V10.530 暫停 recover-stale 主操作入口

2026-06-01 01:29:22 +08:00
parent 4557533c52
commit 0fed025c58
10 changed files with 66 additions and 106 deletions
--- a/TODO_NEXT_STEPS.txt
+++ b/TODO_NEXT_STEPS.txt
@@ -4,6 +4,7 @@
 ================================================================================

 【已完成】
+   - V10.530 輕量化 PChome 狀態 preview 並暫停 `recover-stale` 主操作入口：`_fetch_retryable_candidate_skus()` 先從最新 `competitor_match_attempts` 縮小可重評候選，再用 `JOIN LATERAL` 只取該 SKU 最新 MOMO 價，避免 `/api/ai/pchome-match/backfill/status` 因 `price_records` 全量 window scan 超時；正式 smoke 同時顯示過期 identity fresh-search rescue 5 筆耗時約 109 秒且 0 筆成功，因此 Dashboard 移除「救援過期 40 筆」按鈕，只保留 `stale_recovery_preview` 的只讀「可救援」觀測；後端 `/api/ai/pchome-match/recover-stale` 改由 `PCHOME_STALE_RECOVERY_ENABLED=true` 顯式開關才可執行，避免操作員誤按低成功率慢路徑拖住 worker。
   - V10.529 補強 `recover-stale` 名稱風險擋詞：過期 identity 搜尋救援會先排除 `+`、`x2`、`*2` 等組合暗示，以及湛藍、麋香、海洋、玫瑰、薰衣草、生理呵護、日用型、清爽、潤澤等常見變體 / 香味 / 版本詞，避免同品牌同規格但不同香味、不同膚感、不同使用情境的 stale pair 進慢速 fresh search。
   - V10.528 將 `recover-stale` 救援 preview 改成輕量雙階段篩選：SQL 從過期 `competitor_prices` 小集合出發，只做 identity_v2、過期、exact/total_price/price_alert_exact 等必要條件並限制候選量，再用 `JOIN LATERAL` 取 ACTIVE 商品最新 MOMO 價；variant / catalog / commercial condition / 高風險名稱訊號改在 Python 對小樣本過濾，避免正式站看板狀態端點因全量 price_records、JSONB + regex 過重查詢拖垮 app worker。
   - V10.527 收斂 PChome 過期 identity 搜尋救援隊列：`recover-stale` 不再直接吃全部過期 `identity_v2`，改走 `_fetch_expired_identity_recovery_skus()`，只收既有正式診斷為 `exact_identity / total_price / price_alert_exact` 且無 variant、catalog、commercial condition、count、bundle、unit-price 等阻擋理由的舊配對；名稱含任選、多款、香味、色號、即期、融燭燈、香氛蠟燭等高風險訊號也先排除，避免慢速 fresh search 把人工覆核型 stale pair 全部掃進來。
--- a/config.py
+++ b/config.py
@@ -402,7 +402,7 @@ YOUTUBE_API_KEY = os.getenv('YOUTUBE_API_KEY', '')
 # ==========================================
 # 系統版本與路徑
 # ==========================================
-SYSTEM_VERSION = "V10.529"
+SYSTEM_VERSION = "V10.530"
 LOG_FILE_PATH = os.path.join(BASE_DIR, 'logs/system.log')
 public_url = PUBLIC_URL  # 用於模板顯示

--- a/docs/AI_INTELLIGENCE_MODULE_SOT.md
+++ b/docs/AI_INTELLIGENCE_MODULE_SOT.md
@@ -90,7 +90,7 @@ SQL漏斗(~300筆)
 - 配對來源仍以 PChome crawler 真實搜尋結果為準；無競品資料時不生成挑品。
 - 比對覆蓋率補強入口：`POST /api/ai/pchome-match/backfill`，優先補抓仍無有效 PChome 配對的高價 ACTIVE 商品，完成後自動重算 AI 挑品清單。
 - 過期價格刷新入口：`POST /api/ai/pchome-match/refresh-stale`，只針對已建立 `identity_v2` 但 `expires_at` 過期的 PChome product_id 執行 `run_expired_identity_refresh()`；不得跑 fresh search recovery，不得呼叫 LLM，完成後重算 AI 挑品並清除 Dashboard / competitor intel cache。
- 過期 identity 搜尋救援入口：`POST /api/ai/pchome-match/recover-stale` 僅供操作員手動觸發，對已過期 `identity_v2` 先走既有 PChome product_id refresh；只有舊 ID 查無商品或重評低於門檻時，才允許受控 fresh search recovery。救援隊列必須先排除 variant、catalog、commercial condition、count、bundle、unit-price 與任選 / 多款 / 香味 / 色號 / 即期 / 融燭燈 / 香氛蠟燭 / `+` / `xN` / `*N` / 具名香味或膚感版本等高風險名稱訊號。這條路徑可抓 PChome，但不得呼叫 LLM；正式寫入仍必須通過 matcher、hard veto、auto price write safety 與 overwrite protection。
