chore(secrets-mcp): 0.5.1 — 移除 entry type 归一化，MCP 参数兼容字符串形式

- 去掉 taxonomy 对 entry type 的自动映射与 metadata.subtype 回填；仅 trim 后入库
- MCP tools：Vec/Map/bool 等可选字段支持 JSON 内嵌字符串解析，并改进解析失败提示
- 新增 deser 单元测试；README/AGENTS 与 models 注释同步

Made-with: Cursor

.gitignore

@@ -2,7 +2,7 @@
 .env
 .DS_Store
 .cursor/
-# Google OAuth 下载的 JSON 凭据文件
-client_secret_*.apps.googleusercontent.com.json
 *.pem
 tmp/
+client_secret_*.apps.googleusercontent.com.json
+node_modules/

AGENTS.md

@@ -55,13 +55,24 @@ entries (
 ```sql
 secrets (
   id          UUID PRIMARY KEY DEFAULT uuidv7(),
-  entry_id    UUID         NOT NULL REFERENCES entries(id) ON DELETE CASCADE,
+  user_id     UUID,
-  field_name  VARCHAR(256) NOT NULL,
+  name        VARCHAR(256) NOT NULL,
+  type        VARCHAR(64)  NOT NULL DEFAULT 'text',
   encrypted   BYTEA        NOT NULL DEFAULT '\x',
   version     BIGINT       NOT NULL DEFAULT 1,
   created_at  TIMESTAMPTZ  NOT NULL DEFAULT NOW(),
-  updated_at  TIMESTAMPTZ  NOT NULL DEFAULT NOW(),
+  updated_at  TIMESTAMPTZ  NOT NULL DEFAULT NOW()
-  UNIQUE(entry_id, field_name)
+)
+-- 唯一：UNIQUE(user_id, name) WHERE user_id IS NOT NULL
+```
+
+```sql
+entry_secrets (
+  entry_id    UUID NOT NULL REFERENCES entries(id) ON DELETE CASCADE,
+  secret_id   UUID NOT NULL REFERENCES secrets(id) ON DELETE CASCADE,
+  sort_order  INT  NOT NULL DEFAULT 0,
+  created_at  TIMESTAMPTZ NOT NULL DEFAULT NOW(),
+  PRIMARY KEY(entry_id, secret_id)
 )
 ```
 
@@ -108,17 +119,20 @@ oauth_accounts (
 | 字段 | 含义 | 示例 |
 |------|------|------|
 | `folder` | 隔离空间（参与唯一键） | `refining` |
-| `type` | 软分类（不参与唯一键） | `server`, `service`, `key`, `person` |
+| `type` | 软分类（不参与唯一键，用户自定义） | `server`, `service`, `account`, `person`, `document` |
 | `name` | 标识名 | `gitea`, `aliyun` |
 | `notes` | 非敏感说明 | 自由文本 |
 | `tags` | 标签 | `["aliyun","prod"]` |
-| `metadata` | 明文描述 | `ip`、`url`、`key_ref` |
+| `metadata` | 明文描述 | `ip`、`url`、`subtype` |
-| `secrets.field_name` | 加密字段名（明文） | `token`, `ssh_key` |
+| `secrets.name` | 密钥名称（调用方提供） | `token`, `ssh_key`, `password` |
+| `secrets.type` | 密钥类型（调用方提供，默认 `text`） | `text`, `password`, `key` |
 | `secrets.encrypted` | 密文 | AES-GCM |
 
-### PEM 共享（`key_ref`）
+### 共享密钥（N:N 关联）
 
-建议将共享 PEM 存为 **`type=key`** 的 entry；其它记录在 `metadata.key_ref` 指向目标 entry 的 `name`（支持 `folder/name` 格式消歧）。删除被引用 key 时，服务会自动迁移为单副本 + 重定向（复制到首个引用方，其余引用方改指向新 owner）；解析逻辑见 `secrets_core::service::env_map`。
+多个 entry 可共享同一 secret 字段，通过 `entry_secrets` 中间表关联。
+添加条目时通过 `link_secret_names` 参数指定要关联的已有 secret name（按 `(user_id, name)` 精确匹配）。
+删除 entry 时仅解除关联，secret 本身若仍被引用则保留；不再被任何 entry 引用时自动清理。
 
 ## 代码规范
 
@@ -166,6 +180,9 @@ git tag -l 'secrets-mcp-*'
 | 变量 | 说明 |
 |------|------|
 | `SECRETS_DATABASE_URL` | **必填**。PostgreSQL URL。 |
+| `SECRETS_DATABASE_SSL_MODE` | 可选但强烈建议生产必填。推荐 `verify-full`（至少 `verify-ca`）。 |
+| `SECRETS_DATABASE_SSL_ROOT_CERT` | 可选。私有 CA 或自签链路时指定 CA 根证书路径。 |
+| `SECRETS_ENV` | 可选。设为 `prod` / `production` 时会拒绝弱 PostgreSQL TLS 模式。 |
 | `BASE_URL` | 对外基址；OAuth 回调 `${BASE_URL}/auth/google/callback`。 |
 | `SECRETS_MCP_BIND` | 监听地址，默认 `127.0.0.1:9315`（容器/远程直接暴露时需改为 `0.0.0.0:9315`）。 |
 | `GOOGLE_CLIENT_ID` / `GOOGLE_CLIENT_SECRET` | 可选；仅运行时配置。 |

Cargo.lock

@@ -1960,6 +1960,7 @@ dependencies = [
  "sha2",
  "sqlx",
  "tempfile",
+ "thiserror",
  "tokio",
  "toml",
  "tracing",
@@ -1968,7 +1969,7 @@ dependencies = [
 
 [[package]]
 name = "secrets-mcp"
-version = "0.3.8"
+version = "0.5.1"
 dependencies = [
  "anyhow",
  "askama",

Cargo.toml

@@ -28,6 +28,7 @@ rand = "^0.10.0"
 
 # Utils
 anyhow = "^1.0.102"
+thiserror = "^2"
 chrono = { version = "^0.4.44", features = ["serde"] }
 uuid = { version = "^1.22.0", features = ["serde"] }
 tracing = "^0.1"

README.md

@@ -54,10 +54,31 @@ SECRETS_ENV=production
 
 条目在逻辑上以 **`(folder, name)`** 在用户内唯一（数据库唯一索引：`user_id + folder + name`）。同名可在不同 folder 下各存一条（例如 `refining/aliyun` 与 `ricnsmart/aliyun`）。
 
-- **`secrets_search`**：发现条目（可按 query / folder / type / name 过滤）；不要求加密头。
+### 工具列表
-- **`secrets_get` / `secrets_update` / `secrets_delete`（按 name）/ `secrets_history` / `secrets_rollback`**：仅 `name` 且全局唯一则直接命中；若多条同名，返回消歧错误，需在参数中补 **`folder`**。
+
-- **`secrets_delete`**：`dry_run=true` 时与真实删除相同的消歧规则——唯一则预览一条，多条则报错并要求 `folder`。
+| 工具 | 需要加密密钥 | 说明 |
-- **共享 key 自动迁移删除**：删除仍被 `metadata.key_ref` 引用的 key 条目时，系统会自动迁移：把密文复制到首个引用方，并将其余引用方的 `key_ref` 重定向到新 owner，然后继续删除。
+|------|-------------|------|
+| `secrets_find` | 否 | 发现条目（返回含 secret_fields schema），支持 `name_query` 模糊匹配 |
+| `secrets_search` | 否 | 搜索条目，支持 `query`/`folder`/`type`/`name` 过滤、`sort`/`offset` 分页、`summary` 摘要模式 |
+| `secrets_get` | 是 | 按 UUID `id` 获取单条条目及解密后的 secrets |
+| `secrets_add` | 是 | 添加新条目，支持 `meta_obj`/`secrets_obj` JSON 对象参数、`secret_types` 指定密钥类型、`link_secret_names` 关联已有 secret |
+| `secrets_update` | 是 | 更新条目，支持 `id` 或 `name`+`folder` 定位 |
+| `secrets_delete` | 否 | 删除条目，支持 `id` 或 `name`+`folder` 定位；`dry_run=true` 预览删除 |
+| `secrets_history` | 否 | 查看条目历史，支持 `id` 或 `name`+`folder` 定位 |
+| `secrets_rollback` | 是 | 回滚条目到指定历史版本，支持 `id` 或 `name`+`folder` 定位 |
+| `secrets_export` | 是 | 导出条目（含解密明文），支持 JSON/TOML/YAML 格式 |
+| `secrets_env_map` | 是 | 将 secrets 转换为环境变量映射（`UPPER(entry)_UPPER(field)` 格式），支持 `prefix` |
+| `secrets_overview` | 否 | 返回各 folder 和 type 的 entry 计数概览 |
+
+### 消歧规则
+
+- **按 `name` 定位的工具**（`secrets_update` / `secrets_delete` / `secrets_history` / `secrets_rollback`）：若该用户下仅一条匹配则直接执行；若多条（同 `name`、不同 `folder`）则返回错误并提示补全 `folder`。也可直接传 `id`（UUID）跳过消歧。
+- **`secrets_get`** 仅支持通过 `id`（UUID）获取。
+- **`secrets_delete`** 的 `dry_run=true` 与真实删除使用相同消歧规则——唯一则预览一条，多条则报错并要求 `folder`。
+
+### 共享密钥
+
+N:N 关联下，删除 entry 仅解除关联，被共享的 secret 若仍被其他 entry 引用则保留；无引用时自动清理。
 
 ## 加密架构（混合 E2EE）
 
@@ -151,25 +172,32 @@ flowchart LR
 
 ## 数据模型
 
-主表 **`entries`**（`folder`、`type`、`name`、`notes`、`tags`、`metadata`，多租户时带 `user_id`）+ 子表 **`secrets`**（每行一个加密字段：`field_name`、`encrypted`）。**唯一性**：`UNIQUE(user_id, folder, name)`（`user_id` 为空时为遗留行唯一 `(folder, name)`）。另有 `entries_history`、`secrets_history`、`audit_log`，以及 **`users`**（含 `key_salt`、`key_check`、`key_params`、`api_key`）、**`oauth_accounts`**。首次连库自动迁移建表（`secrets-core` 的 `migrate`）；已有库可对照 [`scripts/migrate-v0.3.0.sql`](scripts/migrate-v0.3.0.sql) 做列重命名与索引重建。**Web 登录会话**（tower-sessions）使用同一 `SECRETS_DATABASE_URL`，进程启动时对会话存储执行迁移（见 `secrets-mcp` 中 `PostgresStore::migrate`），无需额外环境变量。
+主表 **`entries`**（`folder`、`type`、`name`、`notes`、`tags`、`metadata`，多租户时带 `user_id`）+ 子表 **`secrets`**（每行一个加密字段：`name`、`type`、`encrypted`，通过 `entry_secrets` 中间表与 entry 建立 N:N 关联）。**唯一性**：`UNIQUE(user_id, folder, name)`（`user_id` 为空时为遗留行唯一 `(folder, name)`）。另有 `entries_history`、`secrets_history`、`audit_log`，以及 **`users`**（含 `key_salt`、`key_check`、`key_params`、`api_key`）、**`oauth_accounts`**。首次连库自动迁移建表（`secrets-core` 的 `migrate`）；已有库在进程启动时亦由同一 `migrate()` 增量补齐表、索引与 N:N 结构。若需从更早版本对照一次性 SQL，可在 git 历史中检索已移除的 `scripts/migrate-v0.3.0.sql`。**Web 登录会话**（tower-sessions）使用同一 `SECRETS_DATABASE_URL`，进程启动时对会话存储执行迁移（见 `secrets-mcp` 中 `PostgresStore::migrate`），无需额外环境变量。
 
 | 位置 | 字段 | 说明 |
 |------|------|------|
 | entries | folder | 组织/隔离空间，如 `refining`、`ricnsmart`；参与唯一键 |
-| entries | type | 软分类，如 `server`、`service`、`key`、`person`（可扩展，不参与唯一键） |
+| entries | type | 软分类，用户自定义，如 `server`、`service`、`account`、`person`、`document`（不参与唯一键） |
 | entries | name | 人类可读标识；与 `folder` 一起在用户内唯一 |
 | entries | notes | 非敏感说明文本 |
-| entries | metadata | 明文 JSON（ip、url、`key_ref` 等） |
+| entries | metadata | 明文 JSON（ip、url、subtype 等） |
-| secrets | field_name | 明文字段名，便于 schema 展示 |
+| secrets | name | 密钥名称（调用方提供） |
+| secrets | type | 密钥类型（调用方提供，默认 `text`） |
 | secrets | encrypted | AES-GCM 密文（含 nonce） |
 | users | key_salt | PBKDF2 salt（32B），首次设置密码短语时写入 |
 | users | key_check | 派生密钥加密已知常量，用于验证密码短语 |
 | users | key_params | 派生算法参数，如 `{"alg":"pbkdf2-sha256","iterations":600000}` |
 
