Reference

SDKs & frameworks

No ThaiRouter SDK to install. Anything that speaks OpenAI works.

Every integration below is the stock library with two settings changed: the base URL and the key. Model ids are the ThaiRouter ids from the catalog.

OpenAI Python

pip install openai

from openai import OpenAI

client = OpenAI(base_url="https://api.thairouter.ai/v1", api_key="sk-tr-...")

r = client.chat.completions.create(
    model="thairouter/glm-5.3-flash",
    messages=[{"role": "user", "content": "สวัสดี"}],
)
print(r.choices[0].message.content)
print(r.model_extra["thairouter"])   # {'usage_id': '…', 'cost': 0.0007}

OpenAI Node

npm i openai

import OpenAI from "openai";

const client = new OpenAI({ baseURL: "https://api.thairouter.ai/v1", apiKey: process.env.THAIROUTER_KEY });

const r = await client.chat.completions.create({
  model: "thairouter/glm-5.3-flash",
  messages: [{ role: "user", content: "สวัสดี" }],
});
console.log(r.choices[0].message.content);
console.log((r as unknown as { thairouter: { usage_id: string; cost: number } }).thairouter);

Vercel AI SDK

Use the OpenAI-compatible provider. Streaming, useChat and route handlers work unchanged.

npm i ai @ai-sdk/openai-compatible

import { createOpenAICompatible } from "@ai-sdk/openai-compatible";
import { streamText } from "ai";

const thairouter = createOpenAICompatible({
  name: "thairouter",
  baseURL: "https://api.thairouter.ai/v1",
  apiKey: process.env.THAIROUTER_KEY,
});

const result = streamText({
  model: thairouter("thairouter/glm-5.3-flash"),
  prompt: "สรุปข้อดีของ connection pooling",
});
for await (const chunk of result.textStream) process.stdout.write(chunk);

LangChain

Python · pip install langchain-openai

from langchain_openai import ChatOpenAI

llm = ChatOpenAI(
    model="thairouter/glm-5.3-flash",
    base_url="https://api.thairouter.ai/v1",
    api_key="sk-tr-...",
    max_tokens=1024,
)
print(llm.invoke("สวัสดี").content)

TypeScript · npm i @langchain/openai

import { ChatOpenAI } from "@langchain/openai";

const llm = new ChatOpenAI({
  model: "thairouter/glm-5.3-flash",
  apiKey: process.env.THAIROUTER_KEY,
  configuration: { baseURL: "https://api.thairouter.ai/v1" },
});
const res = await llm.invoke("สวัสดี");
console.log(res.content);

LiteLLM

Prefix the model with openai/ so LiteLLM uses the OpenAI adapter, and pass the base URL.

pip install litellm

import litellm

r = litellm.completion(
    model="openai/thairouter/glm-5.3-flash",
    api_base="https://api.thairouter.ai/v1",
    api_key="sk-tr-...",
    messages=[{"role": "user", "content": "สวัสดี"}],
)
print(r.choices[0].message.content)

Environment variables

The OpenAI SDKs read these automatically, so existing code can switch with no edits at all.

export OPENAI_BASE_URL="https://api.thairouter.ai/v1"
export OPENAI_API_KEY="sk-tr-..."

Tools that only accept OPENAI_API_KEY and a base URL (Cursor, Continue, Open WebUI, aider, most CLI agents) work the same way. If one refuses unknown model ids, add thairouter/glm-5.3-flash to its model list.