@weweb/backend-openai

WeWeb OpenAI Integration

A WeWeb backend integration for OpenAI's API, providing access to AI models for chat completions, embeddings, content moderation, and image generation within WeWeb backend workflows.

Features

• Simple integration with OpenAI API
• Support for Chat Completions (GPT-3.5/GPT-4)
• Text embeddings generation
• Content moderation
• Image generation with DALL-E

Installation

This package is designed to work with the WeWeb Supabase Backend Builder and Deno.

import { serve } from '@weweb/backend-core';
import { createOpenAIIntegration } from '@weweb/backend-openai';

Usage

Basic Setup

import type { BackendConfig } from '@weweb/backend-core';
import { serve } from '@weweb/backend-core';
import OpenAI from '@weweb/backend-openai';

// Define your backend configuration
const config: BackendConfig = {
  workflows: [
    // Your workflows here
  ],
  integrations: [
    // Use the default OpenAI integration
    OpenAI,
    // Or add other integrations
  ],
  production: false,
};

// Start the server
const server = serve(config);

Custom Configuration

You can customize the OpenAI client by using the createOpenAIIntegration function:

import { createOpenAIIntegration } from '@weweb/backend-openai';

// Create a custom OpenAI integration
const customOpenAI = createOpenAIIntegration({
  apiKey: 'your-api-key', // Override environment variable
  organization: 'your-org-id', // Optional
  baseURL: 'https://your-custom-endpoint.com', // Optional
});

If not specified, the integration will use the following environment variables:

• OPENAI_API_KEY - Your OpenAI API key
• OPENAI_ORGANIZATION - Your OpenAI organization ID (optional)
• OPENAI_BASE_URL - Custom API endpoint URL (optional)

Available Methods

Chat Completions

Generate responses from OpenAI's GPT models.

// Example workflow action
const config = {
  type: 'action',
  id: 'generate_response',
  actionId: 'openai.create_chat_completion',
  inputMapping: [
    {
      model: 'gpt-4',
      messages: [
        { role: 'system', content: 'You are a helpful assistant.' },
        { role: 'user', content: '$body.question' }
      ],
      temperature: 0.7,
      max_tokens: 500
    }
  ]
};

Embeddings

Generate vector embeddings from text for semantic search and similarity.

// Example workflow action
const config = {
  type: 'action',
  id: 'create_embedding',
  actionId: 'openai.create_embeddings',
  inputMapping: [
    {
      model: 'text-embedding-ada-002',
      input: '$body.text'
    }
  ]
};

Content Moderation

Check content for potentially harmful or sensitive material.

// Example workflow action
const config = {
  type: 'action',
  id: 'moderate_content',
  actionId: 'openai.create_moderation',
  inputMapping: [
    {
      input: '$body.text',
      model: 'text-moderation-latest'
    }
  ]
};

Image Generation

Generate images from text descriptions using DALL-E.

// Example workflow action
const config = {
  type: 'action',
  id: 'generate_image',
  actionId: 'openai.generate_image',
  inputMapping: [
    {
      prompt: '$body.description',
      model: 'dall-e-3',
      size: '1024x1024',
      quality: 'standard',
      style: 'vivid'
    }
  ]
};

Input and Output Schema

The OpenAI integration includes a detailed schema that defines all input parameters and output structures for each method. This schema is used for validation and documentation.

Chat Completion Inputs

• messages: Array of message objects with role and content
• model: ID of the model to use (e.g., gpt-4, gpt-3.5-turbo)
• temperature: Controls randomness (0-2)
• max_tokens: Maximum tokens to generate
• top_p: Alternative to temperature for nucleus sampling
• frequency_penalty: Decreases likelihood of repeating tokens
• presence_penalty: Increases likelihood of new topics
• stream: Whether to stream the response
• stop: Sequences where the API will stop generating

Embedding Inputs

• input: Text to embed (string or array of strings)
• model: Model to use (e.g., text-embedding-ada-002)
• encoding_format: Format for embeddings (float or base64)
• dimensions: Number of dimensions for embeddings

Moderation Inputs

• input: Text to moderate (string or array of strings)
• model: Moderation model to use

Image Generation Inputs

• prompt: Text description of the desired image
• model: Model to use (e.g., dall-e-3)
• n: Number of images to generate
• size: Size of the generated images
• quality: Quality of the generated images
• style: Style of the generated images
• response_format: Format to return the images (url or b64_json)

Development

Testing

deno test

Code Quality

Format and lint your code:

deno fmt
deno lint

Example: Complete Chat Application

import type { BackendConfig } from '@weweb/backend-core';
import { serve } from '@weweb/backend-core';
import OpenAI from '@weweb/backend-openai';

const config: BackendConfig = {
  workflows: [
    {
      path: '/chat',
      methods: ['POST'],
      security: {
        accessRule: 'public',
      },
      inputsValidation: {
        body: {
          type: 'object',
          properties: {
            messages: {
              type: 'array',
              items: {
                type: 'object',
                properties: {
                  role: { type: 'string' },
                  content: { type: 'string' },
                },
                required: ['role', 'content'],
              },
            },
          },
          required: ['messages'],
        },
      },
      workflow: [
        {
          type: 'action',
          id: 'chat_completion',
          actionId: 'openai.create_chat_completion',
          inputMapping: [
            {
              messages: '$body.messages',
              model: 'gpt-3.5-turbo',
              temperature: 0.7,
              max_tokens: 1000,
            },
          ],
        },
      ],
    },
  ],
  integrations: [OpenAI],
  production: false,
};

console.log('Starting OpenAI chat server on http://localhost:8000/chat');
serve(config);