-### PEM 共享（`key_ref`）
+### 共享密钥（N:N 关联）
 
-同一 PEM 可被多条 `server` 等记录引用：建议将 PEM 存为 **`type=key`** 的 entry，在其它条目的 `metadata.key_ref` 中写目标 entry 的 `name`（支持 `folder/name` 格式消歧）；轮换时只更新该目标记录即可。
+多个条目可共享同一密文字段，通过 `entry_secrets` 中间表实现 N:N 关联：
-删除共享 key 时，系统会自动迁移引用：将密文复制到首个引用方（单副本），其余引用方的 `key_ref` 自动重定向到该新 owner，再删除原 key 记录。
+- 添加条目时可通过 `link_secret_names` 参数关联已有的 secret（按 `(user_id, name)` 精确匹配查找）
+- 同一 secret 可被多个 entry 引用，删除某 entry 不会级联删除被共享的 secret
+- 当 secret 不再被任何 entry 引用时，自动清理（`NOT EXISTS` 子查询）
+
+### 类型（Type）
+
+`type` 字段用于软分类，由用户自由填写，不做任何自动转换或归一化。常见示例：`server`、`service`、`account`、`person`、`document`，但任何值均可接受。
 
 ## 审计日志
 
@@ -188,10 +216,18 @@ LIMIT 20;
 ```
 Cargo.toml
 crates/secrets-core/     # db / crypto / models / audit / service
+  src/
+    taxonomy.rs          # SECRET_TYPE_OPTIONS（secret 字段类型下拉选项）
+    service/             # 业务逻辑（add, search, update, delete, export, env_map 等）
 crates/secrets-mcp/     # MCP HTTP、Web、OAuth、API Key
 scripts/
-  migrate-v0.3.0.sql    # 可选：手动 SQL 迁移（namespace/kind → folder/type、唯一键含 folder）
+  release-check.sh      # 发版前 fmt / clippy / test
-deploy/                 # systemd、.env 示例
+  setup-gitea-actions.sh
+  sync-test-to-prod.sh  # 测试库同步到生产（按需）
+deploy/
+  .env.example          # 环境变量模板
+  secrets-mcp.service   # systemd 服务文件（生产部署用）
+  postgres-tls-hardening.md  # PostgreSQL TLS 加固运维手册
 ```
 
 ## CI/CD（Gitea Actions）

crates/secrets-core/Cargo.toml

@@ -10,6 +10,7 @@ path = "src/lib.rs"
 [dependencies]
 aes-gcm.workspace = true
 anyhow.workspace = true
+thiserror.workspace = true
 chrono.workspace = true
 rand.workspace = true
 serde.workspace = true

crates/secrets-core/src/crypto.rs

@@ -5,6 +5,8 @@ use aes_gcm::{
 use anyhow::{Context, Result, bail};
 use serde_json::Value;
 
+use crate::error::AppError;
+
 const NONCE_LEN: usize = 12;
 
 // ─── AES-256-GCM encrypt / decrypt ───────────────────────────────────────────
@@ -38,7 +40,7 @@ pub fn decrypt(master_key: &[u8; 32], data: &[u8]) -> Result<Vec<u8>> {
     let nonce = Nonce::from_slice(nonce_bytes);
     cipher
         .decrypt(nonce, ciphertext)
-        .map_err(|_| anyhow::anyhow!("decryption failed — wrong master key or corrupted data"))
+        .map_err(|_| AppError::DecryptionFailed.into())
 }
 
 // ─── JSON helpers ─────────────────────────────────────────────────────────────

crates/secrets-core/src/error.rs

@@ -0,0 +1,142 @@
+use sqlx::error::DatabaseError;
+
+/// Structured business errors for the secrets service.
+///
+/// These replace ad-hoc `anyhow` strings for expected failure modes,
+/// allowing MCP and Web layers to map to appropriate protocol-level errors.
+#[derive(Debug, thiserror::Error)]
+pub enum AppError {
+    #[error("A secret with the name '{secret_name}' already exists for this user")]
+    ConflictSecretName { secret_name: String },
+
+    #[error("An entry with folder='{folder}' and name='{name}' already exists")]
+    ConflictEntryName { folder: String, name: String },
+
+    #[error("Entry not found")]
+    NotFoundEntry,
+
+    #[error("Validation failed: {message}")]
+    Validation { message: String },
+
+    #[error("Concurrent modification detected")]
+    ConcurrentModification,
+
+    #[error("Decryption failed — the encryption key may be incorrect")]
+    DecryptionFailed,
+
+    #[error(transparent)]
+    Internal(#[from] anyhow::Error),
+}
+
+impl AppError {
+    /// Try to convert a sqlx database error into a structured `AppError`.
+    ///
+    /// The caller should provide the context (which table was being written,
+    /// what values were being inserted) so we can produce a meaningful error.
+    pub fn from_db_error(err: sqlx::Error, ctx: DbErrorContext<'_>) -> Self {
+        if let sqlx::Error::Database(ref db_err) = err
+            && db_err.code().as_deref() == Some("23505")
+        {
+            return Self::from_unique_violation(db_err.as_ref(), ctx);
+        }
+        AppError::Internal(err.into())
+    }
+
+    fn from_unique_violation(db_err: &dyn DatabaseError, ctx: DbErrorContext<'_>) -> Self {
+        let constraint = db_err.constraint();
+
+        match constraint {
+            Some("idx_secrets_unique_user_name") => AppError::ConflictSecretName {
+                secret_name: ctx.secret_name.unwrap_or("unknown").to_string(),
+            },
+            Some("idx_entries_unique_user") | Some("idx_entries_unique_legacy") => {
+                AppError::ConflictEntryName {
+                    folder: ctx.folder.unwrap_or("").to_string(),
+                    name: ctx.name.unwrap_or("unknown").to_string(),
+                }
+            }
+            _ => {
+                // Fall back to message-based detection for unnamed constraints
+                let msg = db_err.message();
+                if msg.contains("secrets") {
+                    AppError::ConflictSecretName {
+                        secret_name: ctx.secret_name.unwrap_or("unknown").to_string(),
+                    }
+                } else {
+                    AppError::ConflictEntryName {
+                        folder: ctx.folder.unwrap_or("").to_string(),
+                        name: ctx.name.unwrap_or("unknown").to_string(),
+                    }
+                }
+            }
+        }
+    }
+}
+
+/// Context hints used when converting a database error to `AppError`.
+#[derive(Debug, Default, Clone, Copy)]
+pub struct DbErrorContext<'a> {
+    pub secret_name: Option<&'a str>,
+    pub folder: Option<&'a str>,
+    pub name: Option<&'a str>,
+}
+
+impl<'a> DbErrorContext<'a> {
+    pub fn secret_name(name: &'a str) -> Self {
+        Self {
+            secret_name: Some(name),
+            ..Default::default()
+        }
+    }
+
+    pub fn entry(folder: &'a str, name: &'a str) -> Self {
+        Self {
+            folder: Some(folder),
+            name: Some(name),
+            ..Default::default()
+        }
+    }
+}
+
+#[cfg(test)]
+mod tests {
+    use super::*;
+
+    #[test]
+    fn app_error_display_messages() {
+        let err = AppError::ConflictSecretName {
+            secret_name: "token".to_string(),
+        };
+        assert!(err.to_string().contains("token"));
+
+        let err = AppError::ConflictEntryName {
+            folder: "refining".to_string(),
+            name: "gitea".to_string(),
+        };
+        assert!(err.to_string().contains("refining"));
+        assert!(err.to_string().contains("gitea"));
+
+        let err = AppError::NotFoundEntry;
+        assert_eq!(err.to_string(), "Entry not found");
+
+        let err = AppError::Validation {
+            message: "too long".to_string(),
+        };
+        assert!(err.to_string().contains("too long"));
+
+        let err = AppError::ConcurrentModification;
+        assert!(err.to_string().contains("Concurrent modification"));
+    }
+
+    #[test]
+    fn db_error_context_helpers() {
+        let ctx = DbErrorContext::secret_name("my_key");
+        assert_eq!(ctx.secret_name, Some("my_key"));
+        assert!(ctx.folder.is_none());
+
+        let ctx = DbErrorContext::entry("prod", "db-creds");
+        assert_eq!(ctx.folder, Some("prod"));
+        assert_eq!(ctx.name, Some("db-creds"));
+        assert!(ctx.secret_name.is_none());
+    }
+}

crates/secrets-core/src/lib.rs

@@ -2,5 +2,7 @@ pub mod audit;
 pub mod config;
 pub mod crypto;
 pub mod db;
+pub mod error;
 pub mod models;
 pub mod service;
+pub mod taxonomy;

crates/secrets-core/src/models.rs

@@ -4,7 +4,7 @@ use serde_json::Value;
 use std::collections::BTreeMap;
 use uuid::Uuid;
 
-/// A top-level entry (server, service, key, person, …).
+/// A top-level entry (server, service, account, person, …).
 /// Sensitive fields are stored separately in `secrets`.
 #[derive(Debug, Serialize, Deserialize, sqlx::FromRow)]
 pub struct Entry {

crates/secrets-core/src/service/add.rs

@@ -7,6 +7,7 @@ use uuid::Uuid;
 
 use crate::crypto;
 use crate::db;
+use crate::error::{AppError, DbErrorContext};
 use crate::models::EntryRow;
 
 // ── Key/value parsing helpers ─────────────────────────────────────────────────
@@ -177,13 +178,18 @@ pub struct AddParams<'a> {
     pub tags: &'a [String],
     pub meta_entries: &'a [String],
     pub secret_entries: &'a [String],
+    pub secret_types: &'a std::collections::HashMap<String, String>,
     pub link_secret_names: &'a [String],
     /// Optional user_id for multi-user isolation (None = single-user CLI mode)
     pub user_id: Option<Uuid>,
 }
 
 pub async fn run(pool: &PgPool, params: AddParams<'_>, master_key: &[u8; 32]) -> Result<AddResult> {
-    let metadata = build_json(params.meta_entries)?;
+    let Value::Object(metadata_map) = build_json(params.meta_entries)? else {
+        unreachable!("build_json always returns a JSON object");
+    };
+    let entry_type = params.entry_type.trim();
+    let metadata = Value::Object(metadata_map);
     let secret_json = build_json(params.secret_entries)?;
     let meta_keys = collect_key_paths(params.meta_entries)?;
     let secret_keys = collect_key_paths(params.secret_entries)?;
@@ -224,7 +230,7 @@ pub async fn run(pool: &PgPool, params: AddParams<'_>, master_key: &[u8; 32]) ->
                 entry_id: ex.id,
                 user_id: params.user_id,
                 folder: params.folder,
-                entry_type: params.entry_type,
+                entry_type,
                 name: params.name,
                 version: ex.version,
                 action: "add",
@@ -254,7 +260,7 @@ pub async fn run(pool: &PgPool, params: AddParams<'_>, master_key: &[u8; 32]) ->
         )
         .bind(uid)
         .bind(params.folder)
-        .bind(params.entry_type)
+        .bind(entry_type)
         .bind(params.name)
         .bind(params.notes)
         .bind(params.tags)
@@ -277,7 +283,7 @@ pub async fn run(pool: &PgPool, params: AddParams<'_>, master_key: &[u8; 32]) ->
                RETURNING id"#,
         )
         .bind(params.folder)
-        .bind(params.entry_type)
+        .bind(entry_type)
         .bind(params.name)
         .bind(params.notes)
         .bind(params.tags)
@@ -299,7 +305,7 @@ pub async fn run(pool: &PgPool, params: AddParams<'_>, master_key: &[u8; 32]) ->
                 entry_id,
                 user_id: params.user_id,
                 folder: params.folder,
-                entry_type: params.entry_type,
+                entry_type,
                 name: params.name,
                 version: current_entry_version,
                 action: "create",
@@ -345,30 +351,42 @@ pub async fn run(pool: &PgPool, params: AddParams<'_>, master_key: &[u8; 32]) ->
             }
         }
 
+        let orphan_candidates: Vec<Uuid> = existing_fields.iter().map(|f| f.id).collect();
+
         sqlx::query("DELETE FROM entry_secrets WHERE entry_id = $1")
             .bind(entry_id)
             .execute(&mut *tx)
             .await?;
 