+- 過期 identity 搜尋救援入口：`POST /api/ai/pchome-match/recover-stale` 預設必須關閉主操作入口，僅保留只讀 preview；正式 smoke 顯示小批次成功率不足且耗時偏高時，不得在 Dashboard 顯示日常操作按鈕。若需操作員手動執行，必須先明確設定 `PCHOME_STALE_RECOVERY_ENABLED=true`，再對已過期 `identity_v2` 先走既有 PChome product_id refresh；只有舊 ID 查無商品或重評低於門檻時，才允許受控 fresh search recovery。救援隊列必須先排除 variant、catalog、commercial condition、count、bundle、unit-price 與任選 / 多款 / 香味 / 色號 / 即期 / 融燭燈 / 香氛蠟燭 / `+` / `xN` / `*N` / 具名香味或膚感版本等高風險名稱訊號。這條路徑可抓 PChome，但不得呼叫 LLM；正式寫入仍必須通過 matcher、hard veto、auto price write safety 與 overwrite protection。
 - 補抓狀態入口：`GET /api/ai/pchome-match/backfill/status` 除背景任務狀態外，必須回傳 read-only coverage snapshot：`active_with_price` / `valid_matches` / `match_rate` / `fresh_matches` / `fresh_match_rate` / `stale_matches` / `pending` / `actionable_review_count`，供 Dashboard 顯示目前該刷新過期價格或補抓未搜尋商品；此端點不寫 DB、不呼叫 LLM、不抓外站。
 - 排程閉環：`run_pchome_match_backfill_task` 每日 10:30 執行，補抓 PChome 待比對商品、寫入歷史價格，再重算 `strategy='product_pick'` 清單。
 - PChome / MOMO 競價摘要出口 `services/competitor_intel_repository.py` 使用 30 分鐘共享快取（`COMPETITOR_INTEL_CACHE_TTL_SECONDS` 可調），避免 `/growth_analysis`、`/daily_sales`、PPT/AI 報表每次請求重跑昂貴覆蓋率與價差趨勢查詢；`run_competitor_price_feeder_task` 與 PChome backfill 完成後會主動清除快取。快取只包摘要輸出，不改 matcher 的高信心門檻與 identity_v2 準確性規則。
@@ -98,7 +98,7 @@ SQL漏斗(~300筆)
 - PChome re-score 回收線：`rescore_accepted_current` 只能表示最新版 matcher 判定「可人工採用」，不可直接寫入正式 `competitor_prices`；`fetch_competitor_coverage()` 必須輸出 `rescore_accepted_count`，Dashboard、daily/growth 與 OpenClaw 競品摘要都要把「重算可採用待審」獨立呈現，避免和一般低信心/單位價覆核混在一起。
 - PChome 低信心操作分流：Dashboard 與 read-only `/api/pchome-review/queue` 必須把近門檻可救、證據不足、低信心舊候選拆成 `recoverable_low_score`、`true_low_confidence`、`legacy_low_score` 三個可篩選桶；廣義 `low_score` 僅作 repository/export 相容查詢，不可在 UI 中冒充單一操作分流。
 - `run_retryable_candidate_revalidation()` 的自動回刷主戰場仍限 `low_score` / `refresh_low_score` / `recoverable_low_score`；`true_low_confidence` 只有在已補 focused exact 規則的窄範圍品線、舊分數 >= 0.95、`comparison_mode='exact_identity'`、含 `strong_exact_spec_match` 且不含 commercial / variant / count / bundle / refill 等阻擋理由時，才可進入重評，不得全面打開人工審核池。
- `/api/ai/pchome-match/backfill/status` 必須把近門檻重評池與過期 identity 救援池以只讀 `revalidation_preview` / `stale_recovery_preview` 曝光給操作員；預覽只復用正式候選 SQL 並受 limit / 60 秒快取限制，不啟動 PChome 搜尋、不呼叫 LLM、不寫 `competitor_match_attempts` / `competitor_prices`。救援 preview 必須從過期 `competitor_prices` 小集合出發並用 `JOIN LATERAL` 取最新 MOMO 價，不得掃全量 `price_records`；其中 `review_gated_count` 僅代表窄門 `true_low_confidence` exact 候選，不得被解讀為全量人工池可自動回刷。
+- `/api/ai/pchome-match/backfill/status` 必須把近門檻重評池與過期 identity 救援池以只讀 `revalidation_preview` / `stale_recovery_preview` 曝光給操作員；預覽只復用正式候選 SQL 並受 limit / 60 秒快取限制，不啟動 PChome 搜尋、不呼叫 LLM、不寫 `competitor_match_attempts` / `competitor_prices`。重評 preview 必須先從最新 `competitor_match_attempts` 縮小候選，再用 `JOIN LATERAL` 取單一最新 MOMO 價；救援 preview 必須從過期 `competitor_prices` 小集合出發並用 `JOIN LATERAL` 取最新 MOMO 價，兩者都不得掃全量 `price_records`；Dashboard 只能顯示「可救援」觀測值，不得在未開啟 `PCHOME_STALE_RECOVERY_ENABLED` 時提供 recover-stale 執行按鈕；其中 `review_gated_count` 僅代表窄門 `true_low_confidence` exact 候選，不得被解讀為全量人工池可自動回刷。
 - PChome re-score audit 預設必須先取每個 SKU 的最新 `competitor_match_attempts` 狀態，再套用 status / reason 篩選；舊低信心歷史候選只能透過 `--include-historical-candidates` 明確進入考古掃描，避免已入隊、已否決或已修正 SKU 被舊紀錄重新推回報表。
 - production re-score `--apply-accepted` 僅可追加 `rescore_accepted_current` attempt 給人工覆核；執行後需清除 Dashboard / competitor intel cache，且必須抽查 `competitor_prices` / `competitor_price_history` 未新增正式價差。
