moderato-app · purr100 · Nov 29, 2025 · Nov 29, 2025
diff --git a/Chato.xcodeproj/project.pbxproj b/Chato.xcodeproj/project.pbxproj
@@ -9,7 +9,6 @@
 /* Begin PBXBuildFile section */
 		EB1B46DF2EDAF0AD00FD053A /* AIProxy in Frameworks */ = {isa = PBXBuildFile; productRef = EB1B46DE2EDAF0AD00FD053A /* AIProxy */; };
 		EB4A0EE52B6F908D0065BCDC /* OpenAI in Frameworks */ = {isa = PBXBuildFile; productRef = EB4A0EE42B6F908D0065BCDC /* OpenAI */; };
-		EB80897D2CE12A100003BACD /* SwiftOpenAI in Frameworks */ = {isa = PBXBuildFile; productRef = EB80897C2CE12A100003BACD /* SwiftOpenAI */; };
 		EB8FB4F32C3940D60041833B /* MarkdownUI in Frameworks */ = {isa = PBXBuildFile; productRef = EB8FB4F22C3940D60041833B /* MarkdownUI */; };
 		EB9E72042BCAB7DF00D5B110 /* ConfettiSwiftUI in Frameworks */ = {isa = PBXBuildFile; productRef = EB9E72032BCAB7DF00D5B110 /* ConfettiSwiftUI */; };
 		EBA73E2F2CDD3006007E480C /* VisualEffectView in Frameworks */ = {isa = PBXBuildFile; productRef = EBA73E2E2CDD3006007E480C /* VisualEffectView */; };
@@ -62,7 +61,6 @@
 			files = (
 				EBC1F9092D1853F50008D5E7 /* Fakery in Frameworks */,
 				EBCFF3952B6FF4CC00498E1E /* Throttler in Frameworks */,
-				EB80897D2CE12A100003BACD /* SwiftOpenAI in Frameworks */,
 				EB4A0EE52B6F908D0065BCDC /* OpenAI in Frameworks */,
 				EBA73E2F2CDD3006007E480C /* VisualEffectView in Frameworks */,
 				EBB4EF292CDDC02200F38178 /* VisualEffectView in Frameworks */,
@@ -114,7 +112,6 @@
 				EB8FB4F22C3940D60041833B /* MarkdownUI */,
 				EBA73E2E2CDD3006007E480C /* VisualEffectView */,
 				EBB4EF282CDDC02200F38178 /* VisualEffectView */,
-				EB80897C2CE12A100003BACD /* SwiftOpenAI */,
 				EBC1F9082D1853F50008D5E7 /* Fakery */,
 				EB1B46DE2EDAF0AD00FD053A /* AIProxy */,
 			);
@@ -154,7 +151,6 @@
 				EB9E72022BCAB7DF00D5B110 /* XCRemoteSwiftPackageReference "ConfettiSwiftUI" */,
 				EB8FB4F12C3940D60041833B /* XCRemoteSwiftPackageReference "swift-markdown-ui" */,
 				EBB4EF272CDDC02200F38178 /* XCRemoteSwiftPackageReference "VisualEffectView" */,
-				EB80897B2CE12A100003BACD /* XCRemoteSwiftPackageReference "SwiftOpenAI" */,
 				EBC1F9072D1853F50008D5E7 /* XCRemoteSwiftPackageReference "Fakery" */,
 				EB1B46DD2EDAF0AD00FD053A /* XCRemoteSwiftPackageReference "AIProxySwift" */,
 			);
@@ -430,14 +426,6 @@
 				version = 0.2.6;
 			};
 		};
-		EB80897B2CE12A100003BACD /* XCRemoteSwiftPackageReference "SwiftOpenAI" */ = {
-			isa = XCRemoteSwiftPackageReference;
-			repositoryURL = "https://github.com/moderato-app/SwiftOpenAI.git";
-			requirement = {
-				kind = revision;
-				revision = a9aa7214d6fdb113da56fb2be6d44b755ed4686c;
-			};
-		};
 		EB8FB4F12C3940D60041833B /* XCRemoteSwiftPackageReference "swift-markdown-ui" */ = {
 			isa = XCRemoteSwiftPackageReference;
 			repositoryURL = "https://github.com/gonzalezreal/swift-markdown-ui.git";
@@ -499,11 +487,6 @@
 			package = EB6422702B6F7D1800119DE6 /* XCRemoteSwiftPackageReference "OpenAI" */;
 			productName = OpenAI;
 		};
-		EB80897C2CE12A100003BACD /* SwiftOpenAI */ = {
-			isa = XCSwiftPackageProductDependency;
-			package = EB80897B2CE12A100003BACD /* XCRemoteSwiftPackageReference "SwiftOpenAI" */;
-			productName = SwiftOpenAI;
-		};
 		EB8FB4F22C3940D60041833B /* MarkdownUI */ = {
 			isa = XCSwiftPackageProductDependency;
 			package = EB8FB4F12C3940D60041833B /* XCRemoteSwiftPackageReference "swift-markdown-ui" */;