+        if !orphan_candidates.is_empty() {
             sqlx::query(
                 "DELETE FROM secrets s \
-             WHERE NOT EXISTS (SELECT 1 FROM entry_secrets es WHERE es.secret_id = s.id)",
+                 WHERE s.id = ANY($1) \
+                   AND NOT EXISTS (SELECT 1 FROM entry_secrets es WHERE es.secret_id = s.id)",
             )
+            .bind(&orphan_candidates)
             .execute(&mut *tx)
             .await?;
         }
+    }
 
     for (field_name, field_value) in &flat_fields {
         let encrypted = crypto::encrypt_json(master_key, field_value)?;
+        let secret_type = params
+            .secret_types
+            .get(field_name)
+            .map(|s| s.as_str())
+            .unwrap_or("text");
         let secret_id: Uuid = sqlx::query_scalar(
             "INSERT INTO secrets (user_id, name, type, encrypted) VALUES ($1, $2, $3, $4) RETURNING id",
         )
             .bind(params.user_id)
             .bind(field_name)
-            .bind(infer_secret_type(field_name))
+            .bind(secret_type)
             .bind(&encrypted)
             .fetch_one(&mut *tx)
-            .await?;
+            .await
+            .map_err(|e| AppError::from_db_error(e, DbErrorContext::secret_name(field_name)))?;
         sqlx::query("INSERT INTO entry_secrets (entry_id, secret_id) VALUES ($1, $2)")
             .bind(entry_id)
             .bind(secret_id)
@@ -414,7 +432,7 @@ pub async fn run(pool: &PgPool, params: AddParams<'_>, master_key: &[u8; 32]) ->
         params.user_id,
         "add",
         params.folder,
-        params.entry_type,
+        entry_type,
         params.name,
         serde_json::json!({
             "tags": params.tags,
@@ -429,32 +447,13 @@ pub async fn run(pool: &PgPool, params: AddParams<'_>, master_key: &[u8; 32]) ->
     Ok(AddResult {
         name: params.name.to_string(),
         folder: params.folder.to_string(),
-        entry_type: params.entry_type.to_string(),
+        entry_type: entry_type.to_string(),
         tags: params.tags.to_vec(),
         meta_keys,
         secret_keys,
     })
 }
 
-pub(crate) fn infer_secret_type(name: &str) -> &'static str {
-    match name {
-        "ssh_key" => "pem",
-        "password" => "password",
-        "phone" | "phone_2" => "phone",
-        "webhook_url" | "address" => "url",
-        "access_key_id"
-        | "access_key_secret"
-        | "global_api_key"
-        | "api_key"
-        | "secret_key"
-        | "personal_access_token"
-        | "runner_token"
-        | "GOOGLE_CLIENT_ID"
-        | "GOOGLE_CLIENT_SECRET" => "token",
-        _ => "text",
-    }
-}
-
 fn validate_link_secret_names(
     link_secret_names: &[String],
     new_secret_names: &BTreeSet<String>,
@@ -601,6 +600,7 @@ mod tests {
                 tags: &[],
                 meta_entries: &[],
                 secret_entries: &[],
+                secret_types: &Default::default(),
                 link_secret_names: std::slice::from_ref(&secret_name),
                 user_id: None,
             },
@@ -647,6 +647,7 @@ mod tests {
                 tags: &[],
                 meta_entries: &[],
                 secret_entries: &[],
+                secret_types: &Default::default(),
                 link_secret_names: std::slice::from_ref(&secret_name),
                 user_id: None,
             },
@@ -697,6 +698,7 @@ mod tests {
                 tags: &[],
                 meta_entries: &[],
                 secret_entries: &[],
+                secret_types: &Default::default(),
                 link_secret_names: std::slice::from_ref(&secret_name),
                 user_id: None,
             },
@@ -709,4 +711,69 @@ mod tests {
         cleanup_test_rows(&pool, &marker).await?;
         Ok(())
     }
+
+    #[tokio::test]
+    async fn add_duplicate_secret_name_returns_conflict_error() -> Result<()> {
+        let Some(pool) = maybe_test_pool().await else {
+            return Ok(());
+        };
+        let suffix = Uuid::from_u128(rand::random()).to_string();
+        let marker = format!("dup_secret_{}", &suffix[..8]);
+        let entry_name = format!("{}_entry", marker);
+        let secret_name = "shared_token";
+
+        cleanup_test_rows(&pool, &marker).await?;
+
+        // First add succeeds
+        run(
+            &pool,
+            AddParams {
+                name: &entry_name,
+                folder: &marker,
+                entry_type: "service",
+                notes: "",
+                tags: &[],
+                meta_entries: &[],
+                secret_entries: &[format!("{}=value1", secret_name)],
+                secret_types: &Default::default(),
+                link_secret_names: &[],
+                user_id: None,
+            },
+            &[0_u8; 32],
+        )
+        .await?;
+
+        // Second add with same secret name under same user_id should fail with ConflictSecretName
+        let entry_name2 = format!("{}_entry2", marker);
+        let err = run(
+            &pool,
+            AddParams {
+                name: &entry_name2,
+                folder: &marker,
+                entry_type: "service",
+                notes: "",
+                tags: &[],
+                meta_entries: &[],
+                secret_entries: &[format!("{}=value2", secret_name)],
+                secret_types: &Default::default(),
+                link_secret_names: &[],
+                user_id: None,
+            },
+            &[0_u8; 32],
+        )
+        .await
+        .expect_err("must fail on duplicate secret name");
+
+        let app_err = err
+            .downcast_ref::<crate::error::AppError>()
+            .expect("error should be AppError");
+        assert!(
+            matches!(app_err, crate::error::AppError::ConflictSecretName { .. }),
+            "expected ConflictSecretName, got: {}",
+            app_err
+        );
+
+        cleanup_test_rows(&pool, &marker).await?;
+        Ok(())
+    }
 }

crates/secrets-core/src/service/delete.rs

@@ -17,7 +17,6 @@ pub struct DeletedEntry {
 #[derive(Debug, serde::Serialize)]
 pub struct DeleteResult {
     pub deleted: Vec<DeletedEntry>,
-    pub migrated: Vec<String>,
     pub dry_run: bool,
 }
 
@@ -32,174 +31,6 @@ pub struct DeleteParams<'a> {
     pub user_id: Option<Uuid>,
 }
 
-#[derive(Debug, sqlx::FromRow)]
-struct KeyReferrer {
-    id: Uuid,
-    folder: String,
-    #[sqlx(rename = "type")]
-    entry_type: String,
-    name: String,
-}
-
-fn ref_label(r: &KeyReferrer) -> String {
-    format!("{}/{} ({})", r.folder, r.name, r.entry_type)
-}
-
-fn ref_path(r: &KeyReferrer) -> String {
-    format!("{}/{}", r.folder, r.name)
-}
-
-async fn fetch_key_referrers_pool(
-    pool: &PgPool,
-    key_entry_id: Uuid,
-    key_folder: &str,
-    key_name: &str,
-    user_id: Option<Uuid>,
-) -> Result<Vec<KeyReferrer>> {
-    let qualified = format!("{}/{}", key_folder, key_name);
-    let refs: Vec<KeyReferrer> = if let Some(uid) = user_id {
-        sqlx::query_as(
-            "SELECT id, folder, type, name FROM entries \
-             WHERE user_id = $1 AND id <> $2 \
-               AND (metadata->>'key_ref' = $3 OR metadata->>'key_ref' = $4) \
-             ORDER BY folder, type, name",
-        )
-        .bind(uid)
-        .bind(key_entry_id)
-        .bind(key_name)
-        .bind(&qualified)
-        .fetch_all(pool)
-        .await?
-    } else {
-        sqlx::query_as(
-            "SELECT id, folder, type, name FROM entries \
-             WHERE user_id IS NULL AND id <> $1 \
-               AND (metadata->>'key_ref' = $2 OR metadata->>'key_ref' = $3) \
-             ORDER BY folder, type, name",
-        )
-        .bind(key_entry_id)
-        .bind(key_name)
-        .bind(&qualified)
-        .fetch_all(pool)
-        .await?
-    };
-    Ok(refs)
-}
-
-async fn migrate_key_refs_if_needed(
-    tx: &mut sqlx::Transaction<'_, sqlx::Postgres>,
-    key_row: &EntryRow,
-    key_name: &str,
-    user_id: Option<Uuid>,
-    dry_run: bool,
-) -> Result<Vec<String>> {
-    let qualified = format!("{}/{}", key_row.folder, key_name);
-    let refs: Vec<KeyReferrer> = if let Some(uid) = user_id {
-        sqlx::query_as(
-            "SELECT id, folder, type, name FROM entries \
-             WHERE user_id = $1 AND id <> $2 \
-               AND (metadata->>'key_ref' = $3 OR metadata->>'key_ref' = $4) \
-             ORDER BY folder, type, name",
-        )
-        .bind(uid)
-        .bind(key_row.id)
-        .bind(key_name)
-        .bind(&qualified)
-        .fetch_all(&mut **tx)
-        .await?
-    } else {
-        sqlx::query_as(
-            "SELECT id, folder, type, name FROM entries \
-             WHERE user_id IS NULL AND id <> $1 \
-               AND (metadata->>'key_ref' = $2 OR metadata->>'key_ref' = $3) \
-             ORDER BY folder, type, name",
-        )
-        .bind(key_row.id)
-        .bind(key_name)
-        .bind(&qualified)
-        .fetch_all(&mut **tx)
-        .await?
-    };
-
-    if refs.is_empty() {
-        return Ok(vec![]);
-    }
-    if dry_run {
-        return Ok(refs.iter().map(ref_label).collect());
-    }
-
-    let owner = &refs[0];
-    let owner_path = ref_path(owner);
-    let key_fields: Vec<SecretFieldRow> = sqlx::query_as(
-        "SELECT s.id, s.name, s.encrypted \
-         FROM entry_secrets es \
-         JOIN secrets s ON s.id = es.secret_id \
-         WHERE es.entry_id = $1",
-    )
-    .bind(key_row.id)
-    .fetch_all(&mut **tx)
-    .await?;
-
-    for f in &key_fields {
-        sqlx::query("INSERT INTO entry_secrets (entry_id, secret_id) VALUES ($1, $2) ON CONFLICT DO NOTHING")
-        .bind(owner.id)
-        .bind(f.id)
-        .execute(&mut **tx)
-        .await?;
-    }
-
-    sqlx::query(
-        "UPDATE entries SET metadata = metadata - 'key_ref', \
-         version = version + 1, updated_at = NOW() WHERE id = $1",
-    )
-    .bind(owner.id)
-    .execute(&mut **tx)
-    .await?;
-
-    crate::audit::log_tx(
-        tx,
-        user_id,
-        "key_migrate",
-        &owner.folder,
-        &owner.entry_type,
-        &owner.name,
-        json!({
-            "from_key": format!("{}/{}", key_row.folder, key_name),
-            "role": "new_owner",
-            "redirect_target": owner_path,
-        }),
-    )
-    .await;
-
-    for r in refs.iter().skip(1) {
-        sqlx::query(
-            "UPDATE entries SET metadata = jsonb_set(metadata, '{key_ref}', to_jsonb($2::text), true), \
-             version = version + 1, updated_at = NOW() WHERE id = $1",
-        )
-        .bind(r.id)
-        .bind(&owner_path)
-        .execute(&mut **tx)
-        .await?;
-
-        crate::audit::log_tx(
-            tx,
-            user_id,
-            "key_migrate",
-            &r.folder,
-            &r.entry_type,
-            &r.name,
-            json!({
-                "from_key": format!("{}/{}", key_row.folder, key_name),
-                "role": "redirected_ref",
-                "redirect_to": owner_path,
-            }),
-        )
-        .await;
-    }
-
-    Ok(refs.iter().map(ref_label).collect())
-}
-
 /// Delete a single entry by id (multi-tenant: `user_id` must match).
 pub async fn delete_by_id(pool: &PgPool, entry_id: Uuid, user_id: Uuid) -> Result<DeleteResult> {
     let mut tx = pool.begin().await?;
@@ -224,8 +55,6 @@ pub async fn delete_by_id(pool: &PgPool, entry_id: Uuid, user_id: Uuid) -> Resul
     let entry_type = row.entry_type.clone();
     let name = row.name.clone();
     let entry_row: EntryRow = (&row).into();
-    let migrated =
-        migrate_key_refs_if_needed(&mut tx, &entry_row, &name, Some(user_id), false).await?;
 
     snapshot_and_delete(
         &mut tx,
@@ -254,7 +83,6 @@ pub async fn delete_by_id(pool: &PgPool, entry_id: Uuid, user_id: Uuid) -> Resul
             folder,
             entry_type,
         }],
-        migrated,
         dry_run: false,
     })
 }
@@ -294,6 +122,7 @@ async fn delete_one(
         // - 2+ matches → disambiguation error (same as non-dry-run)
         #[derive(sqlx::FromRow)]
         struct DryRunRow {
+            #[allow(dead_code)]
             id: Uuid,
             folder: String,
             #[sqlx(rename = "type")]
@@ -339,20 +168,16 @@ async fn delete_one(
         return match rows.len() {
             0 => Ok(DeleteResult {
                 deleted: vec![],
-                migrated: vec![],
                 dry_run: true,
             }),
             1 => {
                 let row = rows.into_iter().next().unwrap();
-                let refs =
-                    fetch_key_referrers_pool(pool, row.id, &row.folder, name, user_id).await?;
                 Ok(DeleteResult {
                     deleted: vec![DeletedEntry {
                         name: name.to_string(),
                         folder: row.folder,
                         entry_type: row.entry_type,
                     }],
-                    migrated: refs.iter().map(ref_label).collect(),
                     dry_run: true,
                 })
             }
@@ -417,7 +242,6 @@ async fn delete_one(
             tx.rollback().await?;
             return Ok(DeleteResult {
                 deleted: vec![],
-                migrated: vec![],
                 dry_run: false,
             });
         }
@@ -437,7 +261,6 @@ async fn delete_one(
 
     let folder = row.folder.clone();
     let entry_type = row.entry_type.clone();
-    let migrated = migrate_key_refs_if_needed(&mut tx, &row, name, user_id, false).await?;
     snapshot_and_delete(&mut tx, &folder, &entry_type, name, &row, user_id).await?;
     crate::audit::log_tx(
         &mut tx,
@@ -457,7 +280,6 @@ async fn delete_one(
             folder,
             entry_type,
         }],
-        migrated,
         dry_run: false,
     })
 }
@@ -497,14 +319,17 @@ async fn delete_bulk(
     }
     if entry_type.is_some() {
         conditions.push(format!("type = ${}", idx));
+        idx += 1;
     }
 
     let where_clause = format!("WHERE {}", conditions.join(" AND "));
+    let _ = idx; // used only for placeholder numbering in conditions
+
+    if dry_run {
         let sql = format!(
             "SELECT id, version, folder, type, name, metadata, tags, notes \
              FROM entries {where_clause} ORDER BY type, name"
         );
-
         let mut q = sqlx::query_as::<_, FullEntryRow>(&sql);
         if let Some(uid) = user_id {
             q = q.bind(uid);
@@ -517,13 +342,6 @@ async fn delete_bulk(
         }
         let rows = q.fetch_all(pool).await?;
 
