上传iOS 的图片识别逻辑代码

local_plugins/agent_service/ios/agent_service/Sources/agent_service/AgentServiceImpl.swift

@@ -154,7 +154,9 @@ class AgentServiceImpl: NSObject {
         if let openaiModel = config["openaiModel"] as? String {
             self.model = openaiModel
         }
-        
+        if let openaivisionModel = config["openaivisionModel"] as? String {
+            self.visionModel = openaivisionModel
+        }
         if let customSystemPrompt = config["systemPrompt"] as? String, !customSystemPrompt.isEmpty {
             self.systemPrompt = customSystemPrompt
         } else {
@@ -501,7 +503,7 @@ audioStream.saveAudioDataTo(data: audioData)
         os_log("sendMessageStream调用完成", log: logger, type: .info)
     }
     
-    func processImageInput(imagePath: String, text: String, speakResponse: Bool) -> Bool {
+    func processImageInput(imagePath: String,imageUrl: String, text: String, speakResponse: Bool) -> Bool {
         if !isInitialized {
             sendError("服务未初始化", code: "NOT_INITIALIZED")
             return false
@@ -525,15 +527,15 @@ audioStream.saveAudioDataTo(data: audioData)
         DispatchQueue.global(qos: .userInitiated).async { [weak self] in
             guard let self = self else { return }
             
-            guard let imageBase64 = chatApiService.fileToBase64(filePath: imagePath) else {
-                DispatchQueue.main.async {
-                    self.sendEvent(name: "error", data: [
-                        "code": "IMAGE_CONVERSION_FAILED",
-                        "message": "图片转换失败"
-                    ])
-                }
-                return
-            }
+//            guard let imageBase64 = chatApiService.fileToBase64(filePath: imagePath) else {
+//                DispatchQueue.main.async {
+//                    self.sendEvent(name: "error", data: [
+//                        "code": "IMAGE_CONVERSION_FAILED",
+//                        "message": "图片转换失败"
+//                    ])
+//                }
+//                return
+//            }
             
             DispatchQueue.main.async {
                 self.sendEvent(name: "image_ready", data: [
@@ -541,7 +543,7 @@ audioStream.saveAudioDataTo(data: audioData)
                     "imagePath": imagePath
                 ])
                 
-                let userMessage = chatApiService.createUserMessageWithImage(text: text, imageBase64: imageBase64)
+                let userMessage = chatApiService.createUserMessageWithImage(text: text, imageUrl: imageUrl)
                 let displayText = text.isEmpty ? "[图片]" : text
                 
                 self.processWithChatApiServiceInternal(
@@ -747,38 +749,16 @@ private func jsonToString(_ json: [String: Any]) -> String? {
     }
 }
 
-// 处理音乐卡片
-private func processMusicCard(_ cardMusic: [String: Any]) {
-    let id = cardMusic["id"] as? Int ?? 0
-    let url = cardMusic["url"] as? String ?? ""
-    let title = cardMusic["name"] as? String ?? ""
-    let artist = cardMusic["sgener"] as? String ?? ""
-    let coverUrl = cardMusic["image"] as? String ?? ""
-    let ts = cardMusic["ts"] as? Int ?? 0
-    
-    processMusicPlay([
-        "id": id,
-        "url": url,
-        "title": title,
-        "artist": artist,
-        "coverUrl": coverUrl,
-        "ts": ts
-    ])
-}
-
-// 处理音乐列表卡片
-private func processMusicListCard(_ musics: [[String: Any]]) {
-    var playlist: [[String: Any]] = []
-    
-    for item in musics {
-        let id = item["id"] as? Int ?? 0
-        let url = item["url"] as? String ?? ""
-        let title = item["name"] as? String ?? ""
-        let artist = item["sgener"] as? String ?? ""
-        let coverUrl = item["image"] as? String ?? ""
-        let ts = item["ts"] as? Int ?? 0
-        
-        playlist.append([
+    // 处理音乐卡片
+    private func processMusicCard(_ cardMusic: [String: Any]) {
+        let id = cardMusic["id"] as? Int ?? 0
+        let url = cardMusic["url"] as? String ?? ""
+        let title = cardMusic["name"] as? String ?? ""
+        let artist = cardMusic["sgener"] as? String ?? ""
+        let coverUrl = cardMusic["image"] as? String ?? ""
+        let ts = cardMusic["ts"] as? Int ?? 0
+        
+        processMusicPlay([
             "id": id,
             "url": url,
             "title": title,
@@ -787,19 +767,41 @@ private func processMusicListCard(_ musics: [[String: Any]]) {
             "ts": ts
         ])
     }
-    
-    if !playlist.isEmpty {
-        processMusicPlayList(playlist)
+
+    // 处理音乐列表卡片
+    private func processMusicListCard(_ musics: [[String: Any]]) {
+        var playlist: [[String: Any]] = []
+        
+        for item in musics {
+            let id = item["id"] as? Int ?? 0
+            let url = item["url"] as? String ?? ""
+            let title = item["name"] as? String ?? ""
+            let artist = item["sgener"] as? String ?? ""
+            let coverUrl = item["image"] as? String ?? ""
+            let ts = item["ts"] as? Int ?? 0
+            
+            playlist.append([
+                "id": id,
+                "url": url,
+                "title": title,
+                "artist": artist,
+                "coverUrl": coverUrl,
+                "ts": ts
+            ])
+        }
+        
+        if !playlist.isEmpty {
+            processMusicPlayList(playlist)
+        }
     }
-}
 
