Files
simple-ai-provider/README.md

487 lines
12 KiB
Markdown

# Simple AI Provider
A professional, extensible TypeScript package for integrating multiple AI providers into your applications with a unified interface. Currently supports **Claude (Anthropic)**, **OpenAI (GPT)**, and **Google Gemini** with plans to add more providers.
## Features
- 🎯 **Unified Interface**: Same API across all AI providers
- 🔒 **Type-Safe**: Full TypeScript support with comprehensive type definitions
- 🚀 **Easy to Use**: Simple factory functions and intuitive configuration
- 📡 **Streaming Support**: Real-time streaming responses where supported
- 🛡️ **Error Handling**: Robust error handling with categorized error types
- 🔧 **Extensible**: Easy to add new AI providers
- 📦 **Modern**: Built with ES modules and modern JavaScript features
- 🌐 **Multi-Provider**: Switch between Claude, OpenAI, and Gemini seamlessly
## Installation
```bash
npm install simple-ai-provider
# or
yarn add simple-ai-provider
# or
bun add simple-ai-provider
```
## Quick Start
### Basic Usage with Claude
```typescript
import { createClaudeProvider } from 'simple-ai-provider';
// Create a Claude provider
const claude = createClaudeProvider('your-anthropic-api-key');
// Initialize the provider
await claude.initialize();
// Generate a completion
const response = await claude.complete({
messages: [
{ role: 'user', content: 'Hello! How are you today?' }
],
maxTokens: 100,
temperature: 0.7
});
console.log(response.content);
```
### Basic Usage with OpenAI
```typescript
import { createOpenAIProvider } from 'simple-ai-provider';
// Create an OpenAI provider
const openai = createOpenAIProvider('your-openai-api-key');
// Initialize the provider
await openai.initialize();
// Generate a completion
const response = await openai.complete({
messages: [
{ role: 'user', content: 'Hello! How are you today?' }
],
maxTokens: 100,
temperature: 0.7
});
console.log(response.content);
```
### Basic Usage with Gemini
```typescript
import { createGeminiProvider } from 'simple-ai-provider';
// Create a Gemini provider
const gemini = createGeminiProvider('your-google-ai-api-key');
// Initialize the provider
await gemini.initialize();
// Generate a completion
const response = await gemini.complete({
messages: [
{ role: 'user', content: 'Hello! How are you today?' }
],
maxTokens: 100,
temperature: 0.7
});
console.log(response.content);
```
### Multi-Provider Usage
```typescript
import { createProvider, createClaudeProvider, createOpenAIProvider, createGeminiProvider } from 'simple-ai-provider';
// Method 1: Using specific factory functions
const claude = createClaudeProvider('your-anthropic-api-key');
const openai = createOpenAIProvider('your-openai-api-key');
const gemini = createGeminiProvider('your-google-ai-api-key');
// Method 2: Using generic factory
const claude2 = createProvider('claude', { apiKey: 'your-anthropic-api-key' });
const openai2 = createProvider('openai', { apiKey: 'your-openai-api-key' });
const gemini2 = createProvider('gemini', { apiKey: 'your-google-ai-api-key' });
// Initialize all
await Promise.all([claude.initialize(), openai.initialize(), gemini.initialize()]);
// Use the same interface for all providers
const prompt = { messages: [{ role: 'user', content: 'Explain AI' }] };
const claudeResponse = await claude.complete(prompt);
const openaiResponse = await openai.complete(prompt);
const geminiResponse = await gemini.complete(prompt);
```
### Streaming Responses
```typescript
import { createGeminiProvider } from 'simple-ai-provider';
const gemini = createGeminiProvider('your-google-ai-api-key');
await gemini.initialize();
// Stream a completion
for await (const chunk of gemini.stream({
messages: [
