macOS 1.8：三评汇总成一份报告

Documentation/Validation.md

@@ -1,3 +1,8 @@
+# 1.8.0 validation · 2026-10-02
+
+- 72 Swift tests pass. New coverage: a full review calls the synthesizer once and a quick review never does; the synthesized report keeps the app's median score and diagnostics, keeps only corrections anchored in the answer, and becomes the conclusion, feedback and improved version; usage includes the synthesis call; a failed synthesis leaves the marks intact, falls back to the local merge and records why; English I translation totals are the sum of per-sentence medians; the synthesis prompt carries the fixed marks and forbids mentioning examiners.
+- The consolidated review page was rendered offscreen. No paid live grading was run.
+
 # 1.7.0 validation · 2026-10-02
 
 - 69 Swift tests pass. New coverage: English I translation totals reconcile to the sum of segment marks (allowing a single 0.5 typo deduction), out-of-range segments are dropped, segments are cleared for other tasks; per-sentence marks come from the median reviewer with every judge's mark; legacy reports without segments decode; prompts require segments only for English I translation and the Codex schema includes them; numbered underlined segments are extracted from the question.

README.md

@@ -137,7 +137,7 @@ platforms/android/   Android Studio 项目与手机界面
 scripts/             构建、图标生成与显式联调脚本
 ```
 
-macOS 自动化测试覆盖 69 个案例，Android 有 5 个领域测试与 2 个实际设备服务测试；Windows 通过评分、持久化、字段校验及实际 OCR 自检。GPT-6 Astra/MAX 已通过实际 Swift 子进程完成样例评卷；DeepSeek 已验证官方模型接口连接。后台、流式输出及版本路径已用隔离测试验证；本次未执行完整付费三评，需用户填入有效 Key 后使用。
+macOS 自动化测试覆盖 72 个案例，Android 有 5 个领域测试与 2 个实际设备服务测试；Windows 通过评分、持久化、字段校验及实际 OCR 自检。GPT-6 Astra/MAX 已通过实际 Swift 子进程完成样例评卷；DeepSeek 已验证官方模型接口连接。后台、流式输出及版本路径已用隔离测试验证；本次未执行完整付费三评，需用户填入有效 Key 后使用。
 
 <details>
 <summary><strong>更多文档</strong></summary>

RELEASE-NOTES.md

@@ -1,3 +1,10 @@
+# 1.8.0 · 三评汇总成一份报告
+
+- 三位评审独立评完后，由一位“主考官”把三份意见汇总成一份报告：一个结论，一份优点、不足和下一稿建议，一份逐句修改，一篇参考改写；英一翻译每句一个分数、一条评语和一组采分点。重复的意见合并，有分歧的地方由主考官核对原文后取舍。
+- 分数仍由本机计算，汇总只整理文字、不改分：总分取三人的中位数；英一翻译每句取三人中位分，总分为各句之和（错别字扣分同样取中位）。
+- 评阅页只呈现这一份报告；三位评审的原始意见收进页面底部默认折叠的“原始评审意见”。复制的评审结果也是一份。
+- 汇总多一次请求（优先 DeepSeek，否则 Codex），可在 Settings 关闭。关闭或汇总失败时，由本机合并三份意见，评分不受影响。快速单评不汇总。
+
 # 1.7.0 · 英一翻译逐句得分与踩点给分
 
 - 英一翻译按大纲规则逐句评分：每句 2 分，拆成 3–4 个意群作为采分点，逐点给分；译文与原意明显不符的句子最多 0.5 分；同一处给出多种译法、其中有错的按错误给分；错别字整题累计满 3 个扣 0.5 分。准确（信）第一、通顺（达）第二，不要求“雅”。

WriteBench.xcodeproj/project.pbxproj

@@ -48,6 +48,7 @@
 		9EA1454709EC8802D43C7CA4 /* HistoryFolderEditor.swift in Sources */ = {isa = PBXBuildFile; fileRef = 419DE9ABE9E26F14128B9B1E /* HistoryFolderEditor.swift */; };
 		A104DD7F66FF512A80BA9717 /* QuestionImportTests.swift in Sources */ = {isa = PBXBuildFile; fileRef = 29453693001AD3ECDE2185B6 /* QuestionImportTests.swift */; };
 		A54BFAD371DEB925FCA1963B /* PracticeLoopTests.swift in Sources */ = {isa = PBXBuildFile; fileRef = 9F5E568498AF07D6BF8C1335 /* PracticeLoopTests.swift */; };
