LangChain

The OpenAI series components in LangChain all support customizing base_url; simply point it to 4ALL API.

Python

from langchain_openai import ChatOpenAI

llm = ChatOpenAI(
    base_url="https://api.4allapi.com/v1",
    api_key="sk-你的密钥",
    model="gpt-5.6",
    temperature=0.7,
)
print(llm.invoke("用一句话解释 RAG").content)

The embedding component is the same:

from langchain_openai import OpenAIEmbeddings

emb = OpenAIEmbeddings(
    base_url="https://api.4allapi.com/v1",
    api_key="sk-你的密钥",
    model="text-embedding-3-large",
)

JavaScript / TypeScript

import { ChatOpenAI } from '@langchain/openai'

const llm = new ChatOpenAI({
  model: 'fable-5',
  apiKey: 'sk-你的密钥',
  configuration: { baseURL: 'https://api.4allapi.com/v1' },
})

Recommendations

• Put base_url, api_key, and model into environment variables so you can switch between environments with zero code changes;
• Different nodes in an Agent/chain can use different models: use a low-cost model for routing and summarization, and a flagship model for core reasoning.