API ガイド

5 分で始める

1. 登録とチャージ

登録・ログイン後、ウォレットで金額と支払方法を選択します。モデル利用はウォレット決済で、入金は支払確認後に反映されます。

2. API キーを作成・保存

コンソールでキーを作成し、必要に応じて月額上限を設定します。キー全体は一度だけ表示されるため安全に保存し、アプリごとに分け、ブラウザーコードや公開リポジトリに含めないでください。

3. モデルとプロトコルを選択

最新モデル一覧の公開名を使います。OpenAI 互換は /v1、Anthropic 互換は /anthropic。プロバイダー内部 ID を公開名の代わりに使わないでください。

4. リクエストを送信

SDK は Base URL にパスを追加しますが、curl は完全な URL が必要です。キーはサーバー環境変数に保存し、認証ヘッダーを送信します。ストリーミングは stream: true です。

5. 使用量と費用を確認

使用量ページでモデル・キー別費用、最近のリクエスト、状態、遅延を確認し CSV 出力できます。ウォレットは人民元決済です。単価の単位はモデル一覧を参照してください。

API キーモデル一覧

コード内のフィールド名、モデル名、API パスは開発用に原文のままです。

完全な技術リファレンス（英語）

Getting started

5-minute quickstart

Five steps from a fresh account to a working LLM call. If you already have a key, jump straight to Use your API key.

Step 1 — Create an account

Go to Sign up.
Enter your email + a password. We send an activation link — click it to switch the account from PENDING_VERIFICATION to ACTIVE.
You land on the developer console.

Already activated? Sign in at /sign-in.

Step 2 — Add funds

Every relay call settles against your wallet, so it starts at zero balance and you need to top up before it works.

Open Billing.
Pick an amount and a payment method (Alipay if configured, or the mock provider in dev environments).
The wallet credit lands when the payment callback fires (seconds for Alipay, instant for mock).

Charges are denominated in CN¥ minor (分). Each model alias has a published per-1k-token price — see Models.

Step 3 — Create an API key

Open API keys.
Pick a name (e.g. local-dev) and an optional monthly spend cap.
Click Create API key. The reveal banner shows the token starting with tsk_. Copy it now — the platform never displays the full token again.

Step 4 — Send a request

Two URL forms come up depending on the client you use:

Base URL (the value you pass to an SDK's baseURL / base_url field): https://chinzy.com/v1
Full endpoint URL (the URL you put in raw HTTP / curl): https://chinzy.com/v1/chat/completions

SDKs (OpenAI, Anthropic, etc.) only need the base URL — they append /chat/completions themselves. Curl needs the full path because it has no concept of a base URL. Quick smoke test with curl:

curl https://chinzy.com/v1/chat/completions \
  -H "Authorization: Bearer tsk_..." \
  -H "Content-Type: application/json" \
  -d '{
    "model": "gpt-5.2",
    "messages": [{"role": "user", "content": "Say hi in one sentence"}]
  }'

Expected response shape: a JSON body with choices[0].message.content populated and a usage block. Status 200.

Got a different status code? Skip to the error table.

Step 5 — Watch your spend

Open Usage. Within seconds your first call shows up:

Cycle spend — total charged this calendar month.
By model / By API key — breakdowns sorted by spend.
Recent calls — last 50 requests with status, latency, and cost. The status code links to the request log row in the operator console for debugging if needed.
Download CSV — pull 7 / 30 / 90 days of activity into a spreadsheet for accounting.

You're live

From here, common next steps:

Switch your existing OpenAI / Anthropic SDK code to point at our base URL — see SDK examples.
Browse the full Models catalog for aliases and pricing.
Set monthly caps on each key so a runaway client can't drain the wallet.