fix(catalog): rename the DeepSeek flash model to its canonical id deepseek-flash

config/config.example.toml

@@ -229,24 +229,27 @@ endpoints = [
 #    高峰时段价（rant 2026-08-20T11:58:40）：peak_input_per_m / peak_output_per_m /
 #    peak_cache_hit_input_per_m 三个可选字段，缺省 0 = 不启用高峰计费（沿用空闲价）。
 #    命中高峰时段（北京时间 9-12、14-18，周一至周日）时按高峰价扣点。
-#    DeepSeek 为官方价（2026-08-20 浏览器核实）：空闲价即高峰价的一半，
-#    高峰 9-12/14-18 北京时翻倍（flash 3.0/0.10/9.0，pro 9.0/0.30/27.0 元/M）。
+#    DeepSeek 为官方价（2026-09-14 浏览器核实，https://api-docs.deepseek.com/zh-cn/quick_start/pricing）：
+#    空闲价即高峰价的一半，高峰 9-12/14-18 北京时翻倍（flash 2.0/0.04/8.0，pro 9.0/0.30/27.0 元/M）。
+#    模型名同样以官方文档为准：V4.1-Flash 的模型名是 `deepseek-flash`（V4 Flash 与 V4 Flash-Vision-Exp
+#    已下线，旧名请求被官方路由到 V4.1-Flash 并按 Flash 价计费 ⇒ 目录中不再单列那两个旧名）。
 # ============================================================
 
 [[models]]
 provider = "deepseek"
-model = "deepseek-v4-flash"
+model = "deepseek-flash"
 currency = "CNY"
-input_per_m = 1.5
-cache_hit_input_per_m = 0.05
-output_per_m = 4.5
+input_per_m = 1.0
+cache_hit_input_per_m = 0.02
+output_per_m = 4.0
 # 高峰时段价（北京 9-12 / 14-18 翻倍；官方价）
-peak_input_per_m = 3.0
-peak_cache_hit_input_per_m = 0.10
-peak_output_per_m = 9.0
+peak_input_per_m = 2.0
+peak_cache_hit_input_per_m = 0.04
+peak_output_per_m = 8.0
 context_length = 1048576
 max_output = 384000
-vision = false
+# DeepSeek-V4.1-Flash 原生多模态（官方「图像理解：支持」）⇒ 旧 vision-exp 条目已并入本条
+vision = true
 
 [[models]]
 provider = "deepseek"
@@ -379,18 +382,5 @@ context_length = 1048576
 max_output = 0
 vision = false
 
-# DeepSeek 视觉模型：价格与 deepseek-v4-flash 完全相同（读图能力增强）
-[[models]]
-provider = "deepseek"
-model = "deepseek-v4-flash-vision-exp"
-currency = "CNY"
-input_per_m = 1.5
-cache_hit_input_per_m = 0.05
-output_per_m = 4.5
-# 高峰时段价（北京 9-12 / 14-18 翻倍；与 flash 一致）
-peak_input_per_m = 3.0
-peak_cache_hit_input_per_m = 0.10
-peak_output_per_m = 9.0
-context_length = 1048576
-max_output = 384000
-vision = true
+# 读图模型：DeepSeek 的视觉能力已并入 `deepseek-flash`（V4.1-Flash 原生多模态），
+# 旧的 `deepseek-v4-flash-vision-exp` 已下线并被官方路由到 V4.1-Flash ⇒ 此处不再单列。

docs/plan-api-matrix.md

@@ -32,4 +32,4 @@ Base URL 统一由 AITokenPool 提供（`[server].public_url` 配置），设置
 
 ## 4. 模型目录
 
-当前内置 14 个模型（DeepSeek / GLM / GPT / Claude / Gemini / 豆包 / MiniMax / 通义等），含 context window、vision 支持、缓存价与高峰价字段（数量与明细以 `config/config.example.toml` 的 `[[models]]` 为唯一真源，由 `src/catalog_gate.rs` 的 `MODEL_COUNT` 守卫）。模型列表可经管理端「模型管理」页 CRUD（需 admin 角色）。
+内置 DeepSeek / GLM / GPT / Claude / Gemini / 豆包 / MiniMax / 通义等模型，含 context window、vision 支持、缓存价与高峰价字段。**数量与明细以 `config/config.example.toml` 的 `[[models]]` 为唯一真源**（由 `src/catalog_gate.rs` 的 `MODEL_COUNT` 守卫，此处不抄写数字以免随之陈旧）。模型列表可经管理端「模型管理」页 CRUD（需 admin 角色）。

