LLM Backend SDK

LLM Backend SDK is a library that simplifies integration of Large Language Models (LLMs) into your backend services.

Change Log

| Version | Changes | |---------|---------| | v0.0.1 | API for fetching models, chat completion/streaming, and image generation |

Features

• getModels: Retrieves a list of available language models from configured providers with their capabilities and metadata
• getCompletion: Sends prompts to LLMs and returns full text responses, supporting various parameters for controlling generation
• streamCompletion: Delivers LLM responses as they're being generated in real-time, enabling progressive rendering in client applications
• generateImage: Creates images from text prompts using AI image generation models, with customizable parameters for style and quality

Installation

npm install @innobridge/llmbackendsdk

In you backend application

import { api } from "@innobridge/llmbackendsdk";


const models = await api.getModels(<baseUrl>, <apiKey>);

Database

Installation

Install postgresql

npm install pg @types/pg 

in your application's main file where you start your express server.

i.e. index.js

import { databaseApi, databaseConfiguration } from '@innobridge/llmbackendsdk';
const app = express();
...

const config = {
    connectionString: <your database url>
} as databaseConfiguration.PostgresConfiguration;

if (<your env is production>) {
    config.ssl = {
        rejectUnauthorized: true
    };
}

initializeDatabase(
    config
).then(() => {
    const server = app.listen(+port, () => {
        console.log(`Listening on http://localhost:${port}`);
        console.log(`Swagger UI: http://localhost:${port}/docs`);
    });
  
    // Handle server shutdown
    const gracefulShutdown = async () => {
      console.log('Shutting down server...');
      server.close(async () => {
        console.log('Server closed. Closing database...');
        await shutdownDatabase();
        process.exit(0);
      });
    };
  
    process.on('SIGTERM', gracefulShutdown);
    process.on('SIGINT', gracefulShutdown);
  });

Running migrations

in index.js

import { databaseApi, databaseConfiguration } from '@innobridge/llmbackendsdk';
import { PoolClient } from 'pg';

...
const config = {
  connectionString: <your database url>
} as databaseConfiguration.PostgresConfiguration;

if (<your env is production>) {
  config.ssl = {
      rejectUnauthorized: true
  };
}

const migration: Map<number, (client: PoolClient) => Promise<void>> = new Map();
migration.set(1, async (client: PoolClient) => {
  await client.query(`
      CREATE TABLE IF NOT EXISTS users (
          id SERIAL PRIMARY KEY,
          title VARCHAR(255) NOT NULL,
          user_id VARCHAR(255) NOT NULL,
          updated_at TIMESTAMP DEFAULT CURRENT_TIMESTAMP,
          deleted_at TIMESTAMP
      );
  `);
});
migration.set(2, async (client: PoolClient) => {
  await client.query(`
      ALTER TABLE users 
      ADD COLUMN username VARCHAR(255);
  `);
});

initializeDatabase(
  config,
  migration
).then(() => {
  const server = app.listen(+port, () => {
      console.log(`Listening on http://localhost:${port}`);
      console.log(`Swagger UI: http://localhost:${port}/docs`);
  });

  // Handle server shutdown
  const gracefulShutdown = async () => {
    console.log('Shutting down server...');
    server.close(async () => {
      console.log('Server closed. Closing database...');
      await shutdownDatabase();
      process.exit(0);
    });
  };

  process.on('SIGTERM', gracefulShutdown);
  process.on('SIGINT', gracefulShutdown);
});