@lwrf42/emerge-provider-openai-compat

OpenAI-compatible provider adapter for Ollama, vLLM, llama.cpp, LM Studio, OpenRouter, and any service that speaks the OpenAI chat API.

v0.1.0 — early. Real-model verified against gpt-5.4 via custom gateway. Ollama, vLLM, llama.cpp shipped but not yet verified end-to-end — see VERIFICATION.md.

Install

npm install @lwrf42/emerge-provider-openai-compat

Quick example

import { OpenAICompatProvider } from "@lwrf42/emerge-provider-openai-compat";

// Ollama running locally
const provider = new OpenAICompatProvider({
  baseURL: "http://localhost:11434/v1",
  model: "llama3.2",
  // apiKey: "not-needed-for-ollama",
});

// Custom gateway
const provider2 = new OpenAICompatProvider({
  baseURL: process.env.EMERGE_LLM_BASE_URL,
  model: process.env.EMERGE_LLM_MODEL,
  apiKey: process.env.EMERGE_LLM_API_KEY,
});

Environment variables

| Variable | Description | |---|---| | EMERGE_LLM_BASE_URL | Base URL of the compatible endpoint | | EMERGE_LLM_MODEL | Model name | | EMERGE_LLM_API_KEY | API key (empty string if not required) | | EMERGE_LLM_PROTOCOL | chat or responses (default: chat) |

Documentation

Full docs: https://github.com/LemonCANDY42/emerge