-    if dry_run {
-        let mut migrated: Vec<String> = Vec::new();
-        for row in &rows {
-            let refs =
-                fetch_key_referrers_pool(pool, row.id, &row.folder, &row.name, user_id).await?;
-            migrated.extend(refs.iter().map(ref_label));
-        }
         let deleted = rows
             .iter()
             .map(|r| DeletedEntry {
@@ -534,15 +352,31 @@ async fn delete_bulk(
             .collect();
         return Ok(DeleteResult {
             deleted,
-            migrated,
             dry_run: true,
         });
     }
 
+    let mut tx = pool.begin().await?;
+
+    let sql = format!(
+        "SELECT id, version, folder, type, name, metadata, tags, notes \
+         FROM entries {where_clause} ORDER BY type, name FOR UPDATE"
+    );
+    let mut q = sqlx::query_as::<_, FullEntryRow>(&sql);
+    if let Some(uid) = user_id {
+        q = q.bind(uid);
+    }
+    if let Some(f) = folder {
+        q = q.bind(f);
+    }
+    if let Some(t) = entry_type {
+        q = q.bind(t);
+    }
+    let rows = q.fetch_all(&mut *tx).await?;
+
     let mut deleted = Vec::with_capacity(rows.len());
-    let mut migrated: Vec<String> = Vec::new();
     for row in &rows {
-        let entry_row = EntryRow {
+        let entry_row: EntryRow = EntryRow {
             id: row.id,
             version: row.version,
             folder: row.folder.clone(),
@@ -551,9 +385,6 @@ async fn delete_bulk(
             metadata: row.metadata.clone(),
             notes: row.notes.clone(),
         };
-        let mut tx = pool.begin().await?;
-        let m = migrate_key_refs_if_needed(&mut tx, &entry_row, &row.name, user_id, false).await?;
-        migrated.extend(m);
         snapshot_and_delete(
             &mut tx,
             &row.folder,
@@ -573,7 +404,6 @@ async fn delete_bulk(
             json!({"bulk": true}),
         )
         .await;
-        tx.commit().await?;
         deleted.push(DeletedEntry {
             name: row.name.clone(),
             folder: row.folder.clone(),
@@ -581,9 +411,10 @@ async fn delete_bulk(
         });
     }
 
+    tx.commit().await?;
+
     Ok(DeleteResult {
         deleted,
-        migrated,
         dry_run: false,
     })
 }
@@ -646,12 +477,17 @@ async fn snapshot_and_delete(
         .execute(&mut **tx)
         .await?;
 
+    let secret_ids: Vec<Uuid> = fields.iter().map(|f| f.id).collect();
+    if !secret_ids.is_empty() {
         sqlx::query(
             "DELETE FROM secrets s \
-         WHERE NOT EXISTS (SELECT 1 FROM entry_secrets es WHERE es.secret_id = s.id)",
+             WHERE s.id = ANY($1) \
+               AND NOT EXISTS (SELECT 1 FROM entry_secrets es WHERE es.secret_id = s.id)",
         )
+        .bind(&secret_ids)
         .execute(&mut **tx)
         .await?;
+    }
 
     Ok(())
 }
@@ -659,280 +495,153 @@ async fn snapshot_and_delete(
 #[cfg(test)]
 mod tests {
     use super::*;
-    use serde_json::json;
+    use sqlx::PgPool;
 
     async fn maybe_test_pool() -> Option<PgPool> {
         let Ok(url) = std::env::var("SECRETS_DATABASE_URL") else {
-            eprintln!("skip delete migration tests: SECRETS_DATABASE_URL is not set");
+            eprintln!("skip delete tests: SECRETS_DATABASE_URL is not set");
             return None;
         };
         let Ok(pool) = PgPool::connect(&url).await else {
-            eprintln!("skip delete migration tests: cannot connect to database");
+            eprintln!("skip delete tests: cannot connect to database");
             return None;
         };
         if let Err(e) = crate::db::migrate(&pool).await {
-            eprintln!("skip delete migration tests: migrate failed: {e}");
+            eprintln!("skip delete tests: migrate failed: {e}");
             return None;
         }
         Some(pool)
     }
 
-    async fn insert_entry(
+    async fn cleanup_single_user_rows(pool: &PgPool, marker: &str) -> Result<()> {
-        pool: &PgPool,
-        id: Uuid,
-        user_id: Uuid,
-        folder: &str,
-        entry_type: &str,
-        name: &str,
-        metadata: serde_json::Value,
-    ) -> Result<()> {
         sqlx::query(
-            "INSERT INTO entries (id, user_id, folder, type, name, notes, tags, metadata, version) \
+            "DELETE FROM entries WHERE user_id IS NULL AND (name LIKE $1 OR folder LIKE $1)",
-             VALUES ($1, $2, $3, $4, $5, '', ARRAY[]::text[], $6, 1)",
         )
-        .bind(id)
+        .bind(format!("%{marker}%"))
-        .bind(user_id)
-        .bind(folder)
-        .bind(entry_type)
-        .bind(name)
-        .bind(metadata)
         .execute(pool)
         .await?;
-        Ok(())
+        sqlx::query(
-    }
+            "DELETE FROM secrets WHERE user_id IS NULL AND name LIKE $1 \
-
+             AND NOT EXISTS (SELECT 1 FROM entry_secrets es WHERE es.secret_id = secrets.id)",
-    async fn insert_secret_for_entry(
-        pool: &PgPool,
-        user_id: Uuid,
-        entry_id: Uuid,
-        name: &str,
-        secret_type: &str,
-        encrypted: Vec<u8>,
-    ) -> Result<()> {
-        let secret_id: Uuid = sqlx::query_scalar(
-            "INSERT INTO secrets (user_id, name, type, encrypted) VALUES ($1, $2, $3, $4) RETURNING id",
         )
-        .bind(user_id)
+        .bind(format!("%{marker}%"))
-        .bind(name)
-        .bind(secret_type)
-        .bind(encrypted)
-        .fetch_one(pool)
-        .await?;
-        sqlx::query("INSERT INTO entry_secrets (entry_id, secret_id) VALUES ($1, $2)")
-            .bind(entry_id)
-            .bind(secret_id)
         .execute(pool)
         .await?;
         Ok(())
     }
 
     #[tokio::test]
-    async fn delete_shared_key_dry_run_reports_migration_without_writes() -> Result<()> {
+    async fn delete_dry_run_reports_matching_entry_without_writes() -> Result<()> {
         let Some(pool) = maybe_test_pool().await else {
             return Ok(());
         };
+        let suffix = Uuid::from_u128(rand::random()).to_string();
+        let marker = format!("delete_dry_{}", &suffix[..8]);
+        let entry_name = format!("{}_entry", marker);
 
-        let user_id = Uuid::from_u128(rand::random());
+        cleanup_single_user_rows(&pool, &marker).await?;
-        let key_id = Uuid::from_u128(rand::random());
-        let ref_a = Uuid::from_u128(rand::random());
-        let ref_b = Uuid::from_u128(rand::random());
 
-        insert_entry(
+        sqlx::query(
-            &pool,
+            "INSERT INTO entries (user_id, folder, type, name, notes, tags, metadata) \
-            key_id,
+             VALUES (NULL, $1, 'service', $2, '', '{}', '{}')",
-            user_id,
-            "kfolder",
-            "key",
-            "shared-key",
-            json!({}),
-        )
-        .await?;
-        insert_secret_for_entry(&pool, user_id, key_id, "pem", "pem", vec![1_u8, 2, 3]).await?;
-
-        insert_entry(
-            &pool,
-            ref_a,
-            user_id,
-            "afolder",
-            "server",
-            "srv-a",
-            json!({"key_ref":"kfolder/shared-key"}),
-        )
-        .await?;
-        insert_entry(
-            &pool,
-            ref_b,
-            user_id,
-            "bfolder",
-            "server",
-            "srv-b",
-            json!({"key_ref":"shared-key"}),
         )
+        .bind(&marker)
+        .bind(&entry_name)
+        .execute(&pool)
         .await?;
 
         let result = run(
             &pool,
             DeleteParams {
-                name: Some("shared-key"),
+                name: Some(&entry_name),
-                folder: Some("kfolder"),
+                folder: Some(&marker),
                 entry_type: None,
                 dry_run: true,
-                user_id: Some(user_id),
+                user_id: None,
             },
         )
         .await?;
 
         assert!(result.dry_run);
         assert_eq!(result.deleted.len(), 1);
-        assert_eq!(result.migrated.len(), 2);
+        assert_eq!(result.deleted[0].name, entry_name);
 
-        let key_exists: bool = sqlx::query_scalar(
+        let still_exists: bool = sqlx::query_scalar(
-            "SELECT EXISTS(SELECT 1 FROM entries WHERE id = $1 AND user_id = $2)",
+            "SELECT EXISTS(SELECT 1 FROM entries WHERE user_id IS NULL AND folder = $1 AND name = $2)",
         )
-        .bind(key_id)
+        .bind(&marker)
-        .bind(user_id)
+        .bind(&entry_name)
         .fetch_one(&pool)
         .await?;
-        assert!(key_exists);
+        assert!(still_exists);
 
-        let ref_a_key_ref: Option<String> =
+        cleanup_single_user_rows(&pool, &marker).await?;
-            sqlx::query_scalar("SELECT metadata->>'key_ref' FROM entries WHERE id = $1")
-                .bind(ref_a)
-                .fetch_one(&pool)
-                .await?;
-        let ref_b_key_ref: Option<String> =
-            sqlx::query_scalar("SELECT metadata->>'key_ref' FROM entries WHERE id = $1")
-                .bind(ref_b)
-                .fetch_one(&pool)
-                .await?;
-        assert_eq!(ref_a_key_ref.as_deref(), Some("kfolder/shared-key"));
-        assert_eq!(ref_b_key_ref.as_deref(), Some("shared-key"));
-
-        sqlx::query("DELETE FROM entries WHERE user_id = $1")
-            .bind(user_id)
-            .execute(&pool)
-            .await?;
         Ok(())
     }
 
     #[tokio::test]
-    async fn delete_shared_key_auto_migrates_single_copy_and_redirects_refs() -> Result<()> {
+    async fn delete_by_id_removes_entry_and_orphan_secret() -> Result<()> {
         let Some(pool) = maybe_test_pool().await else {
             return Ok(());
         };
-
+        let suffix = Uuid::from_u128(rand::random()).to_string();
+        let marker = format!("delete_id_{}", &suffix[..8]);
         let user_id = Uuid::from_u128(rand::random());
-        let key_id = Uuid::from_u128(rand::random());
+        let entry_name = format!("{}_entry", marker);
-        let ref_a = Uuid::from_u128(rand::random());
+        let secret_name = format!("{}_secret", marker);
-        let ref_b = Uuid::from_u128(rand::random());
-        let ref_c = Uuid::from_u128(rand::random());
 