-// 处理导航卡片
-private func processNavigationCard(_ cardNavigation: [String: Any]) {
-    let start = cardNavigation["start"] as? String ?? ""
-    let end = cardNavigation["end"] as? String ?? ""
-    // 实际处理导航逻辑
-    print("启动导航: 从 \(start) 到 \(end)")
-}
+    // 处理导航卡片
+    private func processNavigationCard(_ cardNavigation: [String: Any]) {
+        let start = cardNavigation["start"] as? String ?? ""
+        let end = cardNavigation["end"] as? String ?? ""
+        // 实际处理导航逻辑
+        print("启动导航: 从 \(start) 到 \(end)")
+    }
     // 后台播放单曲
     private func processMusicPlay(_ song: [String: Any]) {
         sendEvent(name: "music_play", data: song)

local_plugins/agent_service/ios/agent_service/Sources/agent_service/AgentServicePlugin.swift

@@ -65,6 +65,7 @@ public class AgentServicePlugin: NSObject, FlutterPlugin {
             
         case "processImageInput":
             guard let arguments = call.arguments as? [String: Any],
+                  let imageUrl = arguments["imageUrl"] as? String,
                   let imagePath = arguments["imagePath"] as? String else {
                 result(FlutterError(code: "INVALID_ARGUMENTS", message: "图片路径不能为空", details: nil))
                 return
@@ -72,7 +73,7 @@ public class AgentServicePlugin: NSObject, FlutterPlugin {
             
             let text = arguments["text"] as? String ?? ""
             let speakResponse = arguments["speakResponse"] as? Bool ?? false
-            let success = impl.processImageInput(imagePath: imagePath, text: text, speakResponse: speakResponse)
+            let success = impl.processImageInput(imagePath: imagePath,imageUrl: imageUrl, text: text, speakResponse: speakResponse)
             result(success)
             
         case "speakText":

local_plugins/chat_api/ios/chat_api/Sources/chat_api/ChatApiService.swift

@@ -169,7 +169,7 @@ public class ChatApiService: NSObject {
     }
     
     /// 创建带图片的用户消息
-    public func createUserMessageWithImage(text: String, imageBase64: String) -> [String: Any] {
+    public func createUserMessageWithImage(text: String, imageUrl: String) -> [String: Any] {
         var contentArray: [[String: Any]] = []
         
         // 添加文本部分
@@ -184,7 +184,8 @@ public class ChatApiService: NSObject {
         contentArray.append([
             "type": "image_url",
             "image_url": [
-                "url": "data:image/jpeg;base64,\(imageBase64)"
+//                "url": "data:image/jpeg;base64,\(imageBase64)"
+                "url": imageUrl
             ]
         ])
         
@@ -202,9 +203,9 @@ public class ChatApiService: NSObject {
         
         // 检查是否包含图片，决定使用哪个模型
         var currentModel = model
-        // if containsImage(in: messages) {
-        //     currentModel = visionModel
-        // }
+         if containsImage(in: messages) {
+             currentModel = visionModel
+         }
         
         // 转换消息格式
         let chatMessages = try convertToChatMessages(messages)