OpenAI SDK baseURL setup for DeepSeek, Qwen, and Kimi routes

AI assistance disclosure: this article was prepared with AI assistance and reviewed before publication.

Disclosure: I work on Botat. This is a technical integration note, not a neutral vendor ranking.

If you are testing DeepSeek, Qwen, Kimi, or other Chinese-strong model routes from outside China, the easiest integration path is often not a new SDK. It is a configurable OpenAI SDK client with a different baseURL, API key, and model ID.

That still needs careful validation. OpenAI-compatible means the request shape is familiar. It does not mean every OpenAI feature behaves identically.

The minimal config shape

Keep provider settings outside application logic:

OPENAI_BASE_URL=https://your-provider.example/v1
OPENAI_API_KEY=your_api_key
OPENAI_MODEL=your_verified_model_id

Do not publish hard-coded production model IDs until you have verified the current model list inside the provider account.

Node.js example

import OpenAI from "openai";

const client = new OpenAI({
  apiKey: process.env.OPENAI_API_KEY,
  baseURL: process.env.OPENAI_BASE_URL,
});

const response = await client.chat.completions.create({
  model: process.env.OPENAI_MODEL,
  messages: [
    { role: "user", content: "Reply with one short sentence." }
  ],
  stream: false,
});

console.log(response.choices[0]?.message?.content);

Python example

from openai import OpenAI
import os

client = OpenAI(
    api_key=os.environ["OPENAI_API_KEY"],
    base_url=os.environ["OPENAI_BASE_URL"],
)

response = client.chat.completions.create(
    model=os.environ["OPENAI_MODEL"],
    messages=[{"role": "user", "content": "Reply with one short sentence."}],
    stream=False,
)

print(response.choices[0].message.content)

Smoke test before debugging wrappers

If your app fails, test the route outside your framework first:

curl "$OPENAI_BASE_URL/chat/completions" \
  -H "Authorization: Bearer $OPENAI_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model":"YOUR_MODEL_ID","messages":[{"role":"user","content":"Reply with one short sentence."}],"stream":false}'

Check these before changing application code:

whether the base URL should include /v1
whether your key can access the selected model
whether the endpoint accepts the SDK's default parameters
whether the provider supports streaming in the same shape
whether errors preserve upstream 400, 401, 403, 429, and 5xx status codes

Capability checks matter

For Chinese-strong OpenAI-compatible routes, I would test these separately:

response_format and JSON mode
tools and tool_choice
strict JSON schema
streaming deltas
sampling parameters such as top_p, stop, penalties, and seed
provider-specific reasoning fields
retry behavior and timeout behavior

A common mistake is to debug a full agent stack before checking whether a single raw chat completion works.

Where Botat fits

Botat is one OpenAI-compatible gateway for overseas developers evaluating Chinese-strong models for non-sensitive Chinese content, batch jobs, coding assistance, localization checks, and retryable tasks.

Do not use Botat or any China-model route for sensitive personal data, confidential customer data, regulated data, or strict data-residency workloads.

I wrote a longer checklist here: https://botat.com/en/openai-sdk-base-url-deepseek-qwen/?utm_source=devto&utm_medium=owned_content&utm_campaign=overseas_10_signup_sprint&utm_content=devto_openai_sdk_base_url_20260611