+		A64FFC28FE1B55A2B44B2E4B /* ReportSynthesis.swift in Sources */ = {isa = PBXBuildFile; fileRef = 99093337348356622254C44E /* ReportSynthesis.swift */; };
 		A840C857A17F4257AD96C9AF /* ReviewTextExporter.swift in Sources */ = {isa = PBXBuildFile; fileRef = 029A81F1C48B2FAE8914576B /* ReviewTextExporter.swift */; };
 		A97D07FFDB81393349741F89 /* PracticeView.swift in Sources */ = {isa = PBXBuildFile; fileRef = 84F1B7E5BB4C4D7689B346DE /* PracticeView.swift */; };
 		C118D4E53ACEC303DE46EA50 /* CorrectionMatcher.swift in Sources */ = {isa = PBXBuildFile; fileRef = 1020C698B4DA8C1C47852419 /* CorrectionMatcher.swift */; };
@@ -131,6 +132,7 @@
 		89F9C06884EE52989CB4491F /* EssayFolderMetadata.swift */ = {isa = PBXFileReference; lastKnownFileType = sourcecode.swift; path = EssayFolderMetadata.swift; sourceTree = "<group>"; };
 		8F892E2F452C5999A8C08043 /* OCRService.swift */ = {isa = PBXFileReference; lastKnownFileType = sourcecode.swift; path = OCRService.swift; sourceTree = "<group>"; };
 		9238EAE831BF69E5F210D72A /* WorkspaceView.swift */ = {isa = PBXFileReference; lastKnownFileType = sourcecode.swift; path = WorkspaceView.swift; sourceTree = "<group>"; };
+		99093337348356622254C44E /* ReportSynthesis.swift */ = {isa = PBXFileReference; lastKnownFileType = sourcecode.swift; path = ReportSynthesis.swift; sourceTree = "<group>"; };
 		9AA993471F8CFF92BD9B2956 /* BrandArt.swift */ = {isa = PBXFileReference; lastKnownFileType = sourcecode.swift; path = BrandArt.swift; sourceTree = "<group>"; };
 		9B600C2CBBB1B477ABB65C42 /* ImmersiveWritingView.swift */ = {isa = PBXFileReference; lastKnownFileType = sourcecode.swift; path = ImmersiveWritingView.swift; sourceTree = "<group>"; };
 		9F5E568498AF07D6BF8C1335 /* PracticeLoopTests.swift */ = {isa = PBXFileReference; lastKnownFileType = sourcecode.swift; path = PracticeLoopTests.swift; sourceTree = "<group>"; };
@@ -244,6 +246,7 @@
 				1020C698B4DA8C1C47852419 /* CorrectionMatcher.swift */,
 				CFD296324E8F63CD0B525458 /* GradingProvider.swift */,
 				634F608AF0DF72D3DE51E438 /* GradingService.swift */,
+				99093337348356622254C44E /* ReportSynthesis.swift */,
 				E6B358A2E1043646F5A65FA4 /* UsageCost.swift */,
 			);
 			path = Grading;
@@ -563,6 +566,7 @@
 				3FF816DF3394076A7EAFCA8E /* QuestionLibraryView.swift in Sources */,
 				6C3E278AD86905C7C0A63BCE /* QuestionText.swift in Sources */,
 				17236CF00A287A2737947138 /* QuestionTextReader.swift in Sources */,
+				A64FFC28FE1B55A2B44B2E4B /* ReportSynthesis.swift in Sources */,
 				042D9A8FAAC22D71BDF6F7E4 /* ReviewCard.swift in Sources */,
 				DB3DEEA68D82D9F8BECF108E /* ReviewOutline.swift in Sources */,
 				A840C857A17F4257AD96C9AF /* ReviewTextExporter.swift in Sources */,
@@ -659,7 +663,7 @@
 				CODE_SIGN_IDENTITY = "-";
 				CODE_SIGN_STYLE = Automatic;
 				COPY_PHASE_STRIP = NO;
-				CURRENT_PROJECT_VERSION = 12;