src/catalog_gate.rs

@@ -32,14 +32,14 @@ const CONFIG_TOML: &str = include_str!("../config/config.example.toml");
 ///
 /// 它们的作用是把「提取器静默失真」与「数据真的变了」区分开：若扫描器写错而返回空集，
 /// 集合断言会**在空集上"通过"**（C2005 坑 68），这些计数会先把运行中止。
-const MODEL_COUNT: usize = 14;
+const MODEL_COUNT: usize = 13;
 const MARKET_COUNT: usize = 7;
 const PLAN_COUNT: usize = 12;
 
 /// 已知真值（从配置里读出的官方价，CNY 计价）——用于确认解析器真的读到了正确字段。
-const KNOWN_MODEL: &str = "deepseek-v4-flash";
-const KNOWN_INPUT: f64 = 1.5;
-const KNOWN_OUTPUT: f64 = 4.5;
+const KNOWN_MODEL: &str = "deepseek-flash";
+const KNOWN_INPUT: f64 = 1.0;
+const KNOWN_OUTPUT: f64 = 4.0;
 
 /// 价格比较的容差。两侧都由十进制字面量解析而来，实际是精确相等；
 /// 留一个极小容差只为避免浮点表示差异造成的假红。
@@ -447,7 +447,7 @@ mod tests {
     ///
     /// 双向是关键，且两个方向对应两类真实事故：
     /// - 「兜底多出来的」= 幽灵模型（`ce6d0db` 塞进 `gemini-3.5-flash-lite`）；
-    /// - 「兜底缺失的」= 新增模型忘了同步（`deepseek-v4-flash-vision-exp`）。
+    /// - 「兜底缺失的」= 新增模型忘了同步（`ce6d0db` 之后新增 `deepseek-flash` 时）。
     ///
     /// 只查一个方向会漏掉其中一类（C2010 更正：此前我误以为该断言应为「子集」）。
     #[test]
@@ -562,7 +562,7 @@ mod tests {
             "阴性对照失败：MARKET 中的改名残留未被报出"
         );
 
-        // ④ 错价（真实事故形态：旧的 flash 价格 1.008/2.016）
+        // ④ 错价（真实事故形态：旧的 DeepSeek flash 价格 1.5/4.5，官方已于 2026-09-10 下调为 1.0/4.0）
         let mut wrong = js_models.get(KNOWN_MODEL).cloned().unwrap();
         wrong.input = 1.008;
         wrong.output = 2.016;

src/config.rs

@@ -338,12 +338,26 @@ mod tests {
         let flash = cfg
             .models
             .iter()
-            .find(|m| m.model == "deepseek-v4-flash")
-            .unwrap();
-        assert_eq!(flash.input_per_m, 1.5);
-        assert_eq!(flash.cache_hit_input_per_m, 0.05);
-        assert_eq!(flash.peak_input_per_m, 3.0);
-        assert_eq!(flash.peak_output_per_m, 9.0);
+            .find(|m| m.model == "deepseek-flash")
+            .expect("deepseek-flash（V4.1-Flash，官方 2026-09-14 的现名）应在模型目录中");
+        assert_eq!(flash.input_per_m, 1.0);
+        assert_eq!(flash.cache_hit_input_per_m, 0.02);
+        assert_eq!(flash.output_per_m, 4.0);
+        assert_eq!(flash.peak_input_per_m, 2.0);
+        assert_eq!(flash.peak_cache_hit_input_per_m, 0.04);
+        assert_eq!(flash.peak_output_per_m, 8.0);
+        assert!(
+            flash.vision,
+            "V4.1-Flash 原生多模态（官方「图像理解：支持」）"
+        );
+        // 已下线的旧名不得再出现在目录中：市场列的是「现在能买的模型」
+        // （官方：deepseek-v4-flash / deepseek-v4-flash-vision-exp「对应模型已下线」）
+        assert!(
+            cfg.models.iter().all(
+                |m| m.model != "deepseek-v4-flash" && m.model != "deepseek-v4-flash-vision-exp"
+            ),
+            "已下线的 DeepSeek 旧模型名不应留在模型目录中"
+        );
         // 未配置高峰价的模型 → 缺省 0（不启用高峰计费）
         let zhipu = cfg.models.iter().find(|m| m.provider == "zhipu").unwrap();
         assert_eq!(zhipu.peak_input_per_m, 0.0, "无高峰价字段 → 缺省 0");