diff --git a/Chato.xcodeproj/project.xcworkspace/xcshareddata/swiftpm/Package.resolved b/Chato.xcodeproj/project.xcworkspace/xcshareddata/swiftpm/Package.resolved
diff --git a/Chato/Extensions/OpenAIExt.swift b/Chato/Extensions/OpenAIExt.swift
@@ -1,15 +1,16 @@
 import Foundation
-import SwiftOpenAI
+import AIProxy
 
+fileprivate let timeout: UInt = 60
 extension OpenAIService {
   func hello(model: String) async throws -> String {
-    let msgs: [ChatCompletionParameters.Message] = [.init(role: .user, content: .text("Hello"))]
-    let parameters = ChatCompletionParameters(
-      messages: msgs,
-      model: .custom(model)
+    let msgs: [OpenAIChatCompletionRequestBody.Message] = [.user(content: .text("Hello"))]
+    let parameters = OpenAIChatCompletionRequestBody(
+      model: model,
+      messages: msgs
     )
 
-    let result = try await self.startChat(parameters: parameters)
+    let result = try await self.chatCompletionRequest(body: parameters, secondsToWait: timeout)
     return result.choices[0].message.content ?? ""
   }
 }
diff --git a/Chato/Models/ChatOption.swift b/Chato/Models/ChatOption.swift
@@ -18,7 +18,7 @@ final class ChatOption {
   @Attribute(originalName: "frequency_penalty") var frequencyPenalty: Double = 0
   // An alternative to sampling with temperature, called nucleus sampling, where the model considers the results of the tokens with top_p probability mass. So 0.1 means only the tokens comprising the top 10% probability mass are considered.
 
-  init(model: String = defaultGPTModel.value, contextLength: Int = 2, prompt: Prompt? = nil) {
+  init(model: String = "gpt-4o-mini", contextLength: Int = 2, prompt: Prompt? = nil) {
     self.model = model
     self.contextLength = contextLength
     self.prompt = prompt

diff --git a/Chato/Models/Data/ChatGPT.swift b/Chato/Models/Data/ChatGPT.swift
@@ -1,31 +1,4 @@
 import Foundation
-import SwiftOpenAI
-
-struct ChatGPTModel {
-  let name: String
-  let value: String
-  let swiftOpenAIModel: SwiftOpenAI.Model
-  let version: Int
-
-  init(_ name: String, _ value: String, _ swiftOpenAIModel: SwiftOpenAI.Model, _ version: Int) {
-    self.name = name
-    self.value = value
-    self.swiftOpenAIModel = swiftOpenAIModel
-    self.version = version
-  }
-}
-
-let chatGPTModels = [
-  ChatGPTModel("GPT3.5 Turbo", "gpt-3.5-turbo", .gpt35Turbo, 3),
-  defaultGPTModel,
-  ChatGPTModel("GPT4", "gpt-4", .gpt4, 4),
-  ChatGPTModel("GPT-4o", "gpt-4o", .gpt4o, 4),
-  ChatGPTModel("o1-preview", "o1-preview", .o1Preview, 4),
-  ChatGPTModel("o1-mini", "o1-mini", .o1Mini, 4),
-  ChatGPTModel("GPT-4 Turbo", "gpt-4-turbo", .gpt4turbo, 4)
-].sorted { a, b in a.value < b.value }
-
-let defaultGPTModel = ChatGPTModel("GPT-4o mini", "gpt-4o-mini", .gpt4omini, 3)
 
 enum ContextLength: Hashable {
   case zero

diff --git a/Chato/Service/Deps/APIClientKey.swift b/Chato/Service/Deps/APIClientKey.swift
@@ -2,12 +2,12 @@
 
 import Foundation
 import os
-import SwiftOpenAI
+import AIProxy
 import SwiftUI
 
 // Define your custom environment key
 struct OpenAIServiceKey: EnvironmentKey {
-  static let defaultValue: OpenAIService = OpenAIServiceFactory.service(apiKey: "your-default-api-key")
+  static let defaultValue: OpenAIService = AIProxy.openAIDirectService(unprotectedAPIKey: "your-default-api-key")
 }
 