-        insert_entry(
+        sqlx::query("DELETE FROM entries WHERE user_id = $1 AND folder = $2")
-            &pool,
-            key_id,
-            user_id,
-            "kfolder",
-            "key",
-            "shared-key",
-            json!({}),
-        )
-        .await?;
-        insert_secret_for_entry(&pool, user_id, key_id, "pem", "pem", vec![7_u8, 8, 9]).await?;
-
-        // owner candidate (sorted first by folder)
-        insert_entry(
-            &pool,
-            ref_a,
-            user_id,
-            "afolder",
-            "server",
-            "srv-a",
-            json!({"key_ref":"kfolder/shared-key"}),
-        )
-        .await?;
-        insert_entry(
-            &pool,
-            ref_b,
-            user_id,
-            "bfolder",
-            "server",
-            "srv-b",
-            json!({"key_ref":"shared-key"}),
-        )
-        .await?;
-        insert_entry(
-            &pool,
-            ref_c,
-            user_id,
-            "cfolder",
-            "service",
-            "svc-c",
-            json!({"key_ref":"kfolder/shared-key"}),
-        )
-        .await?;
-
-        let result = run(
-            &pool,
-            DeleteParams {
-                name: Some("shared-key"),
-                folder: Some("kfolder"),
-                entry_type: None,
-                dry_run: false,
-                user_id: Some(user_id),
-            },
-        )
-        .await?;
-
-        assert!(!result.dry_run);
-        assert_eq!(result.deleted.len(), 1);
-        assert_eq!(result.migrated.len(), 3);
-
-        let key_exists: bool = sqlx::query_scalar(
-            "SELECT EXISTS(SELECT 1 FROM entries WHERE id = $1 AND user_id = $2)",
-        )
-        .bind(key_id)
-        .bind(user_id)
-        .fetch_one(&pool)
-        .await?;
-        assert!(!key_exists);
-
-        let owner_key_ref: Option<String> =
-            sqlx::query_scalar("SELECT metadata->>'key_ref' FROM entries WHERE id = $1")
-                .bind(ref_a)
-                .fetch_one(&pool)
-                .await?;
-        let ref_b_key_ref: Option<String> =
-            sqlx::query_scalar("SELECT metadata->>'key_ref' FROM entries WHERE id = $1")
-                .bind(ref_b)
-                .fetch_one(&pool)
-                .await?;
-        let ref_c_key_ref: Option<String> =
-            sqlx::query_scalar("SELECT metadata->>'key_ref' FROM entries WHERE id = $1")
-                .bind(ref_c)
-                .fetch_one(&pool)
-                .await?;
-
-        assert_eq!(owner_key_ref, None);
-        assert_eq!(ref_b_key_ref.as_deref(), Some("afolder/srv-a"));
-        assert_eq!(ref_c_key_ref.as_deref(), Some("afolder/srv-a"));
-
-        let owner_has_copied: bool = sqlx::query_scalar(
-            "SELECT EXISTS( \
-                SELECT 1 \
-                FROM entry_secrets es \
-                JOIN secrets s ON s.id = es.secret_id \
-                WHERE es.entry_id = $1 AND s.name = 'pem' \
-            )",
-        )
-        .bind(ref_a)
-        .fetch_one(&pool)
-        .await?;
-        assert!(owner_has_copied);
-
-        sqlx::query("DELETE FROM entries WHERE user_id = $1")
             .bind(user_id)
+            .bind(&marker)
             .execute(&pool)
             .await?;
+        sqlx::query("DELETE FROM secrets WHERE user_id = $1 AND name = $2")
+            .bind(user_id)
+            .bind(&secret_name)
+            .execute(&pool)
+            .await?;
+
+        let entry_id: Uuid = sqlx::query_scalar(
+            "INSERT INTO entries (user_id, folder, type, name, notes, tags, metadata) \
+             VALUES ($1, $2, 'service', $3, '', '{}', '{}') RETURNING id",
+        )
+        .bind(user_id)
+        .bind(&marker)
+        .bind(&entry_name)
+        .fetch_one(&pool)
+        .await?;
+        let secret_id: Uuid = sqlx::query_scalar(
+            "INSERT INTO secrets (user_id, name, type, encrypted) VALUES ($1, $2, 'text', $3) RETURNING id",
+        )
+        .bind(user_id)
+        .bind(&secret_name)
+        .bind(vec![1_u8, 2, 3])
+        .fetch_one(&pool)
+        .await?;
+        sqlx::query("INSERT INTO entry_secrets (entry_id, secret_id) VALUES ($1, $2)")
+            .bind(entry_id)
+            .bind(secret_id)
+            .execute(&pool)
+            .await?;
+
+        let result = delete_by_id(&pool, entry_id, user_id).await?;
+        assert!(!result.dry_run);
+        assert_eq!(result.deleted.len(), 1);
+        assert_eq!(result.deleted[0].name, entry_name);
+
+        let entry_exists: bool =
+            sqlx::query_scalar("SELECT EXISTS(SELECT 1 FROM entries WHERE id = $1)")
+                .bind(entry_id)
+                .fetch_one(&pool)
+                .await?;
+        let secret_exists: bool =
+            sqlx::query_scalar("SELECT EXISTS(SELECT 1 FROM secrets WHERE id = $1)")
+                .bind(secret_id)
+                .fetch_one(&pool)
+                .await?;
+        assert!(!entry_exists);
+        assert!(!secret_exists);
+
         Ok(())
     }
 }

crates/secrets-core/src/service/env_map.rs

@@ -40,7 +40,7 @@ async fn build_entry_env_map(
     only_fields: &[String],
     prefix: &str,
     master_key: &[u8; 32],
-    user_id: Option<Uuid>,
+    _user_id: Option<Uuid>,
 ) -> Result<HashMap<String, String>> {
     let entry_ids = vec![entry.id];
     let secrets_map = fetch_secrets_for_entries(pool, &entry_ids).await?;
@@ -68,44 +68,6 @@ async fn build_entry_env_map(
         map.insert(key, json_to_env_string(&decrypted));
     }
 
-    // Resolve key_ref. Supported formats: "name" or "folder/name".
-    if let Some(key_ref) = entry.metadata.get("key_ref").and_then(|v| v.as_str()) {
-        let (ref_folder, ref_name) = if let Some((f, n)) = key_ref.split_once('/') {
-            (Some(f), n)
-        } else {
-            (None, key_ref)
-        };
-        let key_entries =
-            fetch_entries(pool, ref_folder, None, Some(ref_name), &[], None, user_id).await?;
-
-        if key_entries.len() > 1 {
-            anyhow::bail!(
-                "key_ref '{}' matched {} entries; qualify with folder/name to resolve the ambiguity",
-                key_ref,
-                key_entries.len()
-            );
-        }
-
-        if let Some(key_entry) = key_entries.first() {
-            let key_ids = vec![key_entry.id];
-            let key_fields_map = fetch_secrets_for_entries(pool, &key_ids).await?;
-            let empty = vec![];
-            let key_fields = key_fields_map.get(&key_entry.id).unwrap_or(&empty);
-            let key_prefix = env_prefix(key_entry, prefix);
-            for f in key_fields {
-                let decrypted = crypto::decrypt_json(master_key, &f.encrypted)?;
-                let key_var = format!(
-                    "{}_{}",
-                    key_prefix,
-                    f.name.to_uppercase().replace(['-', '.'], "_")
-                );
-                map.insert(key_var, json_to_env_string(&decrypted));
-            }
-        } else {
-            tracing::warn!(key_ref, ?user_id, "key_ref target not found");
-        }
-    }
-
     Ok(map)
 }
 

crates/secrets-core/src/service/import.rs

@@ -85,6 +85,7 @@ pub async fn run(
                 tags: &entry.tags,
                 meta_entries: &meta_entries,
                 secret_entries: &secret_entries,
+                secret_types: &Default::default(),
                 link_secret_names: &[],
                 user_id: params.user_id,
             },

crates/secrets-core/src/service/search.rs

@@ -8,10 +8,23 @@ use crate::models::{Entry, SecretField};
 
 pub const FETCH_ALL_LIMIT: u32 = 100_000;
 
+/// Build an ILIKE pattern for fuzzy matching, escaping `%` and `_` literals.
+pub fn ilike_pattern(value: &str) -> String {
+    format!(
+        "%{}%",
+        value
+            .replace('\\', "\\\\")
+            .replace('%', "\\%")
+            .replace('_', "\\_")
+    )
+}
+
 pub struct SearchParams<'a> {
     pub folder: Option<&'a str>,
     pub entry_type: Option<&'a str>,
     pub name: Option<&'a str>,
+    /// Fuzzy match on `entries.name` only (ILIKE with escaped `%`/`_`).
+    pub name_query: Option<&'a str>,
     pub tags: &'a [String],
     pub query: Option<&'a str>,
     pub sort: &'a str,
@@ -51,11 +64,15 @@ pub async fn count_entries(pool: &PgPool, a: &SearchParams<'_>) -> Result<i64> {
     if let Some(v) = a.name {
         q = q.bind(v);
     }
+    if let Some(v) = a.name_query {
+        let pattern = ilike_pattern(v);
+        q = q.bind(pattern);
+    }
     for tag in a.tags {
         q = q.bind(tag);
     }
     if let Some(v) = a.query {
-        let pattern = format!("%{}%", v.replace('%', "\\%").replace('_', "\\_"));
+        let pattern = ilike_pattern(v);
         q = q.bind(pattern);
     }
     let n = q.fetch_one(pool).await?;
@@ -86,6 +103,10 @@ fn entry_where_clause_and_next_idx(a: &SearchParams<'_>) -> (String, i32) {
         conditions.push(format!("name = ${}", idx));
         idx += 1;
     }
+    if a.name_query.is_some() {
+        conditions.push(format!("name ILIKE ${} ESCAPE '\\'", idx));
+        idx += 1;
+    }
     if !a.tags.is_empty() {
         let placeholders: Vec<String> = a
             .tags
@@ -135,6 +156,7 @@ pub async fn run(pool: &PgPool, params: SearchParams<'_>) -> Result<SearchResult
 }
 
 /// Fetch entries matching the given filters — returns all matching entries up to FETCH_ALL_LIMIT.
+#[allow(clippy::too_many_arguments)]
 pub async fn fetch_entries(
     pool: &PgPool,
     folder: Option<&str>,
@@ -148,6 +170,7 @@ pub async fn fetch_entries(
         folder,
         entry_type,
         name,
+        name_query: None,
         tags,
         query,
         sort: "name",
@@ -189,11 +212,15 @@ async fn fetch_entries_paged(pool: &PgPool, a: &SearchParams<'_>) -> Result<Vec<
     if let Some(v) = a.name {
         q = q.bind(v);
     }
+    if let Some(v) = a.name_query {
+        let pattern = ilike_pattern(v);
+        q = q.bind(pattern);
+    }
     for tag in a.tags {
         q = q.bind(tag);
     }
     if let Some(v) = a.query {
-        let pattern = format!("%{}%", v.replace('%', "\\%").replace('_', "\\_"));
+        let pattern = ilike_pattern(v);
         q = q.bind(pattern);
     }
     q = q.bind(a.limit as i64).bind(a.offset as i64);
@@ -384,3 +411,13 @@ impl EntrySecretRow {
         }
     }
 }
+
+#[cfg(test)]
+mod tests {
+    use super::*;
+
+    #[test]
+    fn ilike_pattern_escapes_backslash_percent_and_underscore() {
+        assert_eq!(ilike_pattern(r"hello\_100%"), r"%hello\\\_100\%%");
+    }
+}

crates/secrets-core/src/service/update.rs

@@ -5,10 +5,11 @@ use uuid::Uuid;
 
 use crate::crypto;
 use crate::db;
+use crate::error::{AppError, DbErrorContext};
 use crate::models::{EntryRow, EntryWriteRow};
 use crate::service::add::{
-    collect_field_paths, collect_key_paths, flatten_json_fields, infer_secret_type, insert_path,
+    collect_field_paths, collect_key_paths, flatten_json_fields, insert_path, parse_key_path,
-    parse_key_path, parse_kv, remove_path,
+    parse_kv, remove_path,
 };
 
 #[derive(Debug, serde::Serialize)]
@@ -23,6 +24,8 @@ pub struct UpdateResult {
     pub remove_meta: Vec<String>,
     pub secret_keys: Vec<String>,
     pub remove_secrets: Vec<String>,
+    pub linked_secrets: Vec<String>,
+    pub unlinked_secrets: Vec<String>,
 }
 
 pub struct UpdateParams<'a> {
@@ -35,7 +38,10 @@ pub struct UpdateParams<'a> {
     pub meta_entries: &'a [String],
     pub remove_meta: &'a [String],
     pub secret_entries: &'a [String],
+    pub secret_types: &'a std::collections::HashMap<String, String>,
     pub remove_secrets: &'a [String],
+    pub link_secret_names: &'a [String],
+    pub unlink_secret_names: &'a [String],
     pub user_id: Option<Uuid>,
 }
 
