Instructor.NET

A .NET port of the Python Instructor library for structured outputs from large language models.

Features

• Strongly typed responses from LLMs
• Built-in support for OpenAI's GPT models
• Easy integration with Azure OpenAI
• JSON schema-based response validation
• Customizable output formatting

Installation

dotnet add package Instructor.NET

Quick Start

1. First, create a model that inherits from ResponseModel:

public class UserProfile : ResponseModel
{
    public string Name { get; set; } = string.Empty;
    public int Age { get; set; }
    public string Email { get; set; } = string.Empty;

    [JsonPropertyName("preferred_language")]
    public string PreferredLanguage { get; set; } = string.Empty;
}

2. Initialize the client with your API key:

var client = new InstructorClient("your-api-key-here");

3. Create structured outputs:

var userProfile = await client.CreateStructuredOutput<UserProfile>(
    "Create a profile for a software developer named Alice who is 28 years old, " +
    "uses alice@example.com, and prefers JavaScript"
);

Console.WriteLine($"Name: {userProfile.Name}");
Console.WriteLine($"Age: {userProfile.Age}");
Console.WriteLine($"Email: {userProfile.Email}");
Console.WriteLine($"Preferred Language: {userProfile.PreferredLanguage}");

Advanced Usage

Custom Temperature and Token Limits

var profile = await client.CreateStructuredOutput<UserProfile>(
    "Create a profile for an experienced developer",
    temperature: 0.9f,
    maxTokens: 1000
);

Using Azure OpenAI

var client = new InstructorClient(
    "your-api-key-here",
    "https://your-resource-name.openai.azure.com/"
);

Accessing Raw Response

var profile = await client.CreateStructuredOutput<UserProfile>(prompt);
Console.WriteLine($"Raw JSON response: {profile.RawResponse}");

Contributing

Contributions are welcome! Please feel free to submit a Pull Request.

License

This project is licensed under the MIT License - see the LICENSE file for details.