 - production re-score 若曾把 `variant_selection_review` 追加成 `rescore_accepted_current`，必須用 `audit_competitor_match_attempt_rescore.py --retract-variant-accepted` 追加最新 `true_low_confidence` 退回列；此路徑只寫 `competitor_match_attempts`，不得刪歷史紀錄，也不得寫 `competitor_prices` / `competitor_price_history`。
--- a/docs/memory/history_logs.md
+++ b/docs/memory/history_logs.md
@@ -13,6 +13,7 @@
 ## 📅 詳細更新日誌 (考古存檔)

 ### 2026-06-01：PChome 比價新鮮度操作閉環
+- **V10.530 retryable preview 輕量化與 recover-stale 安全閘**: 正式站 profiling 證實 `/api/ai/pchome-match/backfill/status` 剩餘瓶頸在 `revalidation_preview`，約 12 秒；`_fetch_retryable_candidate_skus()` 改為先取每個 SKU 最新 attempt 並縮小到可重評候選，再 `JOIN LATERAL` 取單一最新 MOMO 價，不再對全量 `price_records` 做商品 window scan。正式 smoke 也顯示過期 identity fresh-search rescue 小批次 5 筆耗時約 109 秒且 0 筆成功，因此 Dashboard 移除「救援過期 40 筆」按鈕，只保留 `stale_recovery_preview` 只讀觀測；後端 `/api/ai/pchome-match/recover-stale` 保留但需明確設定 `PCHOME_STALE_RECOVERY_ENABLED=true` 才能執行，避免低成功率慢任務拖住正式 worker。
 - **V10.529 recover-stale 名稱風險擋詞補強**: 過期 identity 搜尋救援新增 `+`、`x2`、`*2` 等組合暗示，以及湛藍、麋香、海洋、玫瑰、薰衣草、生理呵護、日用型、清爽、潤澤等常見變體 / 香味 / 版本詞；避免同品牌同規格但不同香味、不同膚感或不同使用情境的 stale pair 進慢速 fresh search。
 - **V10.528 recover-stale preview 輕量化**: V10.527 的救援隊列在正式站 preview 時曾造成 status API 超時。改為雙階段篩選：SQL 從過期 `competitor_prices` 小集合出發，只做 identity_v2、過期、exact/total_price/price_alert_exact 等必要條件並限制候選量，再用 `JOIN LATERAL` 取 ACTIVE 商品最新 MOMO 價；variant / catalog / commercial condition / 高風險名稱訊號改在 Python 對小樣本過濾，避免 `/api/ai/pchome-match/backfill/status` 因全量 price_records、JSONB + regex preview 查詢拖垮。
 - **V10.527 PChome 過期 identity 搜尋救援隊列收斂**: V10.526 production smoke 發現直接對全部過期 `identity_v2` 做 rescue 會把香氛 / 色號 / 目錄款 / 商業狀態差異等人工覆核型 stale pair 送進慢速 fresh search，20 筆耗時 361 秒且 0 筆成功。新增 `_fetch_expired_identity_recovery_skus()` 作為救援專用隊列，只收既有正式診斷為 `exact_identity / total_price / price_alert_exact` 且無 variant、catalog、commercial condition、count、bundle、unit-price 等阻擋理由的舊配對；名稱含任選、多款、香味、色號、即期、融燭燈、香氛蠟燭等高風險訊號先排除。
--- a/routes/ai_routes.py
+++ b/routes/ai_routes.py
@@ -2008,6 +2008,13 @@ def api_pchome_match_recover_stale():
    """背景搜尋救援過期 identity_v2：舊 ID 缺失或低分時才允許 fresh search recovery。"""
    import threading

+    if os.getenv("PCHOME_STALE_RECOVERY_ENABLED", "false").lower() not in {"1", "true", "yes", "on"}:
+        return jsonify({
+            'success': False,
+            'message': 'PChome 過期 identity 搜尋救援目前僅開放只讀預覽；正式 smoke 顯示成功率不足，需開啟 PCHOME_STALE_RECOVERY_ENABLED 才能執行',
+            'data': _get_pchome_backfill_status_payload(),
+        }), 409
+
    payload = request.get_json(silent=True) or {}
    try:
        limit = max(5, min(int(payload.get('limit', 40)), 80))
--- a/services/competitor_price_feeder.py
+++ b/services/competitor_price_feeder.py
@@ -1139,19 +1139,7 @@ class CompetitorPriceFeeder:

        from sqlalchemy import text
        sql = text(f"""
-            WITH latest_momo AS (
-                SELECT
-                    p.id AS product_id,
-                    p.i_code AS sku,
-                    p.name,
-                    p.category,
-                    pr.price AS momo_price,
-                    ROW_NUMBER() OVER (PARTITION BY p.id ORDER BY pr.timestamp DESC) AS rn
