@@ -0,0 +1,134 @@
+//! 概念模型卡命令层（v0.13.2 REQ-206~207 系统层）。
+//!
+//! @ai-context: 概念双面体单向升格——组内 model 卡（记忆面 front=概念名、back=三问）升格为
+//!              体系内概念（思辨面）；只引用不收纳（knowledge_links(target_type=flashcard)），
+//!              不新增 flashcards.concept_id 列。本层只做参数校验、编排数据层/纯函数、错误映射；
+//!              `fn xxx_inner(db, ...)` 为纯函数（:memory: 可测），薄 command 壳只取 state.db。
+//! @ai-context: 线宽豁免登记（AGENTS.md §3 300-600 带）：本文件承载创建/组列表两命令 + 共享
+//!              卡面/时刻工具，≤300 行；卡→概念升格（四分支 + PromoteResult）拆至
+//!              commands_knowledge_cards_promote.rs（同一命令域，避免单文件超限）。
+
+use tauri::State;
+
+use crate::commands::AppState;
+use crate::commands_knowledge::{normalize_text, require_id, NAME_MAX_CHARS};
+use crate::db::Db;
+use crate::db_flashcards::NewFlashcard;
+use crate::knowledge_card::format_model_card_back;
+use crate::scheduler::CardState;
+use crate::types::Flashcard;
+
+/// 组卡类型白名单（list_group_cards 的 kind 过滤；事实/动作/概念模型卡）。
+const KIND_WHITELIST: [&str; 3] = ["fact", "action", "model"];
+
+/// 当前时刻（Unix 毫秒；新卡 due_at 统一口径——与 commands_flashcards 相同）。
+///
+/// @ai-context: 新 model 卡 due_at=now（立即可复习——首次学习也是提取优先）。
+fn now_ms() -> u64 {
+    std::time::SystemTime::now()
+        .duration_since(std::time::UNIX_EPOCH)
+        .map(|d| d.as_millis() as u64)
+        .unwrap_or(0)
+}
+
+/// 卡面可选字段：trim + 空串转 None；非空字段 ≤2000 字符校验（超长报错）。
+///
+/// @ai-context: 卡面契约"缺问 → None"（非"有但为空"——前端按占位渲染）；超长拒绝防污染 DB。
+fn card_field(v: Option<String>) -> Result<Option<String>, String> {
+    match v {
+        Some(s) => {
+            let t = s.trim().to_string();
+            if t.is_empty() {
+                Ok(None)
+            } else if t.chars().count() > NAME_MAX_CHARS {
+                Err(format!("卡面字段超长（上限 {} 字符）", NAME_MAX_CHARS))
+            } else {
+                Ok(Some(t))
+            }
+        }
+        None => Ok(None),
+    }
+}
+
+/// 在组内建概念模型卡（kind='model'；幂等——同组同 front 已有则返回既有卡）。
+///
+/// @ai-context: front=归一化概念名、back=三问契约（format_model_card_back 组合）；组仍是唯一
+///              容器（spec §一）——卡挂组内，升格后才进体系。
+#[tauri::command]
+pub fn create_model_card(
+    state: State<'_, AppState>,
+    group_id: i64,
+    name: String,
+    essence: Option<String>,
+    boundary: Option<String>,
+    relation: Option<String>,
+) -> Result<Flashcard, String> {
+    create_model_card_inner(&state.db, group_id, name, essence, boundary, relation)
+}
+
+/// 组卡列表（kind 过滤可选；模型卡列表数据源）。
+#[tauri::command]
+pub fn list_group_cards(
+    state: State<'_, AppState>,
+    group_id: i64,
+    kind: Option<String>,
+) -> Result<Vec<Flashcard>, String> {
+    list_group_cards_inner(&state.db, group_id, kind)
+}
+
+/// 建概念模型卡（组存在 + 名归一化 + 卡面契约 + 幂等复用）。
+pub(crate) fn create_model_card_inner(
+    db: &Db,
+    group_id: i64,
+    name: String,
+    essence: Option<String>,
+    boundary: Option<String>,
+    relation: Option<String>,
+) -> Result<Flashcard, String> {
+    require_id(group_id)?;
+    if db.get_group(group_id).map_err(|e| e.to_string())?.is_none() {
+        return Err(format!("笔记组不存在: {}", group_id));
+    }
+    let name = normalize_text(&name, "概念名")?;
+    // 幂等：同组同 front（归一化名）已有 → 返回既有卡（不新建，防 UI 双击双卡）
+    if let Some(existing) = db.find_card_by_front(group_id, &name).map_err(|e| e.to_string())? {
+        return Ok(existing);
+    }
+    let essence = card_field(essence)?;
+    let boundary = card_field(boundary)?;
+    let relation = card_field(relation)?;
+    let back = format_model_card_back(essence.as_deref(), boundary.as_deref(), relation.as_deref());
+    let state_json = serde_json::to_string(&CardState::default()).unwrap_or_default();
+    db.create_card(&NewFlashcard {
+        group_id,
+        note_id: None,
+        fragment_id: None,
+        front: name,
+        back,
+        kind: "model".to_string(),
+        state_json,
+        due_at: now_ms() as i64,
+    })
+    .map_err(|e| e.to_string())
+}
+
+/// 组卡列表（require_id + kind 白名单校验）。
+pub(crate) fn list_group_cards_inner(
+    db: &Db,
+    group_id: i64,
+    kind: Option<String>,
+) -> Result<Vec<Flashcard>, String> {
+    require_id(group_id)?;
+    let kind = kind.as_deref().filter(|k| !k.trim().is_empty());
+    if let Some(k) = kind {
+        if !KIND_WHITELIST.contains(&k) {
+            return Err(format!("不支持的卡片类型: {}（支持: {}）", k, KIND_WHITELIST.join("/")));
+        }
+    }
+    db.list_cards_by_group(group_id, kind).map_err(|e| e.to_string())
+}
+
+/// 命令层单测独立文件（保持本文件 ≤300 行，AGENTS.md §3）。
+#[cfg(test)]
+#[path = "commands_knowledge_cards_tests.rs"]
+mod tests;
@@ -0,0 +1,193 @@
+//! 概念模型卡→概念升格命令层（v0.13.2 REQ-207 系统层；commands_knowledge_cards 拆出）。
+//!
