ExisJS

OpenAI Integration

ExisJS provides a zero-config, native integration for OpenAI to instantly harness the power of LLMs.

Following our framework philosophy, the openai SDK is an optional peer dependency. It remains completely invisible and uninstalled until you explicitly need it, ensuring your core application stays blazing fast and lightweight!

Environment Variables

The OpenAI plugin securely reads your API key directly from your .env file.

Before using the plugin, ensure you have your API key set:

.env
OPENAI_API_KEY="sk-proj-123456789"

Installation

ExisJS wraps the official openai Node.js SDK.

npm install openai

Usage

Simply import the openai proxy from exisjs/openai. ExisJS will instantly boot and cache the client connection the very first time you generate a completion!

Functional Paradigm

src/http/ai/route.ts
import { controller, route } from 'exisjs/router'import { openai } from 'exisjs/openai'
export default controller({  generateJoke: route.post('/joke', {    async handle(req) {      // The openai client magically lazy-loads here!      const completion = await openai.chat.completions.create({        messages: [{ role: 'user', content: 'Tell me a programming joke.' }],        model: 'gpt-4o-mini',      })
      return {         success: true,         joke: completion.choices[0].message.content       }    }  })})

Class-Based (OOP) Paradigm

src/http/ai/route.ts
import { Controller, Post } from 'exisjs/decorators'import { openai } from 'exisjs/openai'
@Controller('/ai')export default class AIController {    @Post('/joke')  async generateJoke() {    // The openai client magically lazy-loads here!    const completion = await openai.chat.completions.create({      messages: [{ role: 'user', content: 'Tell me a programming joke.' }],      model: 'gpt-4o-mini',    })
    return {       success: true,       joke: completion.choices[0].message.content     }  }}

Customizing Connection Options

If you need to pass custom options to the OpenAI Client (e.g., overriding the baseURL to use an OpenAI-compatible open-source model via Ollama or vLLM), call configureOpenAI(options) inside your server.ts onStart hook!

src/http/server.ts
import { exis } from 'exisjs'import { configureOpenAI } from 'exisjs/openai'
export default exis({  async onStart() {    // Override the baseURL to point to a local Ollama instance!    configureOpenAI({      baseURL: 'http://localhost:11434/v1',      apiKey: 'ollama' // Local LLMs often just need a dummy key    })  }})
Continue to PostHogEdit this page on GitHub