LLMParser 2.0.5

LLMParser

LLMParser is a .NET library that converts and validates outputs from Large Language Models (LLMs) into strongly typed, reliable objects.

Why LLMParser

• LLM‑focused: Built specifically to handle outputs from Large Language Models, not just generic JSON.
• Automatic schema generation: Create validation schemas directly from your DTOs without boilerplate.
• Reliable validation: Ensures responses strictly match expected structure before deserialization.
• Flexible text conversion: Converts free‑form text like Name: Mark\nAge: 30 into valid JSON.
• Format detection: Recognizes JSON and Key‑Value text formats automatically.
• Error handling: Provides clear exceptions for invalid or unrecognized outputs.

Installation

dotnet add package LLMParser

Basic Usage

Person.cs:

public class Person
{
    public string Name { get; set; }          // Simple text
    public int Age { get; set; }              // Integer
    public decimal Salary { get; set; }       // Decimal
    public DateTime BirthDate { get; set; }   // Date
    public bool IsActive { get; set; }        // Boolean
}

Program.cs:

using LLMParser.Core;
using LLMParser.Json;
using LLMParser.Text;
using LLMParser.Schema;

string llmResponse = "{ \"name\": \"Mark\", \"age\": 30, \"salary\": 3000.50, \"birthDate\": \"1990-05-20\", \"isActive\": true }";

// Check if the response matches a recognized format (JSON or Key:Value text)
if (!ParserUtils.IsRecognizedFormat(llmResponse))
{
    Console.WriteLine("LLM output format not recognized. Cannot parse.");
    return;
}

if (ParserUtils.IsJson(llmResponse))
{
    try
    {
        // Validate and deserialize JSON
        var schema = SchemaFactory.For<Person>();
        var result = JsonParser.EnsureValidJson<Person>(llmResponse, schema);

        Console.WriteLine(
            $"Name: {result.Name}, " +
            $"Age: {result.Age}, " +
            $"Salary: {result.Salary}, " +
            $"BirthDate: {result.BirthDate:yyyy-MM-dd}, " +
            $"IsActive: {result.IsActive}"
        );
        return;
    }
    catch (InvalidLLMOutputException ex)
    {
        Console.WriteLine($"LLM output is invalid: {ex.Message}");
        return;
    }
}

// Expected output:
// Name: Mark, Age: 30, Salary: 3000.50, BirthDate: 1990-05-20, IsActive: True

Usage with Free-form Text

Person.cs:

public class Person
{
    public string Name { get; set; }          // Simple text
    public int Age { get; set; }              // Integer
    public decimal Salary { get; set; }       // Decimal
    public DateTime BirthDate { get; set; }   // Date
    public bool IsActive { get; set; }        // Boolean
}

Program.cs:

using LLMParser.Core;
using LLMParser.Text;
using LLMParser.Json;
using LLMParser.Schema;

string llmResponse = "Name: Mark\nAge: 30\nSalary: 3000.50\nBirthDate: 1990-05-20\nIsActive: true";

// Check if the response matches a recognized format (JSON or Key:Value text)
if (!ParserUtils.IsRecognizedFormat(llmResponse))
{
    Console.WriteLine("LLM output format not recognized. Cannot parse.");
    return;
}

// First check if the response matches Key:Value format
if (ParserUtils.IsKeyValueText(llmResponse))
{
    try
    {
        // Convert free-form text to JSON
        string json = TextToJsonConverter.ConvertKeyValueText(llmResponse);

        // Validate and deserialize
        var schema = SchemaFactory.For<Person>();
        var result = JsonParser.EnsureValidJson<Person>(json, schema);

        Console.WriteLine(
            $"Name: {result.Name}, " +
            $"Age: {result.Age}, " +
            $"Salary: {result.Salary}, " +
            $"BirthDate: {result.BirthDate:yyyy-MM-dd}, " +
            $"IsActive: {result.IsActive}"
        );
    }
    catch (InvalidLLMOutputException ex)
    {
        Console.WriteLine($"LLM output could not be parsed: {ex.Message}");
    }
}

// Expected output:
// Name: Mark, Age: 30, Salary: 3000.50, BirthDate: 1990-05-20, IsActive: True

Key Features

• Schema generation: Automatically builds JSON schemas from DTOs.
• JSON validation: Strictly validates and deserializes JSON against schemas.
• Text‑to‑JSON conversion: Transforms free‑form text into structured JSON.
• Unified format recognition: Detects JSON or Key‑Value text before validation.
• Custom exceptions: InvalidLLMOutputException for clearer debugging.

Quick Reference methods

Support the Project

If you find this library useful, consider supporting its development:

⚖️ License

This project is freely available under the MIT license.
You may use it without restrictions, as long as you retain the reference to the original license.