feat(AI): 统一多模态模型架构,整合文本和视觉推理路径

- 将文本生成和VL(图→文)功能合并到单一的Qwen3.5-4B多模态MNN模型
- 移除独立的Qwen3-VL-4B模型依赖,MLX VL改为使用.llm的多模态模型
- 更新ModelKind枚举,新增userFacing集合用于面向用户展示
- MNN后端现在同时支持文本和视觉任务,模拟器回退到MLX

refactor(models): 模型管理和界面调整以适应新的多模态架构

- 更新模型管理界面,只显示统一的Qwen3.5-4B(MNN)模型给用户
- 修改就绪状态检查逻辑,使用ModelKind.userFacing替代allCases
- 更新模型文件清单,从Qwen3.5-2B升级到Qwen3.5-4B-4bit
- 调整模型管理页面UI,突出MNN+SME2端侧加速功能

feat(camera): 添加拍照识别引擎切换功能

- 实现双路径拍照识别:Apple Vision OCR + 文本模型 和 Qwen3-VL直接识别
- 添加预处理逻辑,优化Qwen3-VL对窄长区域图片的识别效果
- 在模型管理页面添加拍照识别引擎选择组件
- 提供用户界面选项,在两种识别方式间切换

style(ui): 优化输入框样式和颜色主题一致性

- 为指标快速表单添加浅色主题偏好
- 统一所有文本输入框的颜色样式(theme)
- 创建EntryInputField组件,替换原有的单行输入+按钮模式
- 实现聊天框风格的条目输入,支持多行自适应和圆形发送按钮

fix(build): 修正Xcode项目配置中的重复框架搜索路径

- 清理project.pbxproj中重复的FRAMEWORK_SEARCH_PATHS配置
- 重新排列Swift桥接头文件配置确保正确引用
- 修复因路径配置重复导致的编译警告问题

test: 增加区域图片预处理和模型清单测试覆盖

- 添加RegionImageCropper.prepareForQwenVL的单元测试
- 验证宽而矮图片的放大和填充逻辑
- 更新ModelManifestTests中的字节数预期值以匹配新模型
- 修正OCRService中VNRecognizedTextObservation类型的处理
```
This commit is contained in:
link2026
2026-06-08 23:25:31 +08:00
parent b919404412
commit 836f3d4234
20 changed files with 393 additions and 65 deletions

View File

@@ -420,6 +420,7 @@
ENABLE_HARDENED_RUNTIME = YES;
ENABLE_PREVIEWS = YES;
ENABLE_USER_SELECTED_FILES = readonly;
FRAMEWORK_SEARCH_PATHS = "$(PROJECT_DIR)/Frameworks";
GENERATE_INFOPLIST_FILE = YES;
INFOPLIST_KEY_CFBundleDisplayName = "康康";
INFOPLIST_KEY_ITSAppUsesNonExemptEncryption = NO;
@@ -447,8 +448,6 @@
PRODUCT_BUNDLE_IDENTIFIER = com.xuhuayong.kangkang;
PRODUCT_NAME = "$(TARGET_NAME)";
REGISTER_APP_GROUPS = YES;
FRAMEWORK_SEARCH_PATHS = "$(PROJECT_DIR)/Frameworks";
SWIFT_OBJC_BRIDGING_HEADER = "康康/康康-Bridging-Header.h";
SDKROOT = auto;
STRING_CATALOG_GENERATE_SYMBOLS = YES;
SUPPORTED_PLATFORMS = "iphoneos iphonesimulator";
@@ -456,6 +455,7 @@
SWIFT_APPROACHABLE_CONCURRENCY = YES;
SWIFT_DEFAULT_ACTOR_ISOLATION = MainActor;
SWIFT_EMIT_LOC_STRINGS = YES;
SWIFT_OBJC_BRIDGING_HEADER = "康康/康康-Bridging-Header.h";
SWIFT_UPCOMING_FEATURE_MEMBER_IMPORT_VISIBILITY = YES;
SWIFT_VERSION = 5.0;
TARGETED_DEVICE_FAMILY = "1,2";
@@ -476,6 +476,7 @@
ENABLE_HARDENED_RUNTIME = YES;
ENABLE_PREVIEWS = YES;
ENABLE_USER_SELECTED_FILES = readonly;
FRAMEWORK_SEARCH_PATHS = "$(PROJECT_DIR)/Frameworks";
GENERATE_INFOPLIST_FILE = YES;
INFOPLIST_KEY_CFBundleDisplayName = "康康";
INFOPLIST_KEY_ITSAppUsesNonExemptEncryption = NO;
@@ -503,8 +504,6 @@
PRODUCT_BUNDLE_IDENTIFIER = com.xuhuayong.kangkang;
PRODUCT_NAME = "$(TARGET_NAME)";
REGISTER_APP_GROUPS = YES;
FRAMEWORK_SEARCH_PATHS = "$(PROJECT_DIR)/Frameworks";
SWIFT_OBJC_BRIDGING_HEADER = "康康/康康-Bridging-Header.h";
SDKROOT = auto;
STRING_CATALOG_GENERATE_SYMBOLS = YES;
SUPPORTED_PLATFORMS = "iphoneos iphonesimulator";
@@ -512,6 +511,7 @@
SWIFT_APPROACHABLE_CONCURRENCY = YES;
SWIFT_DEFAULT_ACTOR_ISOLATION = MainActor;
SWIFT_EMIT_LOC_STRINGS = YES;
SWIFT_OBJC_BRIDGING_HEADER = "康康/康康-Bridging-Header.h";
SWIFT_UPCOMING_FEATURE_MEMBER_IMPORT_VISIBILITY = YES;
SWIFT_VERSION = 5.0;
TARGETED_DEVICE_FAMILY = "1,2";