From 952aa9a188446d2cb3419ec3ee8a0a895f4cdfd6 Mon Sep 17 00:00:00 2001
From: SwiftBeta <swiftbeta.blog@gmail.com>
Date: Wed, 13 Mar 2024 16:19:12 +0100
Subject: [PATCH] Add new Vison feature to use local images

---
 Demo/Demo/Vision/VisionView.swift      | 58 +++++++++++++++++++++-----
 Demo/Demo/Vision/VisionViewModel.swift | 18 +++++++-
 2 files changed, 64 insertions(+), 12 deletions(-)

diff --git a/Demo/Demo/Vision/VisionView.swift b/Demo/Demo/Vision/VisionView.swift
index f36699d..80ac5a4 100644
--- a/Demo/Demo/Vision/VisionView.swift
+++ b/Demo/Demo/Vision/VisionView.swift
@@ -1,17 +1,51 @@
 import SwiftUI
+import PhotosUI
 
 struct VisionView: View {
-    var viewModel: VisionViewModel
+    @State private var visionStrategy = 0
+    @State var viewModel: VisionViewModel
     
     var body: some View {
         VStack {
-            AsyncImage(url: URL(string: viewModel.imageVisionURL)) { image in
-                image
-                    .resizable()
-                    .scaledToFit()
-                    .frame(width: 300, height: 300)
-            } placeholder: {
-                ProgressView()
+            Picker("What is your favorite color?", selection: $visionStrategy) {
+                Text("URL").tag(0)
+                Text("Gallery").tag(1)
+            }
+            .pickerStyle(.segmented)
+            
+            if visionStrategy == 0 {
+                AsyncImage(url: URL(string: viewModel.imageVisionURL)) { image in
+                    image
+                        .resizable()
+                        .scaledToFit()
+                        .frame(width: 300, height: 300)
+                } placeholder: {
+                    ProgressView()
+                        .padding(.bottom, 20)
+                }
+            } else {
+                PhotosPicker(selection: $viewModel.photoSelection,
+                             matching: .images,
+                             photoLibrary: .shared()) {
+                    Label("Add video or audio",
+                          systemImage: "video.fill")
+                }
+                             .frame(height: 300)
+                             .photosPickerStyle(.inline)
+                             .onChange(of: viewModel.photoSelection!) { oldValue, newValue in
+                                 newValue.loadTransferable(type: Data.self) { [self] result in
+                                     switch result {
+                                     case .success(let data):
+                                         if let data {
+                                             viewModel.currentData = data
+                                         } else {
+                                             print("No supported content type found.")
+                                         }
+                                     case .failure(let error):
+                                         fatalError(error.localizedDescription)
+                                     }
+                                 }
+                             }
             }
             
             if !viewModel.isLoading {
@@ -20,18 +54,22 @@ struct VisionView: View {
                         await viewModel.send(message: "Please analyze the image and describe its contents, providing any relevant details or information")
                     }
                 }, label: {
-                    Text("Describe Image")
+                    Text("Describe Image from URL")
                 })
                 .buttonStyle(.borderedProminent)
             } else {
                 ProgressView()
             }
             
-            TextEditor(text: .constant( viewModel.message))
+            Divider()
+                .padding(.top, 20)
+            
+            TextEditor(text: .constant(viewModel.message))
                 .font(.body)
                 .padding(.top, 12)
                 .padding(.horizontal)
         }
+        .padding(.horizontal, 32)
     }
 }
 
diff --git a/Demo/Demo/Vision/VisionViewModel.swift b/Demo/Demo/Vision/VisionViewModel.swift
index a1e06b4..fbdacb0 100644
--- a/Demo/Demo/Vision/VisionViewModel.swift
+++ b/Demo/Demo/Vision/VisionViewModel.swift
@@ -1,5 +1,7 @@
 import Foundation
 import SwiftOpenAI
+import PhotosUI
+import SwiftUI
 
 @Observable
 final class VisionViewModel {
@@ -8,13 +10,25 @@ final class VisionViewModel {
     var message: String = ""
     var isLoading = false
     
+    // Local Image
+    var photoSelection: PhotosPickerItem? = .init(itemIdentifier: "")
+    var currentData: Data?
+    
     @MainActor
     func send(message: String) async {
         isLoading = true
 
         do {
+            let imageValue: String
+            if let data = currentData {
+                let base64Image = data.base64EncodedString()
+                imageValue = "data:image/jpeg;base64,\(base64Image)"
+            } else {
+                imageValue = imageVisionURL
+            }
+            
             let myMessage = MessageChatImageInput(text: message,
-                                                  imageURL: imageVisionURL,
+                                                  imageURL: imageValue,
                                                   role: .user)
                                     
             let optionalParameters: ChatCompletionsOptionalParameters = .init(temperature: 0.5, 
@@ -25,7 +39,7 @@ final class VisionViewModel {
             let result = try await openAI.createChatCompletionsWithImageInput(model: .gpt4(.gpt_4_vision_preview),
                                                                                              messages: [myMessage],
                                                                                              optionalParameters: optionalParameters)
-            
+            self.currentData = nil
             self.message = result?.choices.first?.message.content ?? "No value"
             self.isLoading = false