terra-llm/Assets/LLM/Editor/Client/GeminiApiClient.cs

using System.Linq;
using System.Text;
using UnityEditor;
using UnityEngine;
using LLM.Editor.Core;
using System.Diagnostics;
using UnityEngine.Networking;
using System.Threading.Tasks;
using System.Collections.Generic;
using Debug = UnityEngine.Debug;

namespace LLM.Editor.Client
{
    public static class GeminiApiClient
    {
        [System.Serializable]
        private class CommandResponse { public List<Data.CommandData> commands; }

        private static Settings.MCPSettings _settings;

        private static bool LoadSettings()
        {
            if (_settings) return true;

            var guids = AssetDatabase.FindAssets("t:MCPSettings");
            if (guids.Length == 0)
            {
                Debug.LogError("[GeminiApiClient] Could not find MCPSettings asset. Please create one via Assets > Create > LLM > MCP Settings.");
                return false;
            }
            var path = AssetDatabase.GUIDToAssetPath(guids[0]);
            _settings = AssetDatabase.LoadAssetAtPath<Settings.MCPSettings>(path);
            return _settings;
        }

        private static string GetAuthToken()
        {
            if (!LoadSettings()) return null;

            try
            {
                var startInfo = new ProcessStartInfo
                {
                    FileName = _settings.gcloudPath,
                    Arguments = "auth print-access-token",
                    RedirectStandardOutput = true,
                    RedirectStandardError = true,
                    UseShellExecute = false,
                    CreateNoWindow = true
                };

                using var process = Process.Start(startInfo);
                if (process == null) return null;
                
                var accessToken = process.StandardOutput.ReadToEnd().Trim();
                var error = process.StandardError.ReadToEnd();
                process.WaitForExit();

                if (process.ExitCode == 0) return accessToken;
                
                Debug.LogError($"[GeminiApiClient] gcloud auth error: {error}");
                return null;
            }
            catch (System.Exception e)
            {
                Debug.LogError($"[GeminiApiClient] Exception while getting auth token: {e.Message}");
                return null;
            }
        }

        private static string GetSystemPrompt()
        {
            return @"You are an expert Unity development assistant. Your goal is to help the user by breaking down their request into a sequence of commands.
You MUST ONLY respond with a single JSON object containing a 'commands' array. Do not use markdown or any other formatting.

Here are the available commands:
- DisplayMessage: Show a text message to the user. Use this for conversational replies, questions, or to provide information.
- CreateScript: Creates a new C# script file.
- CreatePrefab: Creates a new prefab from a source model found in the project.
- AddComponentToAsset: Attaches a script to a prefab asset.
- InstantiatePrefab: Creates an instance of a prefab in the current scene.
- RequestClarification: Ask the user a follow-up question if their request is ambiguous.

Example of a valid response:
{
    ""commands"": [
        {
            ""commandName"": ""DisplayMessage"",
            ""jsonData"": ""{\""message\"":\""Hello! How can I help?""\}""
        }
    ]
}";
        }

        public static async Task SendPrompt(string prompt)
        {
            if (!LoadSettings()) return;

            Debug.Log($"[GeminiApiClient] Getting auth token...");
            var authToken = GetAuthToken();
            if (string.IsNullOrEmpty(authToken))
            {
                Debug.LogError("[GeminiApiClient] Failed to get authentication token. Is gcloud CLI installed and are you logged in?");
                return;
            }
            
            Debug.Log($"[GeminiApiClient] Sending prompt to live API: {prompt}");
            
            var chatHistory = SessionManager.LoadChatHistory();
            chatHistory.Add(new Data.ChatEntry { role = "user", content = prompt });
            SessionManager.SaveChatHistory(chatHistory);

            var url = $"https://{_settings.gcpRegion}-aiplatform.googleapis.com/v1/projects/{_settings.gcpProjectId}/locations/{_settings.gcpRegion}/publishers/google/models/{_settings.modelName}:generateContent";
            
            // Construct the request payload, adding the system prompt first.
            var apiRequest = new Api.ApiRequest
            {
                system_instruction = new Api.SystemInstruction
                {
                    parts = new List<Api.Part> { new() { text = GetSystemPrompt() } }
                },
                contents = chatHistory.Select(entry => new Api.Content
                {
                    // The API expects "model" for the assistant's role.
                    role = entry.role == "assistant" ? "model" : entry.role,
                    parts = new List<Api.Part> { new() { text = entry.content } }
                }).ToList()
            };
            
            var jsonPayload = JsonUtility.ToJson(apiRequest);

            using var request = new UnityWebRequest(url, "POST");
            var bodyRaw = Encoding.UTF8.GetBytes(jsonPayload);
            request.uploadHandler = new UploadHandlerRaw(bodyRaw);
            request.downloadHandler = new DownloadHandlerBuffer();
            request.SetRequestHeader("Content-Type", "application/json");
            request.SetRequestHeader("Authorization", $"Bearer {authToken}");

            var operation = request.SendWebRequest();
            while (!operation.isDone)
            {
                await Task.Yield();
            }

            if (request.result == UnityWebRequest.Result.Success)
            {
                var responseJson = request.downloadHandler.text;
                var apiResponse = JsonUtility.FromJson<Api.ApiResponse>(responseJson);
                    
                if (apiResponse.candidates != null && apiResponse.candidates.Any())
                {
                    var commandJson = apiResponse.candidates[0].content.parts[0].text;
                    Debug.Log($"[GeminiApiClient] Command received: {commandJson}.");
                    chatHistory.Add(new Data.ChatEntry { role = "model", content = commandJson });
                    SessionManager.SaveChatHistory(chatHistory);
                    var commandResponse = JsonUtility.FromJson<CommandResponse>(commandJson);
                    if (commandResponse is { commands: not null })
                    {
                        Debug.Log($"[GeminiApiClient] Received {commandResponse.commands.Count} commands from LLM.");
                        CommandExecutor.SetQueue(commandResponse.commands);
                    }
                    else
                    {
                        Debug.LogError($"[GeminiApiClient] Failed to parse command structure from LLM response text: {commandJson}");
                    }
                }
            }
            else
            {
                Debug.LogError($"[GeminiApiClient] API Error: {request.error}\n{request.downloadHandler.text}");
            }
        }
    }
}