SDK Integration (Drop-in Replacement)

The noris API is fully OpenAI-compatible. Migrating existing applications requires minimal effort: change 3 parameters, everything else stays untouched.

OpenAI Python SDK

Before (against OpenAI):

from openai import OpenAI

client = OpenAI(
    api_key="sk-..."
)

After (against noris):

from openai import OpenAI

client = OpenAI(
    base_url="https://ai.noris.de/v1",
    api_key="YOUR_API_KEY"
)

OpenAI JavaScript SDK

Before (against OpenAI):

import OpenAI from "openai";

const client = new OpenAI({ apiKey: "sk-..." });

After (against noris):

import OpenAI from "openai";

const client = new OpenAI({
  baseURL: "https://ai.noris.de/v1",
  apiKey: "YOUR_API_KEY"
});

Anthropic SDK

noris offers a compatible endpoint that can be addressed directly with the Anthropic SDK:

from anthropic import Anthropic

client = Anthropic(
    base_url="https://ai.noris.de/v1",
    api_key="YOUR_API_KEY"
)

message = client.messages.create(
    model="vllm/release/gpt-oss-120b",
    max_tokens=1024,
    messages=[{"role": "user", "content": "Hello!"}]
)
print(message.content[0].text)

LangChain

In LangChain, it’s enough to override the base URL on the ChatOpenAI object:

Before:

from langchain_openai import ChatOpenAI

llm = ChatOpenAI(model="gpt-4o")

After:

from langchain_openai import ChatOpenAI

llm = ChatOpenAI(
    model="vllm/release/gpt-oss-120b",
    openai_api_base="https://ai.noris.de/v1",
    openai_api_key="YOUR_API_KEY"
)

LlamaIndex

In LlamaIndex, OpenAILike is configured with the noris base URL:

from llama_index.llms.openai_like import OpenAILike

llm = OpenAILike(
    model="vllm/release/gpt-oss-120b",
    api_base="https://ai.noris.de/v1",
    api_key="YOUR_API_KEY",
    is_chat_model=True
)

LiteLLM

In LiteLLM, noris is registered as a provider in the router:

model_list:
  - model_name: noris-gpt-oss
    litellm_params:
      model: vllm/release/gpt-oss-120b
      api_base: https://ai.noris.de/v1
      api_key: YOUR_API_KEY

from litellm import Router

router = Router(model_list=[
    {
        "model_name": "noris-gpt-oss",
        "litellm_params": {
            "model": "vllm/release/gpt-oss-120b",
            "api_base": "https://ai.noris.de/v1",
            "api_key": "YOUR_API_KEY"
        }
    }
])

response = router.completion(
    model="noris-gpt-oss",
    messages=[{"role": "user", "content": "Hello!"}]
)
print(response.choices[0].message.content)

---

Regardless of the SDK you choose, the same rule of thumb always applies: change 3 parameters, everything else stays untouched.