-                FROM products p
-                JOIN price_records pr ON pr.product_id = p.id
-                WHERE p.status = 'ACTIVE'
-            ),
-            latest_attempt AS (
+            WITH latest_attempt AS (
                SELECT DISTINCT ON (cma.sku)
                    cma.sku,
                    cma.best_competitor_product_id,
@@ -1164,33 +1152,44 @@ class CompetitorPriceFeeder:
                FROM competitor_match_attempts cma
                WHERE cma.source = 'pchome'
                ORDER BY cma.sku, cma.attempted_at DESC, cma.id DESC
+            ),
+            candidate_attempt AS (
+                SELECT la.*
+                FROM latest_attempt la
+                WHERE la.best_competitor_product_id IS NOT NULL
+                  AND la.best_competitor_product_id <> ''
+                  AND COALESCE(la.best_match_score, 0) >= :min_score
+                  AND COALESCE(la.hard_veto, false) = false
+                  AND COALESCE(la.match_diagnostic_json->>'comparison_mode', 'exact_identity') = 'exact_identity'
            )
            SELECT
-                lm.product_id,
-                lm.sku,
-                lm.name,
-                lm.category,
-                lm.momo_price,
+                p.id AS product_id,
+                p.i_code AS sku,
+                p.name,
+                p.category,
+                latest_price.price AS momo_price,
                la.best_competitor_product_id AS competitor_product_id,
                la.best_competitor_product_name AS competitor_product_name,
                la.best_match_score,
                la.attempt_status
-            FROM latest_momo lm
-            JOIN latest_attempt la
-              ON la.sku = lm.sku
+            FROM candidate_attempt la
+            JOIN products p
+              ON p.i_code = la.sku
+             AND p.status = 'ACTIVE'
+            JOIN LATERAL (
+                SELECT pr.price
+                FROM price_records pr
+                WHERE pr.product_id = p.id
+                ORDER BY pr.timestamp DESC, pr.id DESC
+                LIMIT 1
+            ) latest_price ON TRUE
            LEFT JOIN competitor_prices cp
-              ON cp.sku = lm.sku
+              ON cp.sku = p.i_code
             AND cp.source = 'pchome'
             AND (cp.expires_at IS NULL OR cp.expires_at > CURRENT_TIMESTAMP)
             AND COALESCE(cp.match_score, 0) >= :match_score_floor
             AND COALESCE(cp.tags, '[]'::jsonb) ? 'identity_v2'
-            WHERE lm.rn = 1
-              AND cp.sku IS NULL
-              AND la.best_competitor_product_id IS NOT NULL
-              AND la.best_competitor_product_id <> ''
-              AND COALESCE(la.best_match_score, 0) >= :min_score
-              AND COALESCE(la.hard_veto, false) = false
-              AND COALESCE(la.match_diagnostic_json->>'comparison_mode', 'exact_identity') = 'exact_identity'
+            WHERE cp.sku IS NULL
              AND (
                  (
                      la.attempt_status IN (
@@ -1214,24 +1213,24 @@ class CompetitorPriceFeeder:
                      )
