Merge pull request #172 from Regenhardt/feature/coding-assistant

Coding assistent example

LLama.Examples/NewVersion/CodingAssistant.cs

@@ -0,0 +1,94 @@
+namespace LLama.Examples.NewVersion
+{
+    using LLama.Common;
+    using System;
+    using System.Reflection;
+
+    internal class CodingAssistant
+    {
+        const string DefaultModelUri = "https://huggingface.co/TheBloke/CodeLlama-7B-Instruct-GGUF/resolve/main/codellama-7b-instruct.Q4_K_S.gguf";
+
+        // Source paper with example prompts:
+        // https://doi.org/10.48550/arXiv.2308.12950
+        const string InstructionPrefix = "[INST]";
+        const string InstructionSuffix = "[/INST]";
+        const string SystemInstruction = "You're an intelligent, concise coding assistant. Wrap code in ``` for readability. Don't repeat yourself. Use best practice and good coding standards.";
+        private static string ModelsDirectory = Path.Combine(Directory.GetParent(Assembly.GetExecutingAssembly().Location)!.FullName, "Models");
+
+        public static async Task Run()
+        {
+            Console.Write("Please input your model path (if left empty, a default model will be downloaded for you): ");
+            var modelPath = Console.ReadLine();
+
+            if(string.IsNullOrWhiteSpace(modelPath) )
+            {
+                modelPath = await GetDefaultModel();
+            }
+
+            var parameters = new ModelParams(modelPath)
+            {
+                ContextSize = 4096
+            };
+            using var model = LLamaWeights.LoadFromFile(parameters);
+            using var context = model.CreateContext(parameters);
+            var executor = new InstructExecutor(context, InstructionPrefix, InstructionSuffix);
+
+            Console.ForegroundColor = ConsoleColor.Yellow;
+            Console.WriteLine("The executor has been enabled. In this example, the LLM will follow your instructions." +
+                "\nIt's a 7B Code Llama, so it's trained for programming tasks like \"Write a C# function reading a file name from a given URI\" or \"Write some programming interview questions\"." +
+                "\nWrite 'exit' to exit");
+            Console.ForegroundColor = ConsoleColor.White;
+
+            var inferenceParams = new InferenceParams() { 
+                Temperature = 0.8f, 
+                MaxTokens = -1,
+            };
+
+            string instruction = $"{SystemInstruction}\n\n";
+            await Console.Out.WriteAsync("Instruction: ");
+            instruction += Console.ReadLine() ?? "Ask me for instructions.";
+            while (instruction != "exit")
+            {
+
+                Console.ForegroundColor = ConsoleColor.Green;
+                foreach (var text in executor.Infer(instruction+System.Environment.NewLine, inferenceParams))
+                {
+                    Console.Write(text);
+                }
+                Console.ForegroundColor = ConsoleColor.White;
+
+                await Console.Out.WriteAsync("Instruction: ");
+                instruction = Console.ReadLine() ?? "Ask me for instructions.";
+            }
+        }
+
+        private static async Task<string> GetDefaultModel()
+        {
+            var uri = new Uri(DefaultModelUri);
+            var modelName = uri.Segments[^1];
+            await Console.Out.WriteLineAsync($"The following model will be used: {modelName}");
+            var modelPath = Path.Combine(ModelsDirectory, modelName);
+            if(!Directory.Exists(ModelsDirectory))
+            {
+                Directory.CreateDirectory(ModelsDirectory);
+            }
+
+            if (File.Exists(modelPath))
+            {
+                await Console.Out.WriteLineAsync($"Existing model found, using {modelPath}");
+            }
+            else
+            {
+                await Console.Out.WriteLineAsync($"Model not found locally, downloading {DefaultModelUri}...");
+                using var http = new HttpClient();
+                await using var downloadStream = await http.GetStreamAsync(uri);
+                await using var fileStream = new FileStream(modelPath, FileMode.Create, FileAccess.Write);
+                await downloadStream.CopyToAsync(fileStream);
+                await Console.Out.WriteLineAsync($"Model downloaded and saved to {modelPath}");
+            }
+
+
+            return modelPath;
+        }
+    }
+}

LLama.Examples/NewVersion/TestRunner.cs

@@ -21,6 +21,7 @@
             Console.WriteLine("11: Semantic Kernel Prompt.");
             Console.WriteLine("12: Semantic Kernel Chat.");
             Console.WriteLine("13: Semantic Kernel Memory.");
+            Console.WriteLine("14: Coding Assistant.");
 
             while (true)
             {
@@ -83,6 +84,10 @@
                 {
                     await SemanticKernelMemory.Run();
                 }
+                else if(choice == 14)
+                {
+                    await CodingAssistant.Run();
+                }
                 else
                 {
                     Console.WriteLine("Cannot parse your choice. Please select again.");