src/db.rs

@@ -463,7 +463,7 @@ pub(crate) fn seed_test_users(conn: &Connection) -> Result<()> {
     if key_count == 0 {
         conn.execute(
             "INSERT INTO keys (provider, plan, model, status, owner_id, encrypted_key, quota, used) \
-             VALUES ('deepseek', 'deepseek-paygo', 'deepseek-v4-flash', 'on', ?1, 'sk-placeholder-encrypted', 1000, 0)",
+             VALUES ('deepseek', 'deepseek-paygo', 'deepseek-flash', 'on', ?1, 'sk-placeholder-encrypted', 1000, 0)",
             [demo_id],
         )?;
     }
@@ -924,19 +924,19 @@ mod tests {
             "peak cache_hit={peak_cache}"
         );
         assert_eq!(currency, "CNY");
-        // flash 也在（config 直接定义；高峰 3.0 / 9.0 / 0.10）
+        // deepseek-flash（V4.1-Flash）也在（config 直接定义；高峰 2.0 / 8.0 / 0.04）
         let (fi, fh, fp_in, fp_out): (f64, f64, f64, f64) = conn
             .query_row(
                 "SELECT input_per_m, cache_hit_input_per_m, peak_input_per_m, peak_output_per_m \
-                 FROM models WHERE provider = 'deepseek' AND model = 'deepseek-v4-flash'",
+                 FROM models WHERE provider = 'deepseek' AND model = 'deepseek-flash'",
                 [],
                 |r| Ok((r.get(0)?, r.get(1)?, r.get(2)?, r.get(3)?)),
             )
             .unwrap();
-        assert!((fi - 1.5).abs() < 1e-9, "flash input={fi}");
-        assert!((fh - 0.05).abs() < 1e-9, "flash cache_hit={fh}");
-        assert!((fp_in - 3.0).abs() < 1e-9, "flash peak input={fp_in}");
-        assert!((fp_out - 9.0).abs() < 1e-9, "flash peak output={fp_out}");
+        assert!((fi - 1.0).abs() < 1e-9, "flash input={fi}");
+        assert!((fh - 0.02).abs() < 1e-9, "flash cache_hit={fh}");
+        assert!((fp_in - 2.0).abs() < 1e-9, "flash peak input={fp_in}");
+        assert!((fp_out - 8.0).abs() < 1e-9, "flash peak output={fp_out}");
         // 无高峰价的模型 → peak 缺省 0（不启用高峰计费）
         let (z_in, z_cache, z_out): (f64, f64, f64) = conn
             .query_row(