                      AND (
                          (
-                              lower(COALESCE(lm.name, '')) LIKE '%beauty foot%'
+                              lower(COALESCE(p.name, '')) LIKE '%beauty foot%'
                              AND lower(COALESCE(la.best_competitor_product_name, '')) LIKE '%beauty foot%'
-                              AND COALESCE(lm.name, '') LIKE '%足膜%'
+                              AND COALESCE(p.name, '') LIKE '%足膜%'
                              AND COALESCE(la.best_competitor_product_name, '') LIKE '%足膜%'
                          )
                          OR (
-                              COALESCE(lm.name, '') LIKE '%足足稱奇%'
+                              COALESCE(p.name, '') LIKE '%足足稱奇%'
                              AND COALESCE(la.best_competitor_product_name, '') LIKE '%足足稱奇%'
-                              AND COALESCE(lm.name, '') LIKE '%足膜%'
+                              AND COALESCE(p.name, '') LIKE '%足膜%'
                              AND COALESCE(la.best_competitor_product_name, '') LIKE '%足膜%'
                          )
                          OR (
-                              lower(COALESCE(lm.name, '')) LIKE '%ts6%'
+                              lower(COALESCE(p.name, '')) LIKE '%ts6%'
                              AND lower(COALESCE(la.best_competitor_product_name, '')) LIKE '%ts6%'
                              AND (
-                                  COALESCE(lm.name, '') LIKE '%蜜愛潤滑液%'
-                                  OR COALESCE(lm.name, '') LIKE '%蜜桃煥白凝膠%'
-                                  OR COALESCE(lm.name, '') LIKE '%極淨白私密潔膚露%'
+                                  COALESCE(p.name, '') LIKE '%蜜愛潤滑液%'
+                                  OR COALESCE(p.name, '') LIKE '%蜜桃煥白凝膠%'
+                                  OR COALESCE(p.name, '') LIKE '%極淨白私密潔膚露%'
                              )
                              AND (
                                  COALESCE(la.best_competitor_product_name, '') LIKE '%蜜愛潤滑液%'
@@ -1241,20 +1240,20 @@ class CompetitorPriceFeeder:
                          )
                          OR (
                              (
-                                  lower(COALESCE(lm.name, '')) LIKE '%vaseline%'
-                                  OR COALESCE(lm.name, '') LIKE '%凡士林%'
+                                  lower(COALESCE(p.name, '')) LIKE '%vaseline%'
+                                  OR COALESCE(p.name, '') LIKE '%凡士林%'
                              )
                              AND (
                                  lower(COALESCE(la.best_competitor_product_name, '')) LIKE '%vaseline%'
                                  OR COALESCE(la.best_competitor_product_name, '') LIKE '%凡士林%'
                              )
-                              AND COALESCE(lm.name, '') LIKE '%嬰兒高純修護凝膠%'
+                              AND COALESCE(p.name, '') LIKE '%嬰兒高純修護凝膠%'
                              AND COALESCE(la.best_competitor_product_name, '') LIKE '%嬰兒高純修護凝膠%'
                          )
                      )
                  )
              )
