Adding AI Providers & Models

This guide will show you how to extend your Enterprise AI Chatbot Platform with additional AI providers and models. The platform is built on the Vercel AI SDK, making it easy to add virtually any AI provider.

Adding a New Provider

1 Install Provider SDK

First, install the AI SDK package for your desired provider:

# Example: Adding Google Gemini
pnpm add @ai-sdk/google

# Example: Adding Mistral AI
pnpm add @ai-sdk/mistral

# Example: Adding Cohere
pnpm add @ai-sdk/cohere

2 Update Environment Variables

Add the API key for your new provider to .env.local:

# Add to your existing .env.local file
GOOGLE_API_KEY=your-google-api-key-here
MISTRAL_API_KEY=your-mistral-api-key-here
COHERE_API_KEY=your-cohere-api-key-here

3 Configure Provider in Code

Edit lib/ai/providers.ts to add your new provider:

import { createOpenAI } from '@ai-sdk/openai';
import { createAnthropic } from '@ai-sdk/anthropic';
import { createGoogleGenerativeAI } from '@ai-sdk/google'; // New import
import { mistral } from '@ai-sdk/mistral'; // New import
import { xai } from '@ai-sdk/xai';

// Existing providers...
const openai = createOpenAI({
  apiKey: process.env.OPENAI_API_KEY,
  compatibility: 'strict',
});

const anthropic = createAnthropic({
  apiKey: process.env.ANTHROPIC_API_KEY,
});

// Add new providers
const google = createGoogleGenerativeAI({
  apiKey: process.env.GOOGLE_API_KEY,
});

const mistralProvider = mistral({
  apiKey: process.env.MISTRAL_API_KEY,
});

export const myProvider = customProvider({
  languageModels: {
    // Existing models...
    'openai-gpt-4.1': openai.chat('gpt-4.1'),
    'claude-3-7-sonnet': anthropic('claude-3-7-sonnet-20250219'),
    
    // Add new models
    'google-gemini-pro': google('gemini-pro'),
    'google-gemini-pro-vision': google('gemini-pro-vision'),
    'mistral-large': mistralProvider('mistral-large-latest'),
    'mistral-medium': mistralProvider('mistral-medium-latest'),
  },
  imageModels: {
    'grok-2-image': xai.image('grok-2-image'),
    // Add image models if supported
    'google-imagen': google.image('imagen-2'),
  },
});

4 Update Model Configuration

Add your new models to the model selector by editing lib/ai/models.tsx:

export const models = [
  // Existing models...
  {
    id: 'openai-gpt-4.1',
    name: 'GPT-4.1',
    provider: 'OpenAI',
    description: 'Most capable GPT-4 model',
  },
  
  // Add new models
  {
    id: 'google-gemini-pro',
    name: 'Gemini Pro',
    provider: 'Google',
    description: 'Google\'s most capable multimodal model',
  },
  {
    id: 'mistral-large',
    name: 'Mistral Large',
    provider: 'Mistral AI',
    description: 'Mistral\'s largest and most capable model',
  },
];

Provider-Specific Configuration

OpenAI Configuration

const openai = createOpenAI({
  apiKey: process.env.OPENAI_API_KEY,
  compatibility: 'strict', // For better compatibility
  baseURL: 'https://api.openai.com/v1', // Custom endpoint if needed
  organization: 'your-org-id', // Optional organization ID
});

Anthropic Configuration

const anthropic = createAnthropic({
  apiKey: process.env.ANTHROPIC_API_KEY,
  baseURL: 'https://api.anthropic.com', // Custom endpoint
});

Google Configuration

const google = createGoogleGenerativeAI({
  apiKey: process.env.GOOGLE_API_KEY,
  baseURL: 'https://generativelanguage.googleapis.com/v1beta',
});

Troubleshooting

Debugging Steps

1. Check Environment Variables:

   echo $GOOGLE_API_KEY

2. Test Provider Connection:

   npx tsx scripts/test-provider.ts

3. Check Application Logs:

   pnpm dev
   # Check console for errors

Best Practices

Security

• Never commit API keys to version control
• Use environment variables for all sensitive configuration
• Rotate API keys regularly in production
• Set up monitoring for API usage and costs

Performance

• Cache model responses when appropriate
• Set reasonable timeouts for API calls
• Implement retry logic for failed requests
• Monitor token usage to prevent budget overruns

User Experience

• Provide clear model descriptions to help users choose
• Implement graceful fallbacks when models are unavailable

Related Documentation

Getting Started

Initial setup and installation

Docker Deployment

Deploy with Docker

Scripts & Tools

Automation scripts

Getting Started Customization