src/routes/sharing.rs

@@ -358,7 +358,7 @@ mod tests {
             st.clone(),
             "POST",
             "/api/sharings",
-            Some(r#"{"provider":"deepseek","plan":"deepseek-paygo","model":"deepseek-v4-flash","key":"sk-realsecret1234","quota":1000,"available":{"days":[1,2,3,4,5],"start":"09:00","end":"18:00"},"note":"工作日共享"}"#),
+            Some(r#"{"provider":"deepseek","plan":"deepseek-paygo","model":"deepseek-flash","key":"sk-realsecret1234","quota":1000,"available":{"days":[1,2,3,4,5],"start":"09:00","end":"18:00"},"note":"工作日共享"}"#),
             &key,
         )
         .await;
@@ -414,7 +414,7 @@ mod tests {
             st.clone(),
             "POST",
             "/api/sharings",
-            Some(r#"{"provider":"deepseek","plan":"deepseek-paygo","model":"deepseek-v4-flash","key":"sk-patchme9999"}"#),
+            Some(r#"{"provider":"deepseek","plan":"deepseek-paygo","model":"deepseek-flash","key":"sk-patchme9999"}"#),
             &key,
         )
         .await;
@@ -492,7 +492,7 @@ mod tests {
             st.clone(),
             "POST",
             "/api/sharings",
-            Some(r#"{"provider":"deepseek","plan":"deepseek-paygo","model":"deepseek-v4-flash","key":"sk-created1234","used":0,"note":"utc"}"#),
+            Some(r#"{"provider":"deepseek","plan":"deepseek-paygo","model":"deepseek-flash","key":"sk-created1234","used":0,"note":"utc"}"#),
             &key,
         )
         .await;
@@ -580,7 +580,7 @@ mod tests {
             st,
             "POST",
             "/api/sharings",
-            Some(r#"{"provider":"deepseek","plan":"deepseek-paygo","model":"deepseek-v4-flash","key":"aa中中中中中","quota":1000,"available":{"days":[1],"start":"09:00","end":"18:00"}}"#),
+            Some(r#"{"provider":"deepseek","plan":"deepseek-paygo","model":"deepseek-flash","key":"aa中中中中中","quota":1000,"available":{"days":[1],"start":"09:00","end":"18:00"}}"#),
             &token,
         )
         .await;
@@ -614,7 +614,7 @@ mod tests {
             st.clone(),
             "POST",
             "/api/sharings",
-            Some(r#"{"provider":"WRONG","plan":"deepseek-paygo","model":"deepseek-v4-flash","key":"sk-wrong1234","quota":100}"#),
+            Some(r#"{"provider":"WRONG","plan":"deepseek-paygo","model":"deepseek-flash","key":"sk-wrong1234","quota":100}"#),
             &key,
         )
         .await;
@@ -639,7 +639,7 @@ mod tests {
             st.clone(),
             "POST",
             "/api/sharings",
-            Some(r#"{"provider":"deepseek","plan":"deepseek-paygo","model":"deepseek-v4-flash","key":"sk-control1234","quota":100}"#),
+            Some(r#"{"provider":"deepseek","plan":"deepseek-paygo","model":"deepseek-flash","key":"sk-control1234","quota":100}"#),
             &key,
         )
         .await;
@@ -676,7 +676,7 @@ mod tests {
             st.clone(),
             "POST",
             "/api/sharings",
-            Some(r#"{"provider":"deepseek","plan":"no-such-plan","model":"deepseek-v4-flash","key":"sk-phantom1111","quota":100}"#),
+            Some(r#"{"provider":"deepseek","plan":"no-such-plan","model":"deepseek-flash","key":"sk-phantom1111","quota":100}"#),
             &key,
         )
         .await;
@@ -687,7 +687,7 @@ mod tests {
             st.clone(),
             "POST",
             "/api/sharings",
-            Some(r#"{"provider":"deepseek","model":"deepseek-v4-flash","key":"sk-noplan2222","quota":100}"#),
+            Some(r#"{"provider":"deepseek","model":"deepseek-flash","key":"sk-noplan2222","quota":100}"#),
             &key,
         )
         .await;
@@ -701,7 +701,7 @@ mod tests {
             st.clone(),
             "POST",
             "/api/sharings",
-            Some(r#"{"provider":"deepseek","plan":"deepseek-paygo","model":"deepseek-v4-flash","key":"sk-control3333","quota":100}"#),
+            Some(r#"{"provider":"deepseek","plan":"deepseek-paygo","model":"deepseek-flash","key":"sk-control3333","quota":100}"#),
             &key,
         )
         .await;
@@ -714,8 +714,8 @@ mod tests {
             crate::dao::list_models_with_availability(&conn)
                 .unwrap()
                 .iter()
-                .find(|m| m["model"] == "deepseek-v4-flash")
-                .expect("市场列表应含 deepseek-v4-flash")["available_keys"]
+                .find(|m| m["model"] == "deepseek-flash")
+                .expect("市场列表应含 deepseek-flash")["available_keys"]
                 .as_i64()
                 .unwrap()
         };
@@ -724,7 +724,7 @@ mod tests {
             let conn = st.db.lock().unwrap();
             conn.execute(
                 "INSERT INTO keys (provider, plan, model, status, owner_id, encrypted_key, quota, used) \
-                 VALUES ('deepseek', 'no-such-plan', 'deepseek-v4-flash', 'on', 1, 'sk-legacy-enc', 100, 0)",
+                 VALUES ('deepseek', 'no-such-plan', 'deepseek-flash', 'on', 1, 'sk-legacy-enc', 100, 0)",
                 [],
             )
             .unwrap();