-            ORDER BY la.best_match_score DESC NULLS LAST, lm.momo_price DESC NULLS LAST, lm.sku
+            ORDER BY la.best_match_score DESC NULLS LAST, latest_price.price DESC NULLS LAST, p.i_code
            LIMIT :limit
        """)
        with self.engine.connect() as conn:
--- a/templates/dashboard_v2.html
+++ b/templates/dashboard_v2.html
@@ -65,7 +65,6 @@
             data-pchome-backfill-card
             data-backfill-endpoint="/api/ai/pchome-match/backfill"
             data-refresh-stale-endpoint="/api/ai/pchome-match/refresh-stale"
-             data-recover-stale-endpoint="/api/ai/pchome-match/recover-stale"
             data-status-endpoint="/api/ai/pchome-match/backfill/status"
             data-pchome-backfill-action="backfillPchomeMatches">
            <div class="dashboard-backfill-main">
@@ -89,12 +88,6 @@
                        data-limit="120">
                    <i class="fas fa-rotate"></i> 刷新過期 120 筆
                </button>
-                <button class="dashboard-action-button"
-                        type="button"
-                        data-pchome-recover-stale-trigger
-                        data-limit="40">
-                    <i class="fas fa-life-ring"></i> 救援過期 40 筆
-                </button>
                <button class="dashboard-action-button is-primary"
                        type="button"
                        data-pchome-backfill-trigger
--- a/tests/test_competitor_match_attempts_persistence.py
+++ b/tests/test_competitor_match_attempts_persistence.py
@@ -124,7 +124,7 @@ def test_competitor_feeder_persists_all_match_attempt_outcomes():
    assert "ROW_NUMBER() OVER (PARTITION BY p.id" not in recovery_source
    assert "candidate_limit" in recovery_source
    retryable_source = source.split("def _fetch_retryable_candidate_skus", 1)[1].split(
-        "def _fetch_expired_identity_skus", 1
+        "def preview_retryable_candidate_revalidation", 1
    )[0]
    for status in (
        "'low_score'",
@@ -152,8 +152,15 @@ def test_competitor_feeder_persists_all_match_attempt_outcomes():
    assert "COALESCE(la.hard_veto, false) = false" in retryable_source
    assert "match_diagnostic_json->>'comparison_mode'" in retryable_source
    assert "?| array[" in retryable_source
+    assert "candidate_attempt AS" in retryable_source
+    assert "FROM candidate_attempt la" in retryable_source
+    assert "JOIN LATERAL" in retryable_source
+    assert "ORDER BY pr.timestamp DESC, pr.id DESC" in retryable_source
+    assert "latest_price.price AS momo_price" in retryable_source
+    assert "ROW_NUMBER() OVER (PARTITION BY p.id" not in retryable_source
+    assert "lm.rn = 1" not in retryable_source
    latest_attempt_source = retryable_source.split("latest_attempt AS", 1)[1].split(
-        "SELECT\n                lm.product_id", 1
+        "candidate_attempt AS", 1
    )[0]
    assert "cma.best_competitor_product_id IS NOT NULL" not in latest_attempt_source
    assert "la.best_competitor_product_id IS NOT NULL" in retryable_source
--- a/tests/test_frontend_v2_assets.py
+++ b/tests/test_frontend_v2_assets.py
@@ -518,6 +518,7 @@ def test_ai_product_pick_agent_uses_real_competitor_data_and_dashboard_action():