 // Add new AIClient environment key
@@ -40,24 +40,21 @@ class OpenAIServiceProvider: ObservableObject {
   let service: OpenAIService
 
   init(apiKey: String, endpint: String? = nil, timeout: TimeInterval = 120) {
-    let conf = URLSessionConfiguration.default
-    conf.timeoutIntervalForRequest = timeout
-
     if let endpint {
       do {
         let res = try parseURL(endpint)
-        self.service = OpenAIServiceFactory.service(apiKey: apiKey, configuration: conf, overrideBaseURL: res.base, proxyPath: res.path)
+        self.service = AIProxy.openAIDirectService(unprotectedAPIKey: apiKey, baseURL: res.base)
       } catch {
         AppLogger.logError(.from(
           error: error,
           operation: "Parse URL",
           component: "OpenAIServiceProvider",
           userMessage: nil
         ))
-        self.service = OpenAIServiceFactory.service(apiKey: apiKey, configuration: conf)
+        self.service = AIProxy.openAIDirectService(unprotectedAPIKey: apiKey)
       }
     } else {
-      self.service = OpenAIServiceFactory.service(apiKey: apiKey, configuration: conf)
+      self.service = AIProxy.openAIDirectService(unprotectedAPIKey: apiKey)
     }
   }
 }

diff --git a/Chato/Views/MessageList/VM.swift b/Chato/Views/MessageList/VM.swift
@@ -1,8 +1,8 @@
+import AIProxy
 import Combine
+import os
 import SwiftData
-import SwiftOpenAI
 import SwiftUI
-import os
 
 extension InputAreaView {
   static let whiteSpaces: [Character] = ["\n", " ", "\t"]
@@ -83,51 +83,62 @@ extension InputAreaView {
   }
 
   func ask2(text: String, contextLength: Int) {
-    let isO1 = chat.option.model.contains("o1-")
-    let timeout: Double = isO1 ? 120.0 : 15.0
-    AppLogger.network.debug("using timeout: \(timeout), contextLength: \(contextLength)")
-
-    var msgs: [ChatCompletionParameters.Message] = [.init(role: .user, content: .text(text))]
+    var msgs: [OpenAIChatCompletionRequestBody.Message] = [.user(content: .text(text))]
 
     var actualCL = 0
     if contextLength > 0 {
       let hist = self.modelContext.recentMessgagesEarlyOnTop(chatId: chat.persistentModelID, limit: contextLength)
       actualCL = hist.count
 
       for item in hist.sorted().reversed() {
-        let msg: ChatCompletionParameters.Message = .init(role: item.role == .user ? .user : (item.role == .assistant ? .assistant : .system), content: .text(item.message.isMeaningful ? item.message : item.errorInfo))
+        let msg: OpenAIChatCompletionRequestBody.Message
+        switch item.role {
+        case .user:
+          msg = .user(content: .text(item.message.isMeaningful ? item.message : item.errorInfo))
+        case .assistant:
+          msg = .assistant(content: .text(item.message.isMeaningful ? item.message : item.errorInfo))
+        default:
+          msg = .system(content: .text(item.message.isMeaningful ? item.message : item.errorInfo))
+        }
         msgs.insert(msg, at: 0)
       }
     }
 
     chat.option.prompt?.messages.sorted().reversed().forEach {
-      let msg: ChatCompletionParameters.Message = .init(role: $0.role == .assistant ? .assistant : ($0.role == .user || isO1 ? .user : .system), content: .text($0.content))
+      let msg: OpenAIChatCompletionRequestBody.Message
+      if $0.role == .assistant {
+        msg = .assistant(content: .text($0.content))
+      } else if $0.role == .user {
+        msg = .user(content: .text($0.content))
+      } else {
+        msg = .system(content: .text($0.content))
+      }
       msgs.insert(msg, at: 0)
     }
 
-    let parameters = ChatCompletionParameters(
+    let streamParameters = OpenAIChatCompletionRequestBody(
+      model: chat.option.model,
       messages: msgs,
-      model: .custom(chat.option.model),
       frequencyPenalty: chat.option.maybeFrequencyPenalty,
       presencePenalty: chat.option.maybePresencePenalty,
+      stream: true,
       temperature: chat.option.maybeTemperature
     )
 
-    let streamParameters = ChatCompletionParameters(
-      messages: msgs,
-      model: .custom(chat.option.model),
-      frequencyPenalty: chat.option.maybeFrequencyPenalty,
-      presencePenalty: chat.option.maybePresencePenalty,
-      temperature: chat.option.maybeTemperature,
-      streamOptions: ChatCompletionParameters.StreamOptions(includeUsage: true)
-    )
-
     AppLogger.network.debug("using temperature: \(String(describing: chat.option.maybeTemperature)), presencePenalty: \(String(describing: chat.option.maybePresencePenalty)), frequencyPenalty: \(String(describing: chat.option.maybeFrequencyPenalty))")
 