{ role: 'user', content: 'Write a short story about a robot.' }
],
maxTokens: 500
})) {
if (!chunk.isComplete) {
process.stdout.write(chunk.content);
} else {
console.log('\n\nUsage:', chunk.usage);
}
}
```
### Advanced Configuration
```typescript
import { ClaudeProvider, OpenAIProvider, GeminiProvider } from 'simple-ai-provider';
// Claude with custom configuration
const claude = new ClaudeProvider({
apiKey: 'your-anthropic-api-key',
defaultModel: 'claude-3-5-sonnet-20241022',
timeout: 30000,
maxRetries: 3,
baseUrl: 'https://api.anthropic.com' // optional custom endpoint
});
// OpenAI with organization and project
const openai = new OpenAIProvider({
apiKey: 'your-openai-api-key',
defaultModel: 'gpt-4',
organization: 'org-your-org-id',
project: 'proj-your-project-id',
timeout: 60000,
maxRetries: 5
});
// Gemini with safety settings and generation config
const gemini = new GeminiProvider({
apiKey: 'your-google-ai-api-key',
defaultModel: 'gemini-1.5-pro',
safetySettings: [], // Configure content filtering
generationConfig: {
temperature: 0.8,
topP: 0.9,
topK: 40,
maxOutputTokens: 2048
},
timeout: 45000
});
await Promise.all([claude.initialize(), openai.initialize(), gemini.initialize()]);
const response = await gemini.complete({
messages: [
{ role: 'system', content: 'You are a helpful assistant.' },
{ role: 'user', content: 'Explain quantum computing in simple terms.' }
],
model: 'gemini-1.5-flash',
maxTokens: 300,
temperature: 0.5,
topP: 0.9,
stopSequences: ['\n\n']
});
```
## API Reference
### Core Types
#### `AIMessage`
```typescript
interface AIMessage {
role: 'system' | 'user' | 'assistant';
content: string;
metadata?: Record<string, any>;
}
```
#### `CompletionParams`
```typescript
interface CompletionParams {
messages: AIMessage[];
model?: string;
maxTokens?: number;
temperature?: number;
topP?: number;
stopSequences?: string[];
stream?: boolean;
}
```
#### `CompletionResponse`
```typescript
interface CompletionResponse {
content: string;
model: string;
usage: TokenUsage;
id: string;
metadata?: Record<string, any>;
}
```
### Factory Functions
#### `createClaudeProvider(apiKey, options?)`
Creates a Claude provider with simplified configuration.
```typescript
const claude = createClaudeProvider('your-api-key', {
defaultModel: 'claude-3-5-sonnet-20241022',
timeout: 30000
});
```
#### `createOpenAIProvider(apiKey, options?)`
Creates an OpenAI provider with simplified configuration.
```typescript
const openai = createOpenAIProvider('your-api-key', {
defaultModel: 'gpt-4',
organization: 'org-123',
timeout: 60000
});
```
#### `createGeminiProvider(apiKey, options?)`
Creates a Gemini provider with simplified configuration.
```typescript
const gemini = createGeminiProvider('your-api-key', {
defaultModel: 'gemini-1.5-pro',
safetySettings: [],
generationConfig: {
temperature: 0.8,
topK: 40
}
});
```
#### `createProvider(type, config)`
Generic factory function for creating any provider type.
```typescript
const claude = createProvider('claude', {
apiKey: 'your-api-key',
defaultModel: 'claude-3-5-sonnet-20241022'
});
const openai = createProvider('openai', {
apiKey: 'your-api-key',
defaultModel: 'gpt-4'
});
const gemini = createProvider('gemini', {
apiKey: 'your-api-key',
defaultModel: 'gemini-1.5-flash'
});
```
### Provider Methods
#### `initialize(): Promise<void>`
Initializes the provider and validates the configuration.
#### `complete(params): Promise<CompletionResponse>`
Generates a completion based on the provided parameters.
#### `stream(params): AsyncIterable<CompletionChunk>`
Generates a streaming completion.
#### `getInfo(): ProviderInfo`
Returns information about the provider and its capabilities.
#### `isInitialized(): boolean`
Checks if the provider has been initialized.