    assert "@ai_bp.route('/api/ai/pchome-match/backfill', methods=['POST'])" in route_source
    assert "@ai_bp.route('/api/ai/pchome-match/refresh-stale', methods=['POST'])" in route_source
    assert "@ai_bp.route('/api/ai/pchome-match/recover-stale', methods=['POST'])" in route_source
+    assert 'PCHOME_STALE_RECOVERY_ENABLED' in route_source
    assert "@ai_bp.route('/api/ai/pchome-match/backfill/status', methods=['GET'])" in route_source
    assert "_build_pchome_backfill_coverage_payload" in route_source
    assert "_build_pchome_revalidation_preview_payload" in route_source
@@ -569,19 +570,19 @@ def test_ai_product_pick_agent_uses_real_competitor_data_and_dashboard_action():
    assert "/api/ai/product-picks/generate" in template
    assert "/api/ai/pchome-match/backfill" in template
    assert "/api/ai/pchome-match/refresh-stale" in dashboard_template
-    assert "/api/ai/pchome-match/recover-stale" in dashboard_template
+    assert "/api/ai/pchome-match/recover-stale" not in dashboard_template
    assert "/api/ai/pchome-match/backfill/status" in dashboard_template
    assert "PCHOME MATCH BACKFILL" in dashboard_template
    assert "data-pchome-backfill-trigger" in dashboard_template
    assert "data-pchome-refresh-stale-trigger" in dashboard_template
-    assert "data-pchome-recover-stale-trigger" in dashboard_template
+    assert "data-pchome-recover-stale-trigger" not in dashboard_template
    assert "刷新過期 120 筆" in dashboard_template
-    assert "救援過期 40 筆" in dashboard_template
+    assert "救援過期 40 筆" not in dashboard_template
    dashboard_js = (ROOT / "web/static/js/page-dashboard-v2.js").read_text(encoding="utf-8")
    assert "loadPchomeBackfillStatus" in dashboard_js
    assert "window.backfillPchomeMatches" in dashboard_js
    assert "window.refreshStalePchomeMatches" in dashboard_js
-    assert "window.recoverStalePchomeMatches" in dashboard_js
+    assert "window.recoverStalePchomeMatches" not in dashboard_js
    assert "formatBackfillCoverageSummary" in dashboard_js
    assert "formatBackfillLimitedCount" in dashboard_js
    assert "status.coverage" in dashboard_js
--- a/web/static/js/page-dashboard-v2.js
+++ b/web/static/js/page-dashboard-v2.js
@@ -288,13 +288,11 @@ let priceChartInstance = null;
            card,
            trigger: document.querySelector('[data-pchome-backfill-trigger]'),
            refreshStaleTrigger: document.querySelector('[data-pchome-refresh-stale-trigger]'),
-            recoverStaleTrigger: document.querySelector('[data-pchome-recover-stale-trigger]'),
            status: document.querySelector('[data-pchome-backfill-status]'),
            result: document.querySelector('[data-pchome-backfill-result]'),
            progress: document.querySelector('[data-pchome-backfill-progress]'),
            backfillEndpoint: card ? card.dataset.backfillEndpoint : '/api/ai/pchome-match/backfill',
            refreshStaleEndpoint: card ? card.dataset.refreshStaleEndpoint : '/api/ai/pchome-match/refresh-stale',
-            recoverStaleEndpoint: card ? card.dataset.recoverStaleEndpoint : '/api/ai/pchome-match/recover-stale',