     AppLogger.network.debug("===whole message list begins===")
 
     for (i, m) in msgs.enumerated() {
-      AppLogger.network.debug("\(i).\(m.role): \(String(describing: m.content))")
+      let roleStr: String
+      switch m {
+      case .user: roleStr = "user"
+      case .assistant: roleStr = "assistant"
+      case .system: roleStr = "system"
+      case .developer: roleStr = "developer"
+      case .tool: roleStr = "tool"
+      }
+      AppLogger.network.debug("\(i).\(roleStr): \(String(describing: m))")
     }
 
     AppLogger.network.debug("===whole message list ends===")
@@ -177,56 +188,34 @@ extension InputAreaView {
 
     Task.detached {
       do {
-        if isO1 {
-          AppLogger.network.info("not using stream")
-          Task { @MainActor in
-            aiMsg.meta?.startedAt = .now
-          }
-          let result = try await openAIService.startChat(parameters: parameters)
+        AppLogger.network.info("using stream")
+        let stream = try await openAIService.streamingChatCompletionRequest(body: streamParameters, secondsToWait: 60)
+        for try await chunk in stream {
           Task { @MainActor in
-            let content = result.choices[0].message.content ?? ""
-            userMsg.onSent()
-            userMsg.meta?.promptTokens = result.usage.promptTokens
-            aiMsg.meta?.completionTokens = result.usage.completionTokens
-            aiMsg.onEOF(text: content)
-            em.messageEvent.send(.eof)
-          }
-        } else {
-          AppLogger.network.info("using stream")
-          let stream = try await openAIService.startStreamedChat(parameters: streamParameters)
-          for try await chunk in stream {
-            Task { @MainActor in
-              if aiMsg.meta?.startedAt == nil {
-                aiMsg.meta?.startedAt = .now
-              }
-              if userMsg.status == .sending {
-                userMsg.onSent()
-              }
+            if aiMsg.meta?.startedAt == nil {
+              aiMsg.meta?.startedAt = .now
+            }
+            if userMsg.status == .sending {
+              userMsg.onSent()
+            }
 
-              if let choice = chunk.choices.first {
-                let text = choice.delta.content ?? "\nchoice has no content\n"
-                if let fr = choice.finishReason {
-                  if case .string(let reason) = fr {
-                    AppLogger.network.debug("finished reason: \(reason)")
-                    aiMsg.onEOF(text: "")
-                    em.messageEvent.send(.eof)
-////                    skip eof; the last chunk's 'usage' isn't nil and it has no 'choices'
-//                  } else {
-//                    // if reason is not stop, something may be wrong
-//                    aiMsg.onError("finished reason: \(fr)", .unknown)
-                  }
-                } else {
-                  aiMsg.onTyping(text: text)
-                }
+            if let choice = chunk.choices.first {
+              let text = choice.delta.content ?? "\nchoice has no content\n"
+              if let fr = choice.finishReason {
+                AppLogger.network.debug("finished reason: \(fr)")
+                aiMsg.onEOF(text: "")
+                em.messageEvent.send(.eof)
+              } else {
+                aiMsg.onTyping(text: text)
+              }
+            } else {
+              if let usage = chunk.usage {
+                userMsg.meta?.promptTokens = usage.promptTokens
+                aiMsg.meta?.completionTokens = usage.completionTokens
+                aiMsg.onEOF(text: "")
+                em.messageEvent.send(.eof)
               } else {
-                if let usage = chunk.usage {
-                  userMsg.meta?.promptTokens = usage.promptTokens
-                  aiMsg.meta?.completionTokens = usage.completionTokens
-                  aiMsg.onEOF(text: "")
-                  em.messageEvent.send(.eof)
-                } else {
-                  AppLogger.network.debug("no text in chunk, chunk: \(String(describing: chunk))")
-                }
+                AppLogger.network.debug("no text in chunk, chunk: \(String(describing: chunk))")
               }
             }
           }

diff --git a/Chato/Views/Settings/ChatGPTSettingSection.swift b/Chato/Views/Settings/ChatGPTSettingSection.swift
@@ -1,6 +1,6 @@
 import Haptico
 import SwiftData
-import SwiftOpenAI
+import AIProxy
 import SwiftUI
 
 struct ChatGPTSettingSection: View {
@@ -76,7 +76,7 @@ struct ChatGPTSettingSection: View {
         TestButton {
           let openai = pref.gptEnableEndpoint ? OpenAIServiceProvider(apiKey: pref.gptApiKey, endpint: pref.gptEndpoint, timeout: 5) : OpenAIServiceProvider(apiKey: pref.gptApiKey, timeout: 5)
           do {
-            let res = try await openai.service.hello(model: models.randomElement()?.modelId ?? Model.gpt4omini.value)
+            let res = try await openai.service.hello(model: models.randomElement()?.modelId ?? "gpt-4o-mini")
             HapticsService.shared.shake(.success)
             return .succeeded(res)
           } catch {