AI Router Library for Golang

This repository provides a Golang library for interacting with various AI models such as OpenAI (GPT), Google Gemini, and AWS Bedrock. The library supports both text completion and multimodal image recognition.

Installation

Ensure you have Go installed (1.18+ recommended). To use this library, install the required dependencies:

go get github.com/NaheedRayan/[email protected]

Also, install godotenv for environment variable management:

go get github.com/joho/godotenv

Environment Variables

Create a .env file in your project root and add your API keys:

OPENAI_API_KEY=your_openai_key
GEMINI_API_KEY=your_gemini_key
AWS_ACCESS_KEY=your_aws_access_key
AWS_SECRET_KEY=your_aws_secret_key

Usage

Import the Library

import (
    "context"
    "github.com/NaheedRayan/openrouter-go/ai"
)

Initialize the Client

You can initialize a client for different AI providers:

OpenAI Example

client, err := ai.InitializeClient(ctx, ai.ProviderOpenAI, ai.ClientOptions{
	APIKey:  apiKey,
	ModelID: "gpt-4o-mini",
})

Gemini Example

client, err := ai.InitializeClient(ctx, ai.ProviderGemini, ai.ClientOptions{
    APIKey:  apiKey,
    ModelID: "models/gemini-2.0-flash-lite-preview-02-05",
})

AWS Bedrock Example

client, err := ai.InitializeClient(ctx, ai.ProviderBedrock, ai.ClientOptions{
    AccessKey: accessKey,
    SecretKey: secretKey,
    Region:    "us-east-1",
    ModelID:   "amazon.nova-lite-v1:0",
})

Sending Requests

Text Completion

messages := []ai.InputMessage{
    { Role: "user", Content: "Write a haiku about programming." },
}

config := ai.ModelConfig{
    Temperature:  0.7,
    TopP:         0.9,
    MaxTokens:    100,
    SystemPrompt: "You are a helpful AI assistant.",
}

response, err := client.TextCompletion(ctx, messages, config)
if err != nil {
    log.Fatalf("Error calling AI model: %v", err)
}

fmt.Println("AI Response:", response.Text)
fmt.Printf("Tokens used: %d input, %d output, %d total\n", response.TokenUsage.InputTokens, response.TokenUsage.OutputTokens, response.TokenUsage.TotalTokens)

Multimodal Image Recognition

image := ai.Image{
    Format: "jpeg",
    Data:   imageBytes,
    URL: "https://example.com/sample.jpg",
}

messages := []ai.InputMessage{
    { Role: "user", Content: "What is shown in this image?", Images: []ai.Image{image} },
}

response, err := client.ImageRecognition(ctx, messages, config)
if err != nil {
    log.Fatalf("Error calling AI model: %v", err)
}

fmt.Println("AI Response:", response.Text)
fmt.Printf("Tokens used: %d input, %d output, %d total\n", response.TokenUsage.InputTokens, response.TokenUsage.OutputTokens, response.TokenUsage.TotalTokens)

Running the Example

To run the example provided in main.go, use:

go run main.go

Expected Output

OpenAI Example Output

-----------------Testing OpenAI client--------------
AI Response:
A code poet dreams,
Logic flows in silent streams,
Night hums with debug.
Tokens used: 10 input, 15 output, 25 total

Gemini Example Output

----------------Testing Gemini client---------------
AI Response:
Patience, practice, learn,
Code with heart and mind as one,
Mastery will come.
Tokens used: 12 input, 18 output, 30 total

Multimodal Image Recognition Output

-------------Testing Gemini Multimodal client--------
AI Response:
The image shows the Palace of Westminster, a historic building in London.
Tokens used: 20 input, 25 output, 45 total

Contributing

Feel free to submit pull requests or report issues if you find any bugs or improvements.

License

This project is licensed under the MIT License.