            statusEndpoint: card ? card.dataset.statusEndpoint : '/api/ai/pchome-match/backfill/status'
        };
    }
@@ -401,13 +399,6 @@ let priceChartInstance = null;
                ? '<i class="fas fa-spinner fa-spin"></i> 執行中'
                : '<i class="fas fa-rotate"></i> 刷新過期 120 筆';
        }
-        if (elements.recoverStaleTrigger) {
-            elements.recoverStaleTrigger.disabled = running;
-            elements.recoverStaleTrigger.classList.toggle('is-loading', running);
-            elements.recoverStaleTrigger.innerHTML = running
-                ? '<i class="fas fa-spinner fa-spin"></i> 執行中'
-                : '<i class="fas fa-life-ring"></i> 救援過期 40 筆';
-        }

        if (running) {
            schedulePchomeBackfillPoll();
@@ -505,54 +496,14 @@ let priceChartInstance = null;
            });
    }

-    function recoverStalePchomeMatches() {
-        const elements = getPchomeBackfillElements();
-        if (!elements.card || !elements.recoverStaleTrigger) return;
-        const limit = Number(elements.recoverStaleTrigger.dataset.limit || 40);
-        if (!confirm(`啟動 PChome 過期 identity 搜尋救援 ${limit} 筆？`)) return;
-
-        elements.recoverStaleTrigger.disabled = true;
-        if (elements.status) {
-            elements.status.textContent = '正在送出過期 identity 搜尋救援任務';
-        }
-        fetch(elements.recoverStaleEndpoint, {
-            method: 'POST',
-            headers: {
-                'Content-Type': 'application/json',
-                'X-CSRFToken': getCSRFToken()
-            },
-            body: JSON.stringify({ limit })
-        })
-            .then(response => response.json().then(data => ({ ok: response.ok, status: response.status, data })))
-            .then(({ ok, status, data }) => {
-                renderPchomeBackfillStatus(data);
-                if (!ok && status !== 409) {
-                    throw new Error(data.message || data.error || 'PChome 過期 identity 搜尋救援啟動失敗');
-                }
-                schedulePchomeBackfillPoll();
-            })
-            .catch(error => {
-                if (elements.status) {
-                    elements.status.textContent = error.message || 'PChome 過期 identity 搜尋救援啟動失敗';
-                }
-                if (elements.recoverStaleTrigger) {
-                    elements.recoverStaleTrigger.disabled = false;
-                }
-            });
-    }
-
    window.backfillPchomeMatches = backfillPchomeMatches;
    window.refreshStalePchomeMatches = refreshStalePchomeMatches;
-    window.recoverStalePchomeMatches = recoverStalePchomeMatches;
    document.querySelectorAll('[data-pchome-backfill-trigger]').forEach(button => {
        button.addEventListener('click', backfillPchomeMatches);
    });
    document.querySelectorAll('[data-pchome-refresh-stale-trigger]').forEach(button => {
        button.addEventListener('click', refreshStalePchomeMatches);
    });
-    document.querySelectorAll('[data-pchome-recover-stale-trigger]').forEach(button => {
-        button.addEventListener('click', recoverStalePchomeMatches);
-    });
    loadPchomeBackfillStatus();

    function runPchomeReviewDecision(button) {