## Error Handling
The package provides comprehensive error handling with categorized error types:
```typescript
import { AIProviderError, AIErrorType } from 'simple-ai-provider';
try {
const response = await openai.complete({
messages: [{ role: 'user', content: 'Hello!' }]
});
} catch (error) {
if (error instanceof AIProviderError) {
switch (error.type) {
case AIErrorType.AUTHENTICATION:
console.error('Invalid API key');
break;
case AIErrorType.RATE_LIMIT:
console.error('Rate limit exceeded');
break;
case AIErrorType.INVALID_REQUEST:
console.error('Invalid request parameters');
break;
default:
console.error('Unknown error:', error.message);
}
}
}
```
## Supported Models
### Claude (Anthropic)
- `claude-3-5-sonnet-20241022` (default)
- `claude-3-5-haiku-20241022`
- `claude-3-opus-20240229`
- `claude-3-sonnet-20240229`
- `claude-3-haiku-20240307`
### OpenAI (GPT)
- `gpt-4` (default)
- `gpt-4-turbo`
- `gpt-4-turbo-preview`
- `gpt-4-0125-preview`
- `gpt-4-1106-preview`
- `gpt-3.5-turbo`
- `gpt-3.5-turbo-0125`
- `gpt-3.5-turbo-1106`
### Google Gemini
- `gemini-1.5-flash` (default)
- `gemini-1.5-flash-8b`
- `gemini-1.5-pro`
- `gemini-1.0-pro`
- `gemini-1.0-pro-vision`
## Environment Variables
You can set your API keys as environment variables:
```bash
export ANTHROPIC_API_KEY="your-anthropic-api-key"
export OPENAI_API_KEY="your-openai-api-key"
export GOOGLE_AI_API_KEY="your-google-ai-api-key"
```
```typescript
const claude = createClaudeProvider(process.env.ANTHROPIC_API_KEY!);
const openai = createOpenAIProvider(process.env.OPENAI_API_KEY!);
const gemini = createGeminiProvider(process.env.GOOGLE_AI_API_KEY!);
```
## Provider Comparison
| Feature | Claude | OpenAI | Gemini |
|---------|--------|--------|--------|
| **Models** | 5 models | 8+ models | 5 models |
| **Max Context** | 200K tokens | 128K tokens | 1M tokens |
| **Streaming** | ✅ | ✅ | ✅ |
| **Vision** | ✅ | ✅ | ✅ |
| **Function Calling** | ✅ | ✅ | ✅ |
| **JSON Mode** | ❌ | ✅ | ❌ |
| **System Messages** | ✅ (separate) | ✅ (inline) | ✅ (separate) |
| **Multimodal** | ✅ | ✅ | ✅ |
| **Safety Controls** | Basic | Basic | Advanced |
| **Special Features** | Advanced reasoning | JSON mode, plugins | Largest context, advanced safety |
## Best Practices
1. **Always initialize providers** before using them
2. **Handle errors gracefully** with proper error types
3. **Use appropriate models** for your use case (speed vs. capability vs. context)
4. **Set reasonable timeouts** for your application
5. **Implement retry logic** for production applications
6. **Monitor token usage** to control costs
7. **Use environment variables** for API keys
8. **Consider provider-specific features** when choosing
9. **Configure safety settings** appropriately for Gemini
10. **Leverage large context** capabilities of Gemini for complex tasks
## Advanced Usage
### Provider Registry
```typescript
import { ProviderRegistry } from 'simple-ai-provider';
// List all registered providers
console.log(ProviderRegistry.getRegisteredProviders()); // ['claude', 'openai', 'gemini']
// Create provider by name
const provider = ProviderRegistry.create('gemini', {
apiKey: 'your-api-key'
});
// Check if provider is registered
if (ProviderRegistry.isRegistered('gemini')) {
console.log('Gemini is available!');
}
```
### Gemini-Specific Features
```typescript
import { createGeminiProvider } from 'simple-ai-provider';
const gemini = createGeminiProvider('your-api-key', {
defaultModel: 'gemini-1.5-pro',
safetySettings: [
{
category: 'HARM_CATEGORY_HARASSMENT',
threshold: 'BLOCK_MEDIUM_AND_ABOVE'
}
],
generationConfig: {
temperature: 0.9,
topP: 0.8,
topK: 40,
maxOutputTokens: 2048,
stopSequences: ['END', 'STOP']
}
});
await gemini.initialize();
// Large context example (up to 1M tokens)
const response = await gemini.complete({
messages: [
{ role: 'system', content: 'You are analyzing a large document.' },
{ role: 'user', content: 'Your very large text here...' }
],
maxTokens: 2048
});
```
## Contributing
Contributions are welcome! Please feel free to submit a Pull Request. For major changes, please open an issue first to discuss what you would like to change.
## License
MIT
## Changelog
### 1.0.0
- Initial release
- Claude provider implementation
- OpenAI provider implementation
- Gemini provider implementation
- Streaming support for all providers
- Comprehensive error handling
- TypeScript support
- Provider registry system
- Multi-provider examples
- Large context support (Gemini)
- Advanced safety controls (Gemini)