SDKs and frameworks

Use the official OpenAI and Anthropic SDKs for Raon's supported text generation features. Set the base URL, API key and model ID, then check the supported features and limits when migrating existing code. Matching an API format does not imply support for every feature of its original provider.

OpenAI SDKCopy Anchor Link

# pip install openai
import os
from openai import OpenAI

client = OpenAI(
    base_url="https://raon-api.krafton.ai/v1",
    api_key=os.environ["RAON_API_KEY"],
)

The same client calls both Chat Completions and Responses.

Anthropic SDKCopy Anchor Link

# pip install anthropic
import os
from anthropic import Anthropic

client = Anthropic(
    base_url="https://raon-api.krafton.ai",
    api_key=os.environ["RAON_API_KEY"],
)

This client calls Messages.

LangChainCopy Anchor Link

ChatOpenAI takes Raon's base URL:

# pip install langchain-openai
import os
from langchain_openai import ChatOpenAI

llm = ChatOpenAI(
    model="raon-1-mini-preview",
    base_url="https://raon-api.krafton.ai/v1",
    api_key=os.environ["RAON_API_KEY"],
)
print(llm.invoke("Hello!").content)

Vercel AI SDKCopy Anchor Link

Use the OpenAI-compatible provider. It returns the model's reasoning as reasoningText:

// npm install ai @ai-sdk/openai-compatible
import { createOpenAICompatible } from "@ai-sdk/openai-compatible";
import { generateText } from "ai";

const raon = createOpenAICompatible({
  name: "raon",
  baseURL: "https://raon-api.krafton.ai/v1",
  apiKey: process.env.RAON_API_KEY,
});

const { text } = await generateText({
  model: raon("raon-1-mini-preview"),
  prompt: "Hello!",
});
console.log(text);

Moving code from OpenAI or AnthropicCopy Anchor Link

• Model ID. Use raon-1-mini-preview. Other providers' model names, such as gpt-4o or claude-sonnet-4-5, return 403 model_not_allowed.
• Parameters. Raon fixes temperature=1.0, top_p=0.95, top_k=20, presence_penalty=1.5, repetition_penalty=1.0, frequency_penalty=0 and min_p=0 for all generation APIs to keep responses stable. These values apply even when you omit the parameters, send null or supply different values. Other parameters such as stop, seed and response_format are passed to the model as they are. Effort settings are accepted but have no effect; see Reasoning.
• No server-side state. store is always off. Send the whole conversation on every request; see What is refused.
• Tools. Function tools work on all three APIs. Of the tools a provider runs for you, only web search is available.
• Text only. The model does not read images, audio or files.