@@ -90,10 +96,7 @@ pub async fn run(
     let row = match rows.len() {
         0 => {
             tx.rollback().await?;
-            anyhow::bail!(
+            return Err(AppError::NotFoundEntry.into());
-                "Not found: '{}'. Use `add` to create it first.",
-                params.name
-            )
         }
         1 => rows.into_iter().next().unwrap(),
         _ => {
@@ -167,10 +170,7 @@ pub async fn run(
 
     if result.rows_affected() == 0 {
         tx.rollback().await?;
-        anyhow::bail!(
+        return Err(AppError::ConcurrentModification.into());
-            "Concurrent modification detected for '{}'. Please retry.",
-            params.name
-        );
     }
 
     for entry in params.secret_entries {
@@ -224,15 +224,21 @@ pub async fn run(
                 .execute(&mut *tx)
                 .await?;
             } else {
+                let secret_type = params
+                    .secret_types
+                    .get(field_name)
+                    .map(|s| s.as_str())
+                    .unwrap_or("text");
                 let secret_id: Uuid = sqlx::query_scalar(
                     "INSERT INTO secrets (user_id, name, type, encrypted) VALUES ($1, $2, $3, $4) RETURNING id",
                 )
                 .bind(params.user_id)
-                .bind(field_name)
+                .bind(field_name.to_string())
-                .bind(infer_secret_type(field_name))
+                .bind(secret_type)
                 .bind(&encrypted)
                 .fetch_one(&mut *tx)
-                .await?;
+                .await
+                .map_err(|e| AppError::from_db_error(e, DbErrorContext::secret_name(field_name)))?;
                 sqlx::query("INSERT INTO entry_secrets (entry_id, secret_id) VALUES ($1, $2)")
                     .bind(row.id)
                     .bind(secret_id)
@@ -292,6 +298,101 @@ pub async fn run(
         }
     }
 
+    // Link existing secrets by name
+    let mut linked_secrets = Vec::new();
+    for link_name in params.link_secret_names {
+        let link_name = link_name.trim();
+        if link_name.is_empty() {
+            anyhow::bail!("link_secret_names contains an empty name");
+        }
+        let secret_ids: Vec<Uuid> = if let Some(uid) = params.user_id {
+            sqlx::query_scalar("SELECT id FROM secrets WHERE user_id = $1 AND name = $2")
+                .bind(uid)
+                .bind(link_name)
+                .fetch_all(&mut *tx)
+                .await?
+        } else {
+            sqlx::query_scalar("SELECT id FROM secrets WHERE user_id IS NULL AND name = $1")
+                .bind(link_name)
+                .fetch_all(&mut *tx)
+                .await?
+        };
+
+        match secret_ids.len() {
+            0 => anyhow::bail!("Not found: secret named '{}'", link_name),
+            1 => {
+                sqlx::query(
+                    "INSERT INTO entry_secrets (entry_id, secret_id) VALUES ($1, $2) ON CONFLICT DO NOTHING",
+                )
+                .bind(row.id)
+                .bind(secret_ids[0])
+                .execute(&mut *tx)
+                .await?;
+                linked_secrets.push(link_name.to_string());
+            }
+            n => anyhow::bail!(
+                "Ambiguous: {} secrets named '{}' found. Please deduplicate names first.",
+                n,
+                link_name
+            ),
+        }
+    }
+
+    // Unlink secrets by name
+    let mut unlinked_secrets = Vec::new();
+    for unlink_name in params.unlink_secret_names {
+        let unlink_name = unlink_name.trim();
+        if unlink_name.is_empty() {
+            continue;
+        }
+
+        #[derive(sqlx::FromRow)]
+        struct SecretToUnlink {
+            id: Uuid,
+            encrypted: Vec<u8>,
+        }
+        let secret: Option<SecretToUnlink> = sqlx::query_as(
+            "SELECT s.id, s.encrypted \
+             FROM entry_secrets es \
+             JOIN secrets s ON s.id = es.secret_id \
+             WHERE es.entry_id = $1 AND s.name = $2",
+        )
+        .bind(row.id)
+        .bind(unlink_name)
+        .fetch_optional(&mut *tx)
+        .await?;
+
+        if let Some(s) = secret {
+            if let Err(e) = db::snapshot_secret_history(
+                &mut tx,
+                db::SecretSnapshotParams {
+                    secret_id: s.id,
+                    name: unlink_name,
+                    encrypted: &s.encrypted,
+                    action: "delete",
+                },
+            )
+            .await
+            {
+                tracing::warn!(error = %e, "failed to snapshot secret field history before unlink");
+            }
+            sqlx::query("DELETE FROM entry_secrets WHERE entry_id = $1 AND secret_id = $2")
+                .bind(row.id)
+                .bind(s.id)
+                .execute(&mut *tx)
+                .await?;
+            sqlx::query(
+                "DELETE FROM secrets s \
+                 WHERE s.id = $1 \
+                   AND NOT EXISTS (SELECT 1 FROM entry_secrets es WHERE es.secret_id = s.id)",
+            )
+            .bind(s.id)
+            .execute(&mut *tx)
+            .await?;
+            unlinked_secrets.push(unlink_name.to_string());
+        }
+    }
+
     let meta_keys = collect_key_paths(params.meta_entries)?;
     let remove_meta_keys = collect_field_paths(params.remove_meta)?;
     let secret_keys = collect_key_paths(params.secret_entries)?;
@@ -311,6 +412,8 @@ pub async fn run(
             "remove_meta": remove_meta_keys,
             "secret_keys": secret_keys,
             "remove_secrets": remove_secret_keys,
+            "linked_secrets": linked_secrets,
+            "unlinked_secrets": unlinked_secrets,
         }),
     )
     .await;
@@ -327,6 +430,8 @@ pub async fn run(
         remove_meta: remove_meta_keys,
         secret_keys,
         remove_secrets: remove_secret_keys,
+        linked_secrets,
+        unlinked_secrets,
     })
 }
 
@@ -347,13 +452,13 @@ pub async fn update_fields_by_id(
     user_id: Uuid,
     params: UpdateEntryFieldsByIdParams<'_>,
 ) -> Result<()> {
-    if params.folder.len() > 128 {
+    if params.folder.chars().count() > 128 {
         anyhow::bail!("folder must be at most 128 characters");
     }
-    if params.entry_type.len() > 64 {
+    if params.entry_type.chars().count() > 64 {
         anyhow::bail!("type must be at most 64 characters");
     }
-    if params.name.len() > 256 {
+    if params.name.chars().count() > 256 {
         anyhow::bail!("name must be at most 256 characters");
     }
 
@@ -372,7 +477,7 @@ pub async fn update_fields_by_id(
         Some(r) => r,
         None => {
             tx.rollback().await?;
-            anyhow::bail!("Entry not found");
+            return Err(AppError::NotFoundEntry.into());
         }
     };
 
@@ -395,13 +500,15 @@ pub async fn update_fields_by_id(
         tracing::warn!(error = %e, "failed to snapshot entry history before web update");
     }
 
+    let entry_type = params.entry_type.trim();
+
     let res = sqlx::query(
         "UPDATE entries SET folder = $1, type = $2, name = $3, notes = $4, tags = $5, metadata = $6, \
          version = version + 1, updated_at = NOW() \
          WHERE id = $7 AND version = $8",
     )
     .bind(params.folder)
-    .bind(params.entry_type)
+    .bind(entry_type)
     .bind(params.name)
     .bind(params.notes)
     .bind(params.tags)
@@ -414,16 +521,17 @@ pub async fn update_fields_by_id(
         if let sqlx::Error::Database(ref d) = e
             && d.code().as_deref() == Some("23505")
         {
-            return anyhow::anyhow!(
+            return AppError::ConflictEntryName {
-                "An entry with this folder and name already exists for your account."
+                folder: params.folder.to_string(),
-            );
+                name: params.name.to_string(),
+            };
         }
-        e.into()
+        AppError::Internal(e.into())
     })?;
 
     if res.rows_affected() == 0 {
         tx.rollback().await?;
-        anyhow::bail!("Concurrent modification detected. Please refresh and try again.");
+        return Err(AppError::ConcurrentModification.into());
     }
 
     crate::audit::log_tx(
@@ -431,7 +539,7 @@ pub async fn update_fields_by_id(
         Some(user_id),
         "update",
         params.folder,
-        params.entry_type,
+        entry_type,
         params.name,
         serde_json::json!({
             "source": "web",

crates/secrets-core/src/taxonomy.rs

@@ -0,0 +1,4 @@
+/// Canonical secret type options for UI dropdowns.
+pub const SECRET_TYPE_OPTIONS: &[&str] = &[
+    "text", "password", "token", "api-key", "ssh-key", "url", "phone", "id-card",
+];

crates/secrets-mcp/Cargo.toml

@@ -1,6 +1,6 @@
 [package]
 name = "secrets-mcp"
-version = "0.3.8"
+version = "0.5.1"
 edition.workspace = true
 
 [[bin]]

crates/secrets-mcp/src/error.rs

@@ -0,0 +1,40 @@
+use secrets_core::error::AppError;
+
+/// Map a structured `AppError` to an MCP protocol error.
+///
+/// This replaces the previous pattern of swallowing all errors into `-32603`.
+pub fn app_error_to_mcp(err: &AppError) -> rmcp::ErrorData {
+    match err {
+        AppError::ConflictSecretName { secret_name } => rmcp::ErrorData::invalid_request(
+            format!(
+                "A secret with the name '{secret_name}' already exists for your account. \
+                 Secret names must be unique per user."
+            ),
+            None,
+        ),
+        AppError::ConflictEntryName { folder, name } => rmcp::ErrorData::invalid_request(
+            format!(
+                "An entry with folder='{folder}' and name='{name}' already exists. \
+                 The combination of folder and name must be unique."
+            ),
+            None,
+        ),
+        AppError::NotFoundEntry => rmcp::ErrorData::invalid_request(
+            "Entry not found. Use secrets_find to discover existing entries.",
+            None,
+        ),
+        AppError::Validation { message } => rmcp::ErrorData::invalid_request(message.clone(), None),
+        AppError::ConcurrentModification => rmcp::ErrorData::invalid_request(
+            "The entry was modified by another request. Please refresh and try again.",
+            None,
+        ),
+        AppError::DecryptionFailed => rmcp::ErrorData::invalid_request(
+            "Decryption failed — the encryption key may be incorrect or does not match the data.",
+            None,
+        ),
+        AppError::Internal(_) => rmcp::ErrorData::internal_error(
+            "Request failed due to a server error. Check service logs if you need details.",
+            None,
+        ),
+    }
+}

crates/secrets-mcp/src/main.rs

@@ -1,4 +1,5 @@
 mod auth;
+mod error;
 mod logging;
 mod oauth;
 mod tools;
@@ -40,6 +41,14 @@ fn load_env_var(name: &str) -> Option<String> {
     std::env::var(name).ok().filter(|s| !s.is_empty())
 }
 
+/// Pretty-print bind address in logs (`127.0.0.1` → `localhost`); actual socket bind unchanged.
+fn listen_addr_log_display(bind_addr: &str) -> String {
+    bind_addr
+        .strip_prefix("127.0.0.1:")
+        .map(|port| format!("localhost:{port}"))
+        .unwrap_or_else(|| bind_addr.to_string())
+}
+
 fn load_oauth_config(prefix: &str, base_url: &str, path: &str) -> Option<OAuthConfig> {
     let client_id = load_env_var(&format!("{}_CLIENT_ID", prefix))?;
     let client_secret = load_env_var(&format!("{}_CLIENT_SECRET", prefix))?;
@@ -168,7 +177,10 @@ async fn main() -> Result<()> {
         .await
         .with_context(|| format!("failed to bind to {}", bind_addr))?;
 
-    tracing::info!("Secrets MCP Server listening on http://{}", bind_addr);
+    tracing::info!(
+        "Secrets MCP Server listening on http://{}",
+        listen_addr_log_display(&bind_addr)
+    );
     tracing::info!("MCP endpoint: {}/mcp", base_url);
 
     axum::serve(

crates/secrets-mcp/src/tools.rs

@@ -13,11 +13,153 @@ use rmcp::{
     tool, tool_handler, tool_router,
 };
 use schemars::JsonSchema;
-use serde::Deserialize;
+use serde::{Deserialize, Deserializer, de};
 use serde_json::{Map, Value};
 use sqlx::PgPool;
 use uuid::Uuid;
 
