SDK y frameworks

Dos URL base

Los ejemplos leen la clave de la variable de entorno NEZHAGATE_API_KEY; nunca pongas una clave directamente en el código.

SDK de OpenAI

import os
from openai import OpenAI

client = OpenAI(base_url="https://nezhagate.com/v1", api_key=os.environ["NEZHAGATE_API_KEY"])
r = client.chat.completions.create(
    model="gpt-5.5",
    messages=[{"role": "user", "content": "Hello"}],
)
print(r.choices[0].message.content)

import OpenAI from "openai";

const client = new OpenAI({ baseURL: "https://nezhagate.com/v1", apiKey: process.env.NEZHAGATE_API_KEY });
const r = await client.chat.completions.create({
  model: "gpt-5.5",
  messages: [{ role: "user", content: "Hello" }],
});
console.log(r.choices[0].message.content);

El SDK de OpenAI también lee las variables de entorno OPENAI_BASE_URL y OPENAI_API_KEY. client.responses.create(...) llama a la API Responses, que solo ofrece los modelos GPT.

SDK de Anthropic

import os
from anthropic import Anthropic

client = Anthropic(base_url="https://nezhagate.com/anthropic", api_key=os.environ["NEZHAGATE_API_KEY"])
m = client.messages.create(
    model="claude-sonnet-5",
    max_tokens=1024,
    messages=[{"role": "user", "content": "Hello"}],
)
print(next(b.text for b in m.content if b.type == "text"))

import Anthropic from "@anthropic-ai/sdk";

const client = new Anthropic({ baseURL: "https://nezhagate.com/anthropic", apiKey: process.env.NEZHAGATE_API_KEY });
const msg = await client.messages.create({
  model: "claude-sonnet-5",
  max_tokens: 1024,
  messages: [{ role: "user", content: "Hello" }],
});
const first = msg.content[0];
if (first.type === "text") console.log(first.text);

Apunta base_url a /anthropic, sin /v1. Las variables de entorno ANTHROPIC_BASE_URL y ANTHROPIC_API_KEY también funcionan.

LangChain

import os
from langchain_openai import ChatOpenAI

llm = ChatOpenAI(
    model="gpt-5.5",
    base_url="https://nezhagate.com/v1",
    api_key=os.environ["NEZHAGATE_API_KEY"],
    use_responses_api=False,
    stream_usage=True,
)
print(llm.invoke("Hello").content)

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

const llm = new ChatOpenAI({
  model: "gpt-5.5",
  apiKey: process.env.NEZHAGATE_API_KEY,
  configuration: { baseURL: "https://nezhagate.com/v1" },
  useResponsesApi: false,
});
console.log((await llm.invoke("Hello")).content);

LangChain cambia por su cuenta a la API Responses algunos modelos GPT, o las peticiones con ciertos parámetros; use_responses_api=False (useResponsesApi: false en JavaScript) lo mantiene en el endpoint de chat. Con una base_url personalizada, el streaming no incluye información de uso a menos que pongas stream_usage=True.

Vercel AI SDK

import { createOpenAI } from "@ai-sdk/openai";
import { generateText } from "ai";

const gw = createOpenAI({ baseURL: "https://nezhagate.com/v1", apiKey: process.env.NEZHAGATE_API_KEY });
const { text } = await generateText({ model: gw.chat("gpt-5.5"), prompt: "Hello" });
console.log(text);

Desde AI SDK 5, gw("model-id") usa la API Responses; gw.chat("model-id") usa el endpoint de chat y funciona con todos los modelos. Con @ai-sdk/anthropic, pon baseURL en https://nezhagate.com/anthropic/v1 (con /v1).

Errores habituales