refactor(rust): 拆 commands_ai_refine 步 2 — 工作台域（392 行外壳）

Author: Aparencia

app/src-tauri/src/app_commands.rs

@@ -400,7 +400,7 @@ pub fn handle() -> impl Fn(tauri::ipc::Invoke<tauri::Wry>) -> bool + Send + Sync
         crate::commands_ai_refine::ai_refine_prompt_preview,
         crate::commands_ai_refine::ai_refine_result,
         crate::commands_ai_refine::ai_refine_apply,
-        crate::commands_ai_refine::refine_workbench,
+        crate::commands_ai_refine::workbench::refine_workbench,
         // v0.17.0（REQ-246）：笔记级 AI 精修——估计/启动/采纳（手写笔记）
         crate::commands_ai_note_refine::ai_note_refine_estimate,
         crate::commands_ai_note_refine::ai_note_refine_start,

app/src-tauri/src/commands_ai_refine/mod.rs

@@ -16,6 +16,7 @@
 //!              · dto.rs        —— IPC 契约类型（4 个 serde 结构体；字段序 = JSON 键序）
 //!              · gate.rs       —— 成本硬拦截（免费档/本地 Provider 跳过；查询失败宽容放行）
 //!              · registry.rs   —— 任务注册表 + id 序列 + 容量守卫 + 状态写入口
+//!              · workbench.rs  —— 精修工作台（三级数据源 + 章节 diff 统计 + 内联契约测试）
 //!              @ai-context: 本外壳只留「跨域共享的命令 + 重导出门面」；外部 10 文件
 //!              20 处 crate::commands_ai_refine::X 引用**零改动**（逐项重导出在下方）；
 //!              #[tauri::command] 定义随域下沉后，注册路径同步改成
@@ -24,6 +25,7 @@
 mod dto;
 mod gate;
 mod registry;
+pub(crate) mod workbench;
 
 pub use dto::{AiRefineResult, AiTaskHandle, RefineEstimateView, RefineStrategyInfo};
 pub use registry::{claim_task_id, task_registry, task_seq, task_seq_lower_bound, AiTaskEntry};
@@ -36,12 +38,9 @@ use crate::ai_cost::estimate_for_content_model;
 use crate::ai_note_refine::NoteRefinePrompt;
 use crate::ai_strategy::{RefineStrategyMeta, StrategyOverride};
 use crate::ai_task::AiTaskState;
-use crate::anchor_strip::strip_anchors;
 use crate::commands::AppState;
 use crate::commands_session_note::build_rule_draft_with_analysis;
-use crate::note_diff::{diff_sections, DiffStats, SectionDiff};
 use crate::note_filter::PurifyEnv;
-use crate::note_version::VersionMeta;
 use crate::types::{NewNote, Note};
 use crate::video_profile::ProfileKind;
 