+// ── Serde helpers for numeric parameters that may arrive as strings ──────────
+
+mod deser {
+    use super::*;
+
+    /// Deserialize a value that may come as a JSON number or a JSON string.
+    pub fn option_u32_from_string<'de, D>(deserializer: D) -> Result<Option<u32>, D::Error>
+    where
+        D: Deserializer<'de>,
+    {
+        #[derive(Deserialize)]
+        #[serde(untagged)]
+        enum NumOrStr {
+            Num(u32),
+            Str(String),
+        }
+
+        match Option::<NumOrStr>::deserialize(deserializer)? {
+            None => Ok(None),
+            Some(NumOrStr::Num(n)) => Ok(Some(n)),
+            Some(NumOrStr::Str(s)) => {
+                if s.is_empty() {
+                    return Ok(None);
+                }
+                s.parse::<u32>().map(Some).map_err(de::Error::custom)
+            }
+        }
+    }
+
+    /// Deserialize an i64 that may come as a JSON number or a JSON string.
+    pub fn option_i64_from_string<'de, D>(deserializer: D) -> Result<Option<i64>, D::Error>
+    where
+        D: Deserializer<'de>,
+    {
+        #[derive(Deserialize)]
+        #[serde(untagged)]
+        enum NumOrStr {
+            Num(i64),
+            Str(String),
+        }
+
+        match Option::<NumOrStr>::deserialize(deserializer)? {
+            None => Ok(None),
+            Some(NumOrStr::Num(n)) => Ok(Some(n)),
+            Some(NumOrStr::Str(s)) => {
+                if s.is_empty() {
+                    return Ok(None);
+                }
+                s.parse::<i64>().map(Some).map_err(de::Error::custom)
+            }
+        }
+    }
+
+    /// Deserialize a bool that may come as a JSON bool or a JSON string ("true"/"false").
+    pub fn option_bool_from_string<'de, D>(deserializer: D) -> Result<Option<bool>, D::Error>
+    where
+        D: Deserializer<'de>,
+    {
+        #[derive(Deserialize)]
+        #[serde(untagged)]
+        enum BoolOrStr {
+            Bool(bool),
+            Str(String),
+        }
+
+        match Option::<BoolOrStr>::deserialize(deserializer)? {
+            None => Ok(None),
+            Some(BoolOrStr::Bool(b)) => Ok(Some(b)),
+            Some(BoolOrStr::Str(s)) => {
+                if s.is_empty() {
+                    return Ok(None);
+                }
+                s.parse::<bool>().map(Some).map_err(de::Error::custom)
+            }
+        }
+    }
+
+    /// Deserialize a Vec<String> that may come as a JSON array or a JSON string containing an array.
+    pub fn option_vec_string_from_string<'de, D>(
+        deserializer: D,
+    ) -> Result<Option<Vec<String>>, D::Error>
+    where
+        D: Deserializer<'de>,
+    {
+        #[derive(Deserialize)]
+        #[serde(untagged)]
+        enum VecOrStr {
+            Vec(Vec<String>),
+            Str(String),
+        }
+
+        match Option::<VecOrStr>::deserialize(deserializer)? {
+            None => Ok(None),
+            Some(VecOrStr::Vec(v)) => Ok(Some(v)),
+            Some(VecOrStr::Str(s)) => {
+                if s.is_empty() {
+                    return Ok(None);
+                }
+                serde_json::from_str(&s)
+                    .map(Some)
+                    .map_err(|e| {
+                        de::Error::custom(format!(
+                            "invalid string value for array field: expected a JSON array, e.g. '[\"a\",\"b\"]': {e}"
+                        ))
+                    })
+            }
+        }
+    }
+
+    /// Deserialize a Map<String, Value> that may come as a JSON object or a JSON string containing an object.
+    pub fn option_map_from_string<'de, D>(
+        deserializer: D,
+    ) -> Result<Option<Map<String, Value>>, D::Error>
+    where
+        D: Deserializer<'de>,
+    {
+        #[derive(Deserialize)]
+        #[serde(untagged)]
+        enum MapOrStr {
+            Map(Map<String, Value>),
+            Str(String),
+        }
+
+        match Option::<MapOrStr>::deserialize(deserializer)? {
+            None => Ok(None),
+            Some(MapOrStr::Map(m)) => Ok(Some(m)),
+            Some(MapOrStr::Str(s)) => {
+                if s.is_empty() {
+                    return Ok(None);
+                }
+                serde_json::from_str(&s)
+                    .map(Some)
+                    .map_err(|e| {
+                        de::Error::custom(format!(
+                            "invalid string value for object field: expected a JSON object, e.g. '{{\"key\":\"value\"}}': {e}"
+                        ))
+                    })
+            }
+        }
+    }
+}
+
 use secrets_core::models::ExportFormat;
 use secrets_core::service::{
     add::{AddParams, run as svc_add},
@@ -31,6 +173,7 @@ use secrets_core::service::{
 };
 
 use crate::auth::AuthUser;
+use crate::error;
 
 // ── MCP client-facing errors (no internal details) ───────────────────────────
 
@@ -50,6 +193,17 @@ fn mcp_err_internal_logged(
     )
 }
 
+fn mcp_err_from_anyhow(
+    tool: &'static str,
+    user_id: Option<Uuid>,
+    err: anyhow::Error,
+) -> rmcp::ErrorData {
+    if let Some(app_err) = err.downcast_ref::<secrets_core::error::AppError>() {
+        return error::app_error_to_mcp(app_err);
+    }
+    mcp_err_internal_logged(tool, user_id, err)
+}
+
 fn mcp_err_invalid_encryption_key_logged(err: impl std::fmt::Display) -> rmcp::ErrorData {
     tracing::warn!(error = %err, "invalid X-Encryption-Key");
     rmcp::ErrorData::invalid_request(
@@ -162,14 +316,22 @@ struct FindInput {
     query: Option<String>,
     #[schemars(description = "Exact folder filter (e.g. 'refining', 'ricnsmart')")]
     folder: Option<String>,
-    #[schemars(description = "Exact type filter (e.g. 'server', 'service', 'person', 'key')")]
+    #[schemars(
+        description = "Exact type filter (e.g. 'server', 'service', 'account', 'person', 'document'). User-defined, any value accepted."
+    )]
     #[serde(rename = "type")]
     entry_type: Option<String>,
-    #[schemars(description = "Exact name filter")]
+    #[schemars(description = "Exact name filter. For fuzzy matching use name_query instead.")]
     name: Option<String>,
+    #[schemars(
+        description = "Fuzzy name filter (ILIKE, case-insensitive partial match). Use this instead of 'name' when you don't know the exact name."
+    )]
+    name_query: Option<String>,
     #[schemars(description = "Tag filters (all must match)")]
+    #[serde(default, deserialize_with = "deser::option_vec_string_from_string")]
     tags: Option<Vec<String>>,
     #[schemars(description = "Max results (default 20)")]
+    #[serde(default, deserialize_with = "deser::option_u32_from_string")]
     limit: Option<u32>,
 }
 
@@ -179,20 +341,30 @@ struct SearchInput {
     query: Option<String>,
     #[schemars(description = "Folder filter (e.g. 'refining', 'personal', 'family')")]
     folder: Option<String>,
-    #[schemars(description = "Type filter (e.g. 'server', 'service', 'person', 'key')")]
+    #[schemars(
+        description = "Type filter (e.g. 'server', 'service', 'account', 'person', 'document'). User-defined, any value accepted."
+    )]
     #[serde(rename = "type")]
     entry_type: Option<String>,
-    #[schemars(description = "Exact name to match")]
+    #[schemars(description = "Exact name to match. For fuzzy matching use name_query instead.")]
     name: Option<String>,
+    #[schemars(
+        description = "Fuzzy name filter (ILIKE, case-insensitive partial match). Use this instead of 'name' when you don't know the exact name."
+    )]
+    name_query: Option<String>,
     #[schemars(description = "Tag filters (all must match)")]
+    #[serde(default, deserialize_with = "deser::option_vec_string_from_string")]
     tags: Option<Vec<String>>,
     #[schemars(description = "Return only summary fields (name/tags/notes/updated_at)")]
+    #[serde(default, deserialize_with = "deser::option_bool_from_string")]
     summary: Option<bool>,
     #[schemars(description = "Sort order: 'name' (default), 'updated', 'created'")]
     sort: Option<String>,
     #[schemars(description = "Max results (default 20)")]
+    #[serde(default, deserialize_with = "deser::option_u32_from_string")]
     limit: Option<u32>,
     #[schemars(description = "Pagination offset (default 0)")]
+    #[serde(default, deserialize_with = "deser::option_u32_from_string")]
     offset: Option<u32>,
 }
 
@@ -211,31 +383,42 @@ struct AddInput {
     #[schemars(description = "Folder for organization (optional, e.g. 'personal', 'refining')")]
     folder: Option<String>,
     #[schemars(
-        description = "Type/category of this entry (optional, e.g. 'server', 'person', 'key')"
+        description = "Type/category of this entry (optional, e.g. 'server', 'service', 'account', 'person', 'document'). Free-form, choose what best describes the entry."
     )]
     #[serde(rename = "type")]
     entry_type: Option<String>,
     #[schemars(description = "Free-text notes for this entry (optional)")]
     notes: Option<String>,
     #[schemars(description = "Tags for this entry")]
+    #[serde(default, deserialize_with = "deser::option_vec_string_from_string")]
     tags: Option<Vec<String>>,
     #[schemars(description = "Metadata fields as 'key=value' or 'key:=json' strings")]
+    #[serde(default, deserialize_with = "deser::option_vec_string_from_string")]
     meta: Option<Vec<String>>,
     #[schemars(
         description = "Metadata fields as a JSON object {\"key\": value}. Merged with 'meta' if both provided."
     )]
+    #[serde(default, deserialize_with = "deser::option_map_from_string")]
     meta_obj: Option<Map<String, Value>>,
     #[schemars(
         description = "Secret fields as 'key=value' strings. Reminder: non-sensitive endpoint/address fields should go to metadata.address instead of secrets."
     )]
+    #[serde(default, deserialize_with = "deser::option_vec_string_from_string")]
     secrets: Option<Vec<String>>,
     #[schemars(
         description = "Secret fields as a JSON object {\"key\": \"value\"}. Merged with 'secrets' if both provided. Reminder: non-sensitive endpoint/address fields should go to metadata.address."
     )]
+    #[serde(default, deserialize_with = "deser::option_map_from_string")]
     secrets_obj: Option<Map<String, Value>>,
+    #[schemars(
+        description = "Secret types as {\"secret_name\": \"type\"}. Keys must match secret field names. Missing keys default to \"text\"."
+    )]
+    #[serde(default, deserialize_with = "deser::option_map_from_string")]
+    secret_types: Option<Map<String, Value>>,
     #[schemars(
         description = "Link existing secrets by secret name. Names must resolve uniquely under current user."
     )]
+    #[serde(default, deserialize_with = "deser::option_vec_string_from_string")]
     link_secret_names: Option<Vec<String>>,
 }
 
@@ -254,27 +437,50 @@ struct UpdateInput {
     #[schemars(description = "Update the notes field")]
     notes: Option<String>,
     #[schemars(description = "Tags to add")]
+    #[serde(default, deserialize_with = "deser::option_vec_string_from_string")]
     add_tags: Option<Vec<String>>,
     #[schemars(description = "Tags to remove")]
+    #[serde(default, deserialize_with = "deser::option_vec_string_from_string")]
     remove_tags: Option<Vec<String>>,
     #[schemars(description = "Metadata fields to update/add as 'key=value' strings")]
+    #[serde(default, deserialize_with = "deser::option_vec_string_from_string")]
     meta: Option<Vec<String>>,
     #[schemars(
         description = "Metadata fields to update/add as a JSON object {\"key\": value}. Merged with 'meta' if both provided."
     )]
+    #[serde(default, deserialize_with = "deser::option_map_from_string")]
     meta_obj: Option<Map<String, Value>>,
     #[schemars(description = "Metadata field keys to remove")]
+    #[serde(default, deserialize_with = "deser::option_vec_string_from_string")]
     remove_meta: Option<Vec<String>>,
     #[schemars(
         description = "Secret fields to update/add as 'key=value' strings. Reminder: non-sensitive endpoint/address fields should go to metadata.address instead of secrets."
     )]
+    #[serde(default, deserialize_with = "deser::option_vec_string_from_string")]
     secrets: Option<Vec<String>>,
     #[schemars(
         description = "Secret fields to update/add as a JSON object {\"key\": \"value\"}. Merged with 'secrets' if both provided. Reminder: non-sensitive endpoint/address fields should go to metadata.address."
     )]
+    #[serde(default, deserialize_with = "deser::option_map_from_string")]
     secrets_obj: Option<Map<String, Value>>,
+    #[schemars(
+        description = "Secret types as {\"secret_name\": \"type\"}. Keys must match secret field names. Missing keys default to \"text\"."
+    )]
+    #[serde(default, deserialize_with = "deser::option_map_from_string")]
+    secret_types: Option<Map<String, Value>>,
     #[schemars(description = "Secret field keys to remove")]
+    #[serde(default, deserialize_with = "deser::option_vec_string_from_string")]
     remove_secrets: Option<Vec<String>>,
+    #[schemars(
+        description = "Link existing secrets by name to this entry. Names must resolve uniquely under current user."
+    )]
+    #[serde(default, deserialize_with = "deser::option_vec_string_from_string")]
+    link_secret_names: Option<Vec<String>>,
+    #[schemars(
+        description = "Unlink secrets by name from this entry. Orphaned secrets are auto-deleted."
+    )]
+    #[serde(default, deserialize_with = "deser::option_vec_string_from_string")]
+    unlink_secret_names: Option<Vec<String>>,
 }
 