+				CURRENT_PROJECT_VERSION = 13;
 				DEBUG_INFORMATION_FORMAT = "dwarf-with-dsym";
 				ENABLE_HARDENED_RUNTIME = YES;
 				ENABLE_NS_ASSERTIONS = NO;
@@ -674,7 +678,7 @@
 				GCC_WARN_UNUSED_VARIABLE = YES;
 				GENERATE_INFOPLIST_FILE = YES;
 				MACOSX_DEPLOYMENT_TARGET = 15.0;
-				MARKETING_VERSION = 1.7.0;
+				MARKETING_VERSION = 1.8.0;
 				MTL_ENABLE_DEBUG_INFO = NO;
 				MTL_FAST_MATH = YES;
 				PRODUCT_NAME = "$(TARGET_NAME)";
@@ -755,7 +759,7 @@
 				CODE_SIGN_IDENTITY = "-";
 				CODE_SIGN_STYLE = Automatic;
 				COPY_PHASE_STRIP = NO;
-				CURRENT_PROJECT_VERSION = 12;
+				CURRENT_PROJECT_VERSION = 13;
 				DEBUG_INFORMATION_FORMAT = dwarf;
 				ENABLE_HARDENED_RUNTIME = YES;
 				ENABLE_STRICT_OBJC_MSGSEND = YES;
@@ -776,7 +780,7 @@
 				GCC_WARN_UNUSED_VARIABLE = YES;
 				GENERATE_INFOPLIST_FILE = YES;
 				MACOSX_DEPLOYMENT_TARGET = 15.0;
-				MARKETING_VERSION = 1.7.0;
+				MARKETING_VERSION = 1.8.0;
 				MTL_ENABLE_DEBUG_INFO = INCLUDE_SOURCE;
 				MTL_FAST_MATH = YES;
 				ONLY_ACTIVE_ARCH = YES;

WriteBench/Features/Review/BackgroundGradingView.swift

@@ -57,7 +57,7 @@ struct GradingProgressView: View {
             }.padding(24)
             ScrollView {
                 VStack(alignment: .leading, spacing: 16) {
-                    Text(job.phase == .completed ? (job.isQuick ? "快速单评已完成，结论与分数已保存到历史。" : "三位独立评审已完成，完整结论与总分已保存到历史。") : job.phase == .failed || job.phase == .cancelled ? "本次未生成总分。已收到的评语片段不作为最终评分，提交原稿保留在下方。" : job.isQuick ? "快速单评：只请一位评审，速度更快、花费更少，没有评审一致性。" : "进度按实际完成的评审计数。下方为实时评语，总分将在三位评审全部完成后生成。").font(.system(size: 12)).foregroundStyle(WB.secondary)
+                    Text(job.phase == .summarizing ? "三位评审已完成，正在把三份意见汇总成一份报告；分数已按中位数确定，汇总不改分。" : job.phase == .completed ? (job.isQuick ? "快速单评已完成，结论与分数已保存到历史。" : job.session?.report?.synthesis != nil ? "三位独立评审已完成，并已汇总成一份报告保存到历史。" : "三位独立评审已完成，完整结论与总分已保存到历史。") : job.phase == .failed || job.phase == .cancelled ? "本次未生成总分。已收到的评语片段不作为最终评分，提交原稿保留在下方。" : job.isQuick ? "快速单评：只请一位评审，速度更快、花费更少，没有评审一致性。" : "进度按实际完成的评审计数。下方为实时评语，总分将在三位评审全部完成后生成。").font(.system(size: 12)).foregroundStyle(WB.secondary)
                     ForEach(job.activeJudges) { judge in
                         Card(padding: 20) {
                             VStack(alignment: .leading, spacing: 12) {

WriteBench/Features/Review/ReviewOutline.swift

@@ -1,15 +1,15 @@
 import SwiftUI
 
 enum ReviewSection: String, CaseIterable, Identifiable {
-    case overview, progress, segments, feedback, examiners, corrections, improved, rewrite
+    case overview, progress, segments, feedback, corrections, improved, rewrite, examiners
     var id: String { rawValue }
     func title(isTranslation: Bool) -> String {
         switch self {
         case .overview: "总分与结论"
         case .progress: "与上一稿对比"
         case .segments: "逐句得分"
         case .feedback: "优点与不足"
-        case .examiners: "评审意见"
+        case .examiners: "原始评审意见"