+//! @ai-context: 卡→概念升格四分支（Create/Merge/Hint/Already）——组内 model 卡经 promote_rules
+//!              决策升格为体系内概念（思辨面）；只引用不收纳（knowledge_links(target_type=flashcard)）；
+//!              锚点为卡背独立行 `→ 概念「name」`（back_has_anchor 幂等防重复回链）。
+//! @ai-context: 线宽豁免登记：本文件承载升格命令 + PromoteResult + 四分支编排（同一命令域），
+//!              自 commands_knowledge_cards.rs 拆出以保持各文件 ≤300 行（AGENTS.md §3）。
+
+use tauri::State;
+
+use serde::Serialize;
+
+use crate::commands::AppState;
+use crate::commands_knowledge::{normalize_text, require_id};
+use crate::commands_knowledge_core::link_knowledge_target_inner;
+use crate::db::Db;
+use crate::knowledge_card::{back_has_anchor, parse_model_card_back};
+use crate::knowledge_pure::{PromoteDecision, PromoteInput};
+use crate::types::{KnowledgeConcept, KnowledgeLink, KnowledgeSystem, NewKnowledgeConcept};
+
+/// 升格结果（camelCase 序列化回传前端）。
+///
+/// @ai-context: action ∈ created/merged/hinted/already；decision 透传 promote_rules 决策；
+///              concept=落库/命中的概念；link=新建引用（created/merged）或 None
+///              （hinted/already——未落库/既有引用）。
+#[derive(Debug, Clone, Serialize)]
+#[serde(rename_all = "camelCase")]
+pub struct PromoteResult {
+    pub action: String,
+    pub decision: PromoteDecision,
+    pub concept: KnowledgeConcept,
+    pub link: Option<KnowledgeLink>,
+}
+
+/// 卡→概念升格（Create/Merge/Hint/Already 四分支）。
+///
+/// @ai-context: target_system_id 缺省取全局体系（无全局 → 引导先建）；前置校验卡存在且
+///              kind='model'（非概念卡拒绝）；空名在 normalize_text 拒空（promote_rules panic 前置）。
+#[tauri::command]
+pub fn promote_card_to_concept(
+    state: State<'_, AppState>,
+    card_id: i64,
+    target_system_id: Option<i64>,
+) -> Result<PromoteResult, String> {
+    promote_card_to_concept_inner(&state.db, card_id, target_system_id)
+}
+
+/// 卡→概念升格（规格 §四 5 步：卡校验→目标体系→AlreadyLinked 免重复→promote_rules→分派）。
+pub(crate) fn promote_card_to_concept_inner(
+    db: &Db,
+    card_id: i64,
+    target_system_id: Option<i64>,
+) -> Result<PromoteResult, String> {
+    require_id(card_id)?;
+    let card = db
+        .get_card(card_id)
+        .map_err(|e| e.to_string())?
+        .ok_or_else(|| format!("卡片不存在: {}", card_id))?;
+    if card.kind != "model" {
+        return Err("仅概念卡（model 卡）可纳入体系".to_string());
+    }
+    let target = resolve_target_system(db, target_system_id)?;
+    // AlreadyLinked 免重复：目标体系内该卡已有闪卡引用 → 直接返回既有关联（不重跑）
+    let links = db
+        .list_knowledge_links(target.id, None, None, None)
+        .map_err(|e| e.to_string())?;
+    if let Some(link) = links.iter().find(|l| {
+        l.target_type == "flashcard" && l.target_id == card_id && l.concept_id.is_some()
+    }) {
+        let cid = link.concept_id.ok_or_else(|| "引用缺少概念".to_string())?;
+        let concept = db
+            .get_knowledge_concept(cid)
+            .map_err(|e| e.to_string())?
