Fluent AI

Fluent AI is a powerful library designed to build and execute advanced prompts using OpenAI's various ChatGPT models. It leverages the capabilities of Azure.AI.OpenAI and NJsonSchema packages to provide a seamless experience for integrating AI into your .NET applications.

Features

Fluent Building of User and System Prompts: Easily construct prompts for ChatGPT models.
Custom Chat Tools Integration: Integrate custom chat functions to enable ChatGPT to call arbitrary .NET methods and utilize returned values.
Structured Responses: Automatically instruct the model to respond in a specified JSON format.
Built-in Validation: Validate ChatGPT responses using JSON Schema and auto-retry in case the model returns incorrectly formatted answers.

Example Usage

var openAiToken = "...";

var client = new ChatCompletionsOpenAiClient(openAiToken);
var builder = client.ToCompletionsBuilder();

var request = builder
    .UseChatGpt4o()
    .UseChatTool(new FetchUrlTool())
    .UserPrompt("Give me short description of the following webpage: https://docs.bland.ai/welcome-to-bland")
    .UseResponseSchema<ChatGptResponse>()
    .BuildCompletionsRequest();

var response = await request.GetStructuredResponse<ChatGptResponse>();

Console.WriteLine(response.Text);

[Description("This is response model you should use to send answer for questions")]
public class ChatGptResponse
{
    [Description("Your response message"), Required]
    public string Text { get; set; }
}

Implementation of FetchUrlTool

/// <summary>
/// Handles requests to fetch content from a specified URL.
/// </summary>
[Description("Gets content by specified URL")]
public class FetchUrlTool : ChatCompletionToolBase<FetchUrlTool.FetchUrlToolRequest, FetchUrlTool.FetchUrlToolResponse>
{
    public FetchUrlTool() : base("fetch_url") {}

    /// <summary>
    /// Handles the specified request to fetch content from a URL.
    /// </summary>
    /// <param name="request">The request containing the URL to fetch.</param>
    /// <returns>A task that represents the asynchronous operation. The task result contains the response with the fetched content.</returns>
    public override async Task<FetchUrlToolResponse> Handle(FetchUrlToolRequest request)
    {
        var httpClient = new HttpClient();
        var response = await httpClient.GetStringAsync(request.Url);
        return new FetchUrlToolResponse() { Content = response };
    }

    /// <summary>
    /// Represents a request to fetch content from a specified URL.
    /// </summary>
    public class FetchUrlToolRequest
    {
        /// <summary>
        /// Gets or sets the URL to fetch.
        /// </summary>
        [Description("The URL"), Required, Url]
        public string Url { get; set; }
    }

    /// <summary>
    /// Represents the response containing the content fetched from a URL.
    /// </summary>
    public class FetchUrlToolResponse
    {
        /// <summary>
        /// Gets or sets the content fetched from the URL.
        /// </summary>
        public string Content { get; set; }
    }
}

Important Note

The DescriptionAttribute is mandatory as it describes the meaning of the properties to the model, ensuring that the AI understands the context and purpose of each property.

Contributing

Contributions are welcome! Please feel free to submit a pull request or open an issue.

License

This project is licensed under the MIT License. See the LICENSE file for more details.