运行基础示例创建These examples require a group that supports Chat Completions. Run the basic example to create client，再按需运行后续片段；Responses 和 Anthropic 分组请使用上方原生协议示例。, then try the snippets below. For Responses or Anthropic groups, use the native protocol examples above.

import os
from openai import OpenAI

client = OpenAI(
    api_key=os.environ["PORTAL_API_KEY"],
    base_url="https://newapi.world/v1",
)
response = client.chat.completions.create(
    model="YOUR_MODEL_ID",
    messages=[{"role": "user", "content": "Hello!"}],
)
print(response.choices[0].message.content)

流式响应 · StreamingStreaming responses

增量读取输出，适合交互式对话。SDK 会处理事件流，空事件无需输出。Read output as it arrives for interactive conversations. The SDK handles the event stream; skip empty events.

stream = client.chat.completions.create(
    model="YOUR_MODEL_ID",
    messages=[{"role": "user", "content": "Write a short poem."}],
    stream=True,
)
for chunk in stream:
    if chunk.choices:
        print(chunk.choices[0].delta.content or "", end="", flush=True)

多模态 · 图片输入Multimodal · Image input

选择支持图片输入的模型，并填写可访问的图片地址。图片大小、格式和计费由模型与分组规则决定。Choose a model that supports image input and provide an accessible image URL. Size, format and billing follow the model and group rules.

response = client.chat.completions.create(
    model="YOUR_VISION_MODEL_ID",
    messages=[{
        "role": "user",
        "content": [
            {"type": "text", "text": "Describe this picture."},
            {"type": "image_url", "image_url": {"url": "YOUR_IMAGE_URL"}},
        ],
    }],
)
print(response.choices[0].message.content)

函数调用 · Tool UseFunction calling · Tool use

模型返回工具调用后，由你的应用验证参数、执行工具，再传回结果。工具不会由门户自动执行。下面的天气结果是示例数据，需要替换为实际服务。When the model requests a tool call, your application validates the arguments, runs the tool and returns the result. The portal does not execute tools. Replace the sample weather data below with a real service.

import json

messages = [{"role": "user", "content": "Weather in Shanghai?"}]
tools = [{
    "type": "function",
    "function": {
        "name": "get_weather",
        "description": "Get the weather for a city",
        "parameters": {
            "type": "object",
            "properties": {"city": {"type": "string"}},
            "required": ["city"],
        },
    },
}]
response = client.chat.completions.create(
    model="YOUR_MODEL_ID", messages=messages, tools=tools,
)
message = response.choices[0].message
messages.append(message.model_dump(exclude_none=True))
for call in message.tool_calls or []:
    if call.function.name != "get_weather":
        raise ValueError("Unexpected tool")
    arguments = json.loads(call.function.arguments)
    # Replace this example result with your actual weather service.
    result = {"city": arguments["city"], "status": "example_only"}
    messages.append({
        "role": "tool", "tool_call_id": call.id,
        "content": json.dumps(result),
    })
answer = client.chat.completions.create(
    model="YOUR_MODEL_ID", messages=messages,
)
print(answer.choices[0].message.content)

模型选择Choosing a model

在In 模型与价格Models & pricing中确认可用模型 ID，再检查你的密钥分组支持哪种协议。优先根据任务复杂度、响应速度、上下文长度和价格选择。, confirm the model ID and check your key group's protocol. Choose based on task complexity, response speed, context length and price.

• 日常对话与摘要：先用较低成本模型验证质量。Everyday chat and summaries: test quality with a lower-cost model first.
• 复杂代码与推理：比较任务成功率、延迟和总消费。Complex code and reasoning: compare task success, latency and total cost.
• 图片任务：确认模型支持图片输入，再参考多模态示例。Image tasks: confirm image input support, then use the multimodal example.
• 工具调用与 Agent：确认模型和分组均支持工具协议。Tools and agents: check that both the model and group support tool calls.

高级技巧Advanced usage

故障切换 · FallbackFallback models

在应用中准备备选模型，并确认协议和权限一致。平台是否自动切换上游由实际分组配置决定；当前门户不提供自动容灾设置。对可能已经执行或扣费的请求，先确认结果再重试。Prepare fallback models in your application and check compatible protocols and permissions. Automatic upstream failover depends on group configuration; the portal does not configure it. Check the outcome before retrying requests that may already have run or incurred charges.

Prompt Caching

稳定的系统提示和公共上下文适合放在请求前部。支持显式缓存的 Anthropic 分组可以按对应协议设置Place stable system prompts and shared context early in the request. Anthropic groups with explicit caching support can use cache_control；其他模型的缓存行为可能自动处理。命中情况与费用以用量记录为准。; other models may handle caching automatically. Check usage records for cache hits and charges.

超时与重试Timeouts & retries

• 给推理任务预留足够超时，并为流式连接设置合理的空闲时间限制。Allow enough time for inference and set an appropriate idle timeout for streaming connections.
• 遇到 429 时按服务端重试提示或指数退避处理。For 429 responses, follow server retry guidance or use exponential backoff.
• 401 / 403 优先检查密钥、权限和分组，避免重复发送同一错误请求。For 401 or 403 responses, check your key, permissions and group before retrying.
• 5xx 与网络中断先判断请求是否已被执行；有副作用的工具操作需要去重。After 5xx errors or network interruptions, check whether the request ran. Deduplicate tool actions with side effects.

用量与计费Usage & billing

账户总览Account overview展示余额、今日调用与实际消费。 shows your balance, today's requests and actual charges. 用量明细Usage支持按日期与 Key ID 查询使用日志，包括模型、输入和输出 Tokens、缓存读写、缓存命中率、首 Token 时间和实际消费。 supports filtering by date and key ID, with model, input/output tokens, cache reads/writes, cache hit rate, time to first token, and actual cost.

模型页面的官方参考价用于比较。实际账单会结合接入分组、专属倍率、缓存、订阅和其他平台规则计算。Official model prices are provided for comparison. Actual billing includes group rates, custom multipliers, caching, subscriptions and other platform rules.

你可以在Open 充值与余额Top up & balance中查看余额、使用兑换码，以及在平台启用时在线充值和查询订单。 to view your balance, redeem codes, and, when enabled, top up online and check orders.

真实性验证Authenticity verification

带有 TEE 证明的响应可以交给开源验证器核对来源和完整性。普通响应中的模型名或模型自述，不能单独证明真实来源。An open-source verifier can check the provenance and integrity of responses carrying TEE evidence. A model name or self-description alone does not establish provenance.

查看第三方验证入口、材料要求与验证边界View verification tools, required evidence, and limitations →

常见问题Frequently asked questions