+            .ok_or_else(|| format!("概念不存在: {}", cid))?;
+        return Ok(PromoteResult {
+            action: "already".to_string(),
+            decision: PromoteDecision::AlreadyLinked { concept_id: cid },
+            concept,
+            link: Some(link.clone()),
+        });
+    }
+    // 组装 PromoteInput：name 归一化前置（拒空——promote_rules 空名 panic 前置）
+    let name = normalize_text(&card.front, "概念名")?;
+    let existing: Vec<(i64, String, i64)> = db
+        .find_concept_by_name(&name)
+        .map_err(|e| e.to_string())?
+        .map(|c| (c.id, c.name, c.system_id))
+        .into_iter()
+        .collect();
+    let input = PromoteInput {
+        card_name: name.clone(),
+        existing,
+        target_system_id: target.id,
+    };
+    let decision = crate::knowledge_pure::promote_rules(&input);
+    match decision {
+        PromoteDecision::Create => {
+            let parsed = parse_model_card_back(&card.back);
+            let concept = db
+                .add_knowledge_concept(&NewKnowledgeConcept {
+                    system_id: target.id,
+                    name: name.clone(),
+                    essence: parsed.essence,
+                    boundary: parsed.boundary,
+                    relation: parsed.relation,
+                })
+                .map_err(|e| {
+                    let msg = e.to_string();
+                    if msg.contains("UNIQUE") {
+                        "概念已存在，请合并".to_string()
+                    } else {
+                        msg
+                    }
+                })?;
+            let link = link_knowledge_target_inner(
+                db, target.id, None, Some(concept.id), None, "flashcard".to_string(), card_id,
+            )?;
+            append_anchor(db, card_id, &card.back, &name)?;
+            record_promoted_metric(db, card_id, concept.id);
+            Ok(PromoteResult { action: "created".to_string(), decision, concept, link: Some(link) })
+        }
+        PromoteDecision::Merge { concept_id } => {
+            let concept = db
+                .get_knowledge_concept(concept_id)
+                .map_err(|e| e.to_string())?
+                .ok_or_else(|| format!("概念不存在: {}", concept_id))?;
+            let link = link_knowledge_target_inner(
+                db, target.id, None, Some(concept.id), None, "flashcard".to_string(), card_id,
+            )?;
+            append_anchor(db, card_id, &card.back, &name)?;
+            record_promoted_metric(db, card_id, concept.id);
+            Ok(PromoteResult { action: "merged".to_string(), decision, concept, link: Some(link) })
+        }
+        PromoteDecision::CrossSystemHint { concept_id, other_system_id: _ } => {
+            // 仅提示不落库（v0.13.4 交叉点数据源）；无 link/无 metric/无锚点追加
+            let concept = db
+                .get_knowledge_concept(concept_id)
+                .map_err(|e| e.to_string())?
+                .ok_or_else(|| format!("概念不存在: {}", concept_id))?;
+            Ok(PromoteResult { action: "hinted".to_string(), decision, concept, link: None })
+        }
+        PromoteDecision::AlreadyLinked { concept_id } => {
+            // 防御分支（步骤 3 已拦截，基本不达；仍透传既有关联）
+            let concept = db
+                .get_knowledge_concept(concept_id)
+                .map_err(|e| e.to_string())?
+                .ok_or_else(|| format!("概念不存在: {}", concept_id))?;
+            let link = db
+                .list_knowledge_links(target.id, None, None, None)
+                .map_err(|e| e.to_string())?
+                .into_iter()
+                .find(|l| {
+                    l.target_type == "flashcard"
+                        && l.target_id == card_id
+                        && l.concept_id == Some(concept_id)
+                });
+            Ok(PromoteResult { action: "already".to_string(), decision, concept, link })
+        }
+    }
+}
+
+/// 目标体系解析：传则校验存在性；未传取全局体系（无→引导先建全局体系）。
+fn resolve_target_system(db: &Db, target_system_id: Option<i64>) -> Result<KnowledgeSystem, String> {
+    match target_system_id {
+        Some(tid) => {
+            require_id(tid)?;
+            db.get_knowledge_system(tid)
+                .map_err(|e| e.to_string())?
+                .ok_or_else(|| format!("体系不存在: {}", tid))
+        }
+        None => db
+            .find_global_system()
+            .map_err(|e| e.to_string())?
+            .ok_or_else(|| "请先创建全局体系".to_string()),
+    }
+}
+
+/// 升格锚点追加（幂等：back_has_anchor 已锚定则不二次追加）。
+fn append_anchor(db: &Db, card_id: i64, back: &str, name: &str) -> Result<(), String> {
+    if back_has_anchor(back) {
+        return Ok(());
+    }
+    let new_back = format!("{}\n→ 概念「{}」", back, name);
+    db.update_card_back(card_id, &new_back).map_err(|e| e.to_string())?;
+    Ok(())
+}
+
+/// 埋 concept_promoted 指标（best-effort——与 card_reviewed/fragment_upgraded 同口径，
+/// Payload 带 cardId/conceptId 供前端与后续体系统计）。
+fn record_promoted_metric(db: &Db, card_id: i64, concept_id: i64) {
+    let payload = serde_json::json!({ "cardId": card_id, "conceptId": concept_id }).to_string();
+    let _ = db.add_metric_event("concept_promoted", &payload);
+}