ui/index.html

@@ -843,7 +843,7 @@ <h3 id="chat-title">使用模型</h3>
 </div>
 
 <script src="js/api.js?v=20260912-5"></script>
-<script src="js/data.js?v=20260912-5"></script>
+<script src="js/data.js?v=20260914-1"></script>
 <script src="js/i18n.js?v=20260914-4"></script>
 <script src="js/app.js?v=20260914-9"></script>
 </body>

ui/js/data.js

@@ -17,8 +17,7 @@
   // 模型价格（对齐 config.toml [[models]] 官方价，折算为点数 / 1M tokens）——上架表单定价兜底
   const MODELS = [
     { provider: "deepseek", model: "deepseek-v4-pro",  in: CNY(4.5),   out: CNY(13.5),  ctx: 1048576, max: 384000, tag: "推理" },
-    { provider: "deepseek", model: "deepseek-v4-flash", in: CNY(1.5),   out: CNY(4.5),   ctx: 1048576, max: 384000, tag: "通用" },
-    { provider: "deepseek", model: "deepseek-v4-flash-vision-exp", in: CNY(1.5), out: CNY(4.5), ctx: 1048576, max: 384000, tag: "读图" },
+    { provider: "deepseek", model: "deepseek-flash",   in: CNY(1.0),   out: CNY(4.0),   ctx: 1048576, max: 384000, tag: "多模态" },
     { provider: "zhipu",    model: "glm-5.3",          in: CNY(8.0),   out: CNY(28.0),  ctx: 1048576, max: 131072, tag: "旗舰" },
     { provider: "openai",   model: "gpt-5.6-sol",      in: USD(5.0),   out: USD(30.0),  ctx: 1050000, max: null,   tag: "旗舰" },
     { provider: "anthropic",model: "claude-opus-5",     in: USD(5.0),   out: USD(25.0),  ctx: 1000000, max: null,   tag: "旗舰" },
@@ -79,7 +78,7 @@
     // 市场在售 key（游客浏览用；rant 2026-08-19T15:54:06：multi/success 为虚构数据已移除；
     // 登录态用 GET /api/models 真实数据，multi=available_keys>=2、ctx=context_window）
     MARKET: [
-      { id: 1, provider: "deepseek", model: "deepseek-v4-flash", in: CNY(1.5),  out: CNY(4.5),   ctx: 1048576, avail: true, peak: true, peakIn: CNY(3.0), peakOut: CNY(9.0), peakMult: 2 },
+      { id: 1, provider: "deepseek", model: "deepseek-flash",    in: CNY(1.0),  out: CNY(4.0),   ctx: 1048576, avail: true, peak: true, peakIn: CNY(2.0), peakOut: CNY(8.0), peakMult: 2 },
       { id: 2, provider: "zhipu",    model: "glm-5.3",           in: CNY(8.0),  out: CNY(28.0),  ctx: 1048576, avail: true },
       { id: 3, provider: "openai",   model: "gpt-5.6-sol",       in: USD(5.0),  out: USD(30.0),  ctx: 1050000, avail: true },
       { id: 4, provider: "anthropic",model: "claude-opus-5",     in: USD(5.0),  out: USD(25.0),  ctx: 1000000, avail: false },