@@ -391,192 +390,3 @@ pub fn ai_refine_apply(
     crate::notify::emit_changed(&state.app, crate::notify::DataDomain::Notes);
     Ok(note_row)
 }
-
-// ────────────────────────────────────────────────────────────
-// 精修工作台（Task 11 / spec 6️⃣——规则草稿+精修结果+章节分组 diff 一次取全）
-// ────────────────────────────────────────────────────────────
-
-/// 工作台数据（前端 RefineWorkbench 模态数据源）。
-///
-/// @ai-context: 必须 camelCase（与 AiRefineResult 等兄弟结构体一致）——本模块
-///              其余结构体均已 2026-08 统一契约；此处缺失曾导致前端按
-///              ruleMarkdown 读取得到 undefined → renderMd(undefined).split
-///              （"Cannot read properties of undefined (reading 'split')"，
-///              v0.11.5 引入、真机验收漏测，v0.12.3 修复补测）。
-/// @ai-context: 嵌套 section（SectionDiff）保持 snake_case 是刻意契约
-///              （首次出现即定，前端类型注释"勿改"），rename 只作用于顶层键。
-#[derive(Debug, Clone, serde::Serialize)]
-#[serde(rename_all = "camelCase")]
-pub struct WorkbenchData {
-    pub rule_markdown: String,
-    pub refined_markdown: Option<String>,
-    pub sections: Vec<SectionDiff>,
-    pub stats: DiffStats,
-    pub meta: Option<VersionMeta>,
-}
-
-/// 工作台数据接口：规则草稿 + 精修结果 + 章节分组 diff 一次取全。
-///
-/// refined_markdown 为 None → 尚未精修；否则包含最新精修版内容与章节 diff。
-///
-/// @ai-context: 精修版三级数据源（修复：原实现只查已落库笔记——采纳前打开
-///              工作台右侧恒空；且任务成功事件先于 DB 落库，存在竞态）：
-///              ① refine_result 参数（调用方内存结果——采纳前刚完成的任务，
-///                 消除竞态）② 最新未采纳成功任务（DB 持久化——重启后恢复）
-///              ③ 已落库笔记最新版本（采纳后）。
-#[tauri::command]
-pub async fn refine_workbench(
-    state: State<'_, AppState>,
-    session_id: i64,
-    refine_result: Option<AiRefineResult>,
-) -> Result<WorkbenchData, String> {
-    if session_id <= 0 {
-        return Err("无效的会话 id".to_string());
-    }
-    // ① 构建规则草稿（与精修任务同管线）
-    let env = PurifyEnv {
-        config: state.purify.clone(),
-        symbol: state.symbol_normalize.clone(),
-        corrections: state.ocr_corrections.clone(),
-    };
-    let (draft, _) = build_rule_draft_with_analysis(
-        &state.db,
-        &state.ui_junk,
-        &env,
-        &state.data_dir,
-        session_id,
-        None,
-    )
-    .map_err(|e| e.to_string())?;
-    let rule_md = strip_anchors(&draft.markdown);
-
-    // ② 精修版数据源：内存结果（优先）＞ 未采纳成功任务（DB 兜底）＞ 已落库笔记
-    let unadopted = state
-        .db
-        .find_latest_unadopted_refine(session_id)
-        .map_err(|e| e.to_string())?;
-    let pending_result: Option<AiRefineResult> = match (refine_result, unadopted) {
-        (Some(r), _) => Some(r),
-        // 解析失败降级（日志可观测——不影响工作台打开，可经任务中心重取）
-        (None, Some(task)) => task.result_json.as_deref().and_then(|j| {
-            serde_json::from_str::<AiRefineResult>(j)
-                .map_err(|e| eprintln!("[refine-workbench] 未采纳任务 {} 结果解析失败: {}", task.task_id, e))
-                .ok()
-        }),
-        (None, None) => None,
-    };
-
-    let note = state.db.find_note_by_session(session_id).map_err(|e| e.to_string())?;
-    let (refined_md, sections, stats, meta) = if let Some(result) = pending_result {
-        let secs = diff_sections(&rule_md, &result.refined_markdown);
-        let st = stats_from(&secs);
-        // 未落库：成本按模型单价在 apply 时核算——此处不做虚假回填
-        let m = VersionMeta {
-            cost_yuan: None,
-            model: Some(result.model.clone()),
-            slices: Some(result.slices),
-            merged_from: None,
-        };
-        (Some(result.refined_markdown), secs, st, Some(m))
-    } else if let Some(ref note) = note {
-        // 取最新版本内容作为精修版
-        let versions = state.db.list_versions(note.id).map_err(|e| e.to_string())?;
-        let latest = versions.last()
-            .map(|v| v.content.clone())
-            .unwrap_or_else(|| note.content.clone());
-        let secs = diff_sections(&rule_md, &latest);
-        let st = stats_from(&secs);
-        // 取最新版本 meta
