@drobs/ai-api

AI API

This is a generic API for calling different Generative AI models (OpenAI, VertexAI, Google, ...). The API could be invoked with some text, images or video prompt to get text or strem. It implements a Node.js module to be imported in your project

Available models

Google

Vertex AI - Recommended

• gemini-2.0-flash

Procédure

Get the following values from the Google credential JSON file by the procedure Get the complete procedure from here

Generative API

Simpler but only for test or non industrial projects

• gemini-1.5-flash

Procédure

Get your API key following this procedure

Open AI

• gpt-3.5
• gpt-3.5-turbo
• gpt-4o-mini

Installation

Add in your project

npm install @drobs/ai-api

Add the environement variables

Google API (Vertex API)

• GOOGLE_API_PROJECT_ID=Your Project ID (ex : the-experience-xxxxx)
• GOOGLE_API_PRIVATE_KEY_ID=Your Private Key ID (ex : 87adbdad1eb9b1540edf867bd8d79926eb2ad5c8)
• GOOGLE_API_PRIVATE_KEY=Your Private Key (ex : -----BEGIN PRIVATE KEY-----\nMIIEvgIBADANBgkq...oq4SQJbV\n-----END PRIVATE KEY-----\n)
• GOOGLE_API_CLIENT_EMAIL=Your Google Account allowed to use VertexAI API (ex : [e-mail])
• GOOGLE_API_CLIENT_ID=Your Google Client ID (ex : 107696212857897829246)
• GOOGLE_API_REGION=Your Google Region (ex : europe-west9)
• GOOGLE_API_CLIENT_CERT_URL=https://www.googleapis.com/...

Google API (Generative API)

• GOOGLE_API_PRIVATE_KEY=Your Private Key

Open API

• OPENAI_API_KEY=Your Open API Key (ex : sk-tfB0mBpLPyQFeUzR7R48issEmHrjNbzCYsT3BlbkFJev3ZGj)

Usage

Procedure

Import functions from @drobs/ai-api module

import { generateContent, getModels } from './index.js';

Choose your model from getModels

(Optional) Choose options for your request

const options = {
   temperature: 0, //from 0 (very deterministic) to 2 (very creative)
   max_tokens: 100, //number of tken for the expected response
   topK: 64, //probability of tokens to consider for each candidate (1=most probable token) (only for Google models)
   topP: 1, //cumulative probability of tokens to consider for each candidate
   candidateCount: 1 //number of candidate responses that will be return by the model
   moderationModel:{} //soon available
}

Create your prompt request

const prompt = 'Tell me a story';

Invoke generateContent with the chosen model

const response = generateContent (prompt, model, false, options);
console.log (response);

Example

import { generateContent } from '@drobs/ai-api';

const prompt = "This is a test";
const model = 4;  //Gemini 1.5 Flash via Vertex AI
const stream = false;
const options = {
    temperature: 0,
    max_tokens: 100,
    topP: 1,
    candidateCount: 1
}

function main () {
    generateContent (prompt, model, stream, options)
    .then (result => {
        console.log(result);
    })
    .catch (error => {
        console.error(error);
    }); 
}

main();

Contribute

GitHub Repository

Please feel free to contribute or report any issue on GitHub

Compile

npm run build

Test

node run test