SDK Integration

Documentation Index

Fetch the complete documentation index at: https://docs.openserv.ai/llms.txt

Use this file to discover all available pages before exploring further.

​

Quickstart

​

OpenAI SDK → SERV (/v1/chat/completions)

import OpenAI from "openai";

const client = new OpenAI({
  baseURL: "https://inference-api.openserv.ai/v1",  // /v1 IS required here
  apiKey: process.env.SERV_API_KEY,
});

const completion = await client.chat.completions.create({
  model: "gpt-5.4-mini",
  messages: [
    // System prompt is required, see Gotcha #3 below
    { role: "system", content: "You are a concise assistant." },
    { role: "user", content: "What is a CPU register?" },
  ],
});

console.log(completion.choices[0].message.content);

​

OpenAI SDK → SERV (/v1/responses)

import OpenAI from "openai";

const client = new OpenAI({
  baseURL: "https://inference-api.openserv.ai/v1",
  apiKey: process.env.SERV_API_KEY,
});

const response = await client.responses.create({
  model: "gpt-5.4",
  instructions: "You are a careful reasoner.",   // top-level, counts as the system prompt
  input: "What is the integral of x^2 from 0 to 3?",
});

console.log(response.output_text);

​

Anthropic SDK → SERV (/v1/messages)

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

const client = new Anthropic({
  baseURL: "https://inference-api.openserv.ai",  // NO /v1, the SDK adds it
  authToken: process.env.SERV_API_KEY,           // authToken, not apiKey
});

const message = await client.messages.create({
  model: "claude-haiku-4.5",
  max_tokens: 1024,                              // required by the Anthropic API
  system: "You answer in one sentence.",         // top-level, required by SERV
  messages: [{ role: "user", content: "What is a CPU register?" }],
});

console.log(message.content.find(b => b.type === "text").text);

​

Gotchas

​

Gotcha #1: baseURL suffix differs by SDK

​

Gotcha #2: auth field name differs

​

Gotcha #3: SERV requires a system prompt

400 A system prompt is required. Please include a system or developer message in your request.

​

Gotcha #4: reasoning_effort allowed values are SERV-specific

​

Parameter map: OpenAI ↔ Anthropic ↔ SERV

​

Available models

Gemma: the Playground displays these as “Gemma 4 31B” and “Gemma 4 26B A4B”, but the API requires the -it (instruction-tuned) suffix. Use gemma-4-31b-it and gemma-4-26b-a4b-it. Without the suffix you get 404 The model 'gemma-...' does not exist.

​

Model × endpoint compatibility matrix

1. /v1/chat/completions is the universal endpoint. If you want one code path that works across every provider in the catalog including Claude, this is it. You can use the OpenAI SDK to call Claude through SERV, no Anthropic SDK required.
2. /v1/responses is OpenAI-family only. Sending a non-OpenAI model returns: 400 The Responses API is not supported with model X. Use the Chat Completions API (POST /v1/chat/completions) instead.
3. /v1/messages is broad, not Anthropic-exclusive. It accepts OpenAI, Anthropic, xAI, Qwen, DeepSeek, and Gemma models. Useful if you have an existing Anthropic-SDK integration you don’t want to rewrite. Swap the model id to e.g. gpt-5.4-mini and keep using @anthropic-ai/sdk.

• Gemini on /v1/messages returns 502. Use /v1/chat/completions for Gemini specifically. Gemma is fine on either.
• Google’s google-genai SDK is not compatible with SERV. It speaks Gemini’s generateContent wire format, which SERV doesn’t expose. Use the OpenAI SDK against /v1/chat/completions for any Gemini or Gemma model.

​

Verifying your setup

​

See also

• SDK Migration Guide for users on Python, Vercel AI SDK, LangChain, Mastra, or raw fetch.
• Models for the full pricing and context-window catalog.
• Playground to compare models and effort levels side by side.