-        let m = versions.last().map(|v| VersionMeta {
-            cost_yuan: v.meta.cost_yuan,
-            model: v.meta.model.clone(),
-            slices: v.meta.slices,
-            merged_from: v.meta.merged_from.clone(),
-        });
-        (Some(latest), secs, st, m)
-    } else {
-        (None, vec![], DiffStats { added: 0, removed: 0, unchanged: 0 }, None)
-    };
-
-    Ok(WorkbenchData {
-        rule_markdown: rule_md,
-        refined_markdown: refined_md,
-        sections,
-        stats,
-        meta,
-    })
-}
-
-/// 章节分组 diff 统计（新增/删除/未变行数——与 diff_sections 同口径）。
-fn stats_from(secs: &[SectionDiff]) -> DiffStats {
-    DiffStats {
-        added: secs.iter().map(|s| s.added_lines.len()).sum(),
-        removed: secs.iter().map(|s| s.removed_lines.len()).sum(),
-        unchanged: secs.iter()
-            .filter(|s| s.status == crate::note_diff::DiffStatus::Unchanged)
-            .count(),
-    }
-}
-
-#[cfg(test)]
-mod tests {
-    use super::*;
-
-    /// v0.12.3 回归（Bug#2）：WorkbenchData 顶层键必须 camelCase——
-    /// 前端按 ruleMarkdown/refinedMarkdown 读取；snake_case 键使字段为
-    /// undefined → renderMd(undefined).split 抛
-    /// "Cannot read properties of undefined (reading 'split')"
-    /// （原实现缺失 rename_all，v0.11.5 引入、真机验收漏测）。
-    #[test]
-    fn workbench_data_serializes_camel_case_top_level() {
-        let data = WorkbenchData {
-            rule_markdown: "# 标题\n正文".to_string(),
-            refined_markdown: Some("# 标题\n精修正文".to_string()),
-            sections: vec![crate::note_diff::SectionDiff {
-                heading: "标题".to_string(),
-                status: crate::note_diff::DiffStatus::Modified,
-                removed_lines: vec!["正文".to_string()],
-                added_lines: vec!["精修正文".to_string()],
-            }],
-            stats: crate::note_diff::DiffStats {
-                added: 1,
-                removed: 1,
-                unchanged: 0,
-            },
-            meta: Some(crate::note_version::VersionMeta {
-                cost_yuan: Some(0.01),
-                model: Some("test-model".to_string()),
-                slices: Some(1),
-                merged_from: None,
-            }),
-        };
-        let v = serde_json::to_value(&data).expect("序列化失败");
-        let obj = v.as_object().expect("应为 JSON 对象");
-        assert!(obj.contains_key("ruleMarkdown"), "顶层键必须为 ruleMarkdown（camelCase）");
-        assert!(obj.contains_key("refinedMarkdown"), "顶层键必须为 refinedMarkdown（camelCase）");
-        assert!(obj.contains_key("sections"));
-        assert!(obj.contains_key("stats"));
-        assert!(obj.contains_key("meta"));
-        // 嵌套 SectionDiff 保持 snake_case（首次出现即定的契约，勿随顶层 rename）
-        let sec = &obj["sections"][0];
-        assert!(sec.get("removed_lines").is_some(), "SectionDiff 嵌套键保持 snake_case");
-        assert!(sec.get("added_lines").is_some());
-        assert_eq!(sec["status"], "modified");
-    }
-
-    /// stats_from：按 diff_sections 分组行数统计（工作台头部 新增/删除/未变 徽标）。
-    #[test]
-    fn stats_from_aggregates_section_lines() {
-        use crate::note_diff::{DiffStatus, SectionDiff};
-        let secs = vec![
-            SectionDiff { heading: "A".into(), status: DiffStatus::Modified, removed_lines: vec!["a".into()], added_lines: vec!["A".into(), "B".into()] },
-            SectionDiff { heading: "B".into(), status: DiffStatus::Unchanged, removed_lines: vec![], added_lines: vec![] },
-            SectionDiff { heading: "C".into(), status: DiffStatus::Added, removed_lines: vec![], added_lines: vec!["c".into()] },
-            SectionDiff { heading: "D".into(), status: DiffStatus::Removed, removed_lines: vec!["d".into()], added_lines: vec![] },
-        ];
-        let s = stats_from(&secs);
-        assert_eq!(s.added, 3);
-        assert_eq!(s.removed, 2);
-        assert_eq!(s.unchanged, 1);
-    }
-}