         case .corrections: "逐句修改"
         case .improved: isTranslation ? "参考改译" : "改进作文"
         case .rewrite: "重写"

WriteBench/Features/Review/ReviewTextExporter.swift

@@ -7,26 +7,21 @@ import Foundation
         var sections = [
             "WriteBench · \(session.task.fullTitle)\n\(session.date.formatted(date: .abbreviated, time: .shortened))",
             "\(report.isDemo ? "演示评分 · " : "")最终得分：\(report.finalScore.scoreText) / \(Int(session.task.maxScore))\n" + (report.gradingMode == .quick ? "评阅方式：快速单评（一位评审）" : "置信度：\(report.confidence.rawValue)\n评审分差：\(report.spread.scoreText)"),
-            "评阅结论（中位分评审）\n\(report.conclusion)",
+            "评阅结论\n\(report.conclusion)",
             feedback("写得好的地方", report.strengths),
             feedback("不足的地方", report.weaknesses),
             feedback("下一稿怎么改", report.improvements),
             "评分维度（诊断分 / 10）\n\(session.task.isTranslation ? "译义与完整性" : "任务完成度")：\(report.dimension(\.taskCompletion).scoreText)\n语言：\(report.dimension(\.language).scoreText)\n连贯性：\(report.dimension(\.coherence).scoreText)\n语域：\(report.dimension(\.register).scoreText)"
         ]
         if !report.segmentScores.isEmpty {
-            sections.append("逐句得分（中位分评审）\n" + report.segmentScores.map { item in
+            sections.append("逐句得分\n" + report.segmentScores.map { item in
                 (["(\(item.segment.number)) \(item.segment.score.scoreText) / \(item.segment.maxScore.scoreText)　\(item.segment.comment)"]
                  + item.segment.points.map { "  · \($0.source)：\($0.earned.scoreText) / \($0.max.scoreText)\($0.note.isEmpty ? "" : "，" + $0.note)" }).joined(separator: "\n")
             }.joined(separator: "\n"))
         }
-        let reviewers = report.reviewers.map { result in
-            var lines = ["\(result.judge.title) · \(result.judge.role) · \(result.response.score.scoreText) / \(Int(session.task.maxScore))",
-                         "\(result.provider?.title ?? result.model) · \(result.model)", result.response.summary]
-            if !result.response.majorErrors.isEmpty { lines.append(feedback("主要问题", result.response.majorErrors)) }
-            if !result.response.minorErrors.isEmpty { lines.append(feedback("次要问题", result.response.minorErrors)) }
-            return lines.joined(separator: "\n")
+        if report.reviewers.count > 1 {
+            sections.append("三位评审独立评分：" + report.reviewers.map { "\($0.judge.title) \($0.response.score.scoreText)" }.joined(separator: "，") + "（取中位数）")
         }
-        sections.append((report.gradingMode == .quick ? "评审意见" : "三位评审的独立意见") + "\n\n" + reviewers.joined(separator: "\n\n"))
         let corrections = report.corrections.enumerated().map { index, correction in
             "\(index + 1). \(correction.category.rawValue) · \(correction.severity == .major ? "主要" : "次要")\n原句：\(correction.original)\n修改：\(correction.corrected)\n说明：\(correction.explanation)"
         }