 #[derive(Debug, Deserialize, JsonSchema)]
@@ -293,6 +499,7 @@ struct DeleteInput {
     #[serde(rename = "type")]
     entry_type: Option<String>,
     #[schemars(description = "Preview deletions without writing")]
+    #[serde(default, deserialize_with = "deser::option_bool_from_string")]
     dry_run: Option<bool>,
 }
 
@@ -309,6 +516,7 @@ struct HistoryInput {
     )]
     id: Option<String>,
     #[schemars(description = "Max history entries to return (default 20)")]
+    #[serde(default, deserialize_with = "deser::option_u32_from_string")]
     limit: Option<u32>,
 }
 
@@ -325,6 +533,7 @@ struct RollbackInput {
     )]
     id: Option<String>,
     #[schemars(description = "Target version number. Omit to restore the most recent snapshot.")]
+    #[serde(default, deserialize_with = "deser::option_i64_from_string")]
     to_version: Option<i64>,
 }
 
@@ -338,6 +547,7 @@ struct ExportInput {
     #[schemars(description = "Exact name filter")]
     name: Option<String>,
     #[schemars(description = "Tag filters")]
+    #[serde(default, deserialize_with = "deser::option_vec_string_from_string")]
     tags: Option<Vec<String>>,
     #[schemars(description = "Fuzzy query")]
     query: Option<String>,
@@ -355,8 +565,10 @@ struct EnvMapInput {
     #[schemars(description = "Exact name filter")]
     name: Option<String>,
     #[schemars(description = "Tag filters")]
+    #[serde(default, deserialize_with = "deser::option_vec_string_from_string")]
     tags: Option<Vec<String>>,
     #[schemars(description = "Only include these secret fields")]
+    #[serde(default, deserialize_with = "deser::option_vec_string_from_string")]
     only_fields: Option<Vec<String>>,
     #[schemars(description = "Environment variable name prefix. \
         Variable names are built as UPPER(prefix)_UPPER(entry_name)_UPPER(field_name), \
@@ -412,6 +624,7 @@ impl SecretsService {
             folder = input.folder.as_deref(),
             entry_type = input.entry_type.as_deref(),
             name = input.name.as_deref(),
+            name_query = input.name_query.as_deref(),
             query = input.query.as_deref(),
             "tool call start",
         );
@@ -422,6 +635,7 @@ impl SecretsService {
                 folder: input.folder.as_deref(),
                 entry_type: input.entry_type.as_deref(),
                 name: input.name.as_deref(),
+                name_query: input.name_query.as_deref(),
                 tags: &tags,
                 query: input.query.as_deref(),
                 sort: "name",
@@ -499,6 +713,7 @@ impl SecretsService {
             folder = input.folder.as_deref(),
             entry_type = input.entry_type.as_deref(),
             name = input.name.as_deref(),
+            name_query = input.name_query.as_deref(),
             query = input.query.as_deref(),
             "tool call start",
         );
@@ -509,6 +724,7 @@ impl SecretsService {
                 folder: input.folder.as_deref(),
                 entry_type: input.entry_type.as_deref(),
                 name: input.name.as_deref(),
+                name_query: input.name_query.as_deref(),
                 tags: &tags,
                 query: input.query.as_deref(),
                 sort: input.sort.as_deref().unwrap_or("name"),
@@ -607,7 +823,7 @@ impl SecretsService {
             let value =
                 get_secret_field_by_id(&self.pool, entry_id, field_name, &user_key, Some(user_id))
                     .await
-                    .map_err(|e| mcp_err_internal_logged("secrets_get", None, e))?;
+                    .map_err(|e| mcp_err_from_anyhow("secrets_get", Some(user_id), e))?;
 
             tracing::info!(
                 tool = "secrets_get",
@@ -621,7 +837,7 @@ impl SecretsService {
         } else {
             let secrets = get_all_secrets_by_id(&self.pool, entry_id, &user_key, Some(user_id))
                 .await
-                .map_err(|e| mcp_err_internal_logged("secrets_get", None, e))?;
+                .map_err(|e| mcp_err_from_anyhow("secrets_get", Some(user_id), e))?;
 
             tracing::info!(
             tool = "secrets_get",
@@ -667,6 +883,11 @@ impl SecretsService {
         if let Some(obj) = input.secrets_obj {
             secrets.extend(map_to_kv_strings(obj));
         }
+        let secret_types = input.secret_types.unwrap_or_default();
+        let secret_types_map: std::collections::HashMap<String, String> = secret_types
+            .into_iter()
+            .filter_map(|(k, v)| v.as_str().map(|s| (k, s.to_string())))
+            .collect();
         let link_secret_names = input.link_secret_names.unwrap_or_default();
         let folder = input.folder.as_deref().unwrap_or("");
         let entry_type = input.entry_type.as_deref().unwrap_or("");
@@ -682,13 +903,14 @@ impl SecretsService {
                 tags: &tags,
                 meta_entries: &meta,
                 secret_entries: &secrets,
+                secret_types: &secret_types_map,
                 link_secret_names: &link_secret_names,
                 user_id: Some(user_id),
             },
             &user_key,
         )
         .await
-        .map_err(|e| mcp_err_internal_logged("secrets_add", Some(user_id), e))?;
+        .map_err(|e| mcp_err_from_anyhow("secrets_add", Some(user_id), e))?;
 
         tracing::info!(
             tool = "secrets_add",
@@ -745,7 +967,14 @@ impl SecretsService {
         if let Some(obj) = input.secrets_obj {
             secrets.extend(map_to_kv_strings(obj));
         }
+        let secret_types = input.secret_types.unwrap_or_default();
+        let secret_types_map: std::collections::HashMap<String, String> = secret_types
+            .into_iter()
+            .filter_map(|(k, v)| v.as_str().map(|s| (k, s.to_string())))
+            .collect();
         let remove_secrets = input.remove_secrets.unwrap_or_default();
+        let link_secret_names = input.link_secret_names.unwrap_or_default();
+        let unlink_secret_names = input.unlink_secret_names.unwrap_or_default();
 
         let result = svc_update(
             &self.pool,
@@ -758,13 +987,16 @@ impl SecretsService {
                 meta_entries: &meta,
                 remove_meta: &remove_meta,
                 secret_entries: &secrets,
+                secret_types: &secret_types_map,
                 remove_secrets: &remove_secrets,
+                link_secret_names: &link_secret_names,
+                unlink_secret_names: &unlink_secret_names,
                 user_id: Some(user_id),
             },
             &user_key,
         )
         .await
-        .map_err(|e| mcp_err_internal_logged("secrets_update", Some(user_id), e))?;
+        .map_err(|e| mcp_err_from_anyhow("secrets_update", Some(user_id), e))?;
 
         tracing::info!(
             tool = "secrets_update",
@@ -1000,7 +1232,7 @@ impl SecretsService {
             Some(&user_key),
         )
         .await
-        .map_err(|e| mcp_err_internal_logged("secrets_export", Some(user_id), e))?;
+        .map_err(|e| mcp_err_from_anyhow("secrets_export", Some(user_id), e))?;
 
         let fmt = format.parse::<ExportFormat>().map_err(|e| {
             tracing::warn!(
@@ -1016,7 +1248,7 @@ impl SecretsService {
         })?;
         let serialized = fmt
             .serialize(&data)
-            .map_err(|e| mcp_err_internal_logged("secrets_export", Some(user_id), e))?;
+            .map_err(|e| mcp_err_from_anyhow("secrets_export", Some(user_id), e))?;
 
         tracing::info!(
             tool = "secrets_export",
@@ -1067,7 +1299,7 @@ impl SecretsService {
             Some(user_id),
         )
         .await
-        .map_err(|e| mcp_err_internal_logged("secrets_env_map", Some(user_id), e))?;
+        .map_err(|e| mcp_err_from_anyhow("secrets_env_map", Some(user_id), e))?;
 
         let entry_count = env_map.len();
         tracing::info!(
@@ -1167,3 +1399,202 @@ impl ServerHandler for SecretsService {
         info
     }
 }
+
+#[cfg(test)]
+mod deser_tests {
+    use super::deser;
+    use serde::Deserialize;
+    use serde_json::json;
+
+    #[derive(Deserialize)]
+    struct TestU32 {
+        #[serde(deserialize_with = "deser::option_u32_from_string")]
+        val: Option<u32>,
+    }
+
+    #[derive(Deserialize)]
+    struct TestI64 {
+        #[serde(deserialize_with = "deser::option_i64_from_string")]
+        val: Option<i64>,
+    }
+
+    #[derive(Deserialize)]
+    struct TestBool {
+        #[serde(deserialize_with = "deser::option_bool_from_string")]
+        val: Option<bool>,
+    }
+
+    #[derive(Debug, Deserialize)]
+    struct TestVec {
+        #[serde(deserialize_with = "deser::option_vec_string_from_string")]
+        val: Option<Vec<String>>,
+    }
+
+    #[derive(Debug, Deserialize)]
+    struct TestMap {
+        #[serde(deserialize_with = "deser::option_map_from_string")]
+        val: Option<serde_json::Map<String, serde_json::Value>>,
+    }
+
+    // option_u32_from_string
+    #[test]
+    fn u32_native_number() {
+        let v: TestU32 = serde_json::from_value(json!({"val": 42})).unwrap();
+        assert_eq!(v.val, Some(42));
+    }
+
+    #[test]
+    fn u32_string_number() {
+        let v: TestU32 = serde_json::from_value(json!({"val": "42"})).unwrap();
+        assert_eq!(v.val, Some(42));
+    }
+
+    #[test]
+    fn u32_empty_string() {
+        let v: TestU32 = serde_json::from_value(json!({"val": ""})).unwrap();
+        assert_eq!(v.val, None);
+    }
+
+    #[test]
+    fn u32_none() {
+        let v: TestU32 = serde_json::from_value(json!({"val": null})).unwrap();
+        assert_eq!(v.val, None);
+    }
+
+    // option_i64_from_string
+    #[test]
+    fn i64_native_number() {
+        let v: TestI64 = serde_json::from_value(json!({"val": -100})).unwrap();
+        assert_eq!(v.val, Some(-100));
+    }
+
+    #[test]
+    fn i64_string_number() {
+        let v: TestI64 = serde_json::from_value(json!({"val": "999"})).unwrap();
+        assert_eq!(v.val, Some(999));
+    }
+
+    #[test]
+    fn i64_empty_string() {
+        let v: TestI64 = serde_json::from_value(json!({"val": ""})).unwrap();
+        assert_eq!(v.val, None);
+    }
+
+    #[test]
+    fn i64_none() {
+        let v: TestI64 = serde_json::from_value(json!({"val": null})).unwrap();
+        assert_eq!(v.val, None);
+    }
+
+    // option_bool_from_string
+    #[test]
+    fn bool_native_true() {
+        let v: TestBool = serde_json::from_value(json!({"val": true})).unwrap();
+        assert_eq!(v.val, Some(true));
+    }
+
+    #[test]
+    fn bool_native_false() {
+        let v: TestBool = serde_json::from_value(json!({"val": false})).unwrap();
+        assert_eq!(v.val, Some(false));
+    }
+
+    #[test]
+    fn bool_string_true() {
+        let v: TestBool = serde_json::from_value(json!({"val": "true"})).unwrap();
+        assert_eq!(v.val, Some(true));
+    }
+
+    #[test]
+    fn bool_string_false() {
+        let v: TestBool = serde_json::from_value(json!({"val": "false"})).unwrap();
+        assert_eq!(v.val, Some(false));
+    }
+
+    #[test]
+    fn bool_empty_string() {
+        let v: TestBool = serde_json::from_value(json!({"val": ""})).unwrap();
+        assert_eq!(v.val, None);
+    }
+
+    #[test]
+    fn bool_none() {
+        let v: TestBool = serde_json::from_value(json!({"val": null})).unwrap();
+        assert_eq!(v.val, None);
+    }
+
+    // option_vec_string_from_string
+    #[test]
+    fn vec_native_array() {
+        let v: TestVec = serde_json::from_value(json!({"val": ["a", "b"]})).unwrap();
+        assert_eq!(v.val, Some(vec!["a".to_string(), "b".to_string()]));
+    }
+
+    #[test]
+    fn vec_json_string_array() {
+        let v: TestVec = serde_json::from_value(json!({"val": "[\"x\",\"y\"]"})).unwrap();
+        assert_eq!(v.val, Some(vec!["x".to_string(), "y".to_string()]));
+    }
+
+    #[test]
+    fn vec_empty_string() {
+        let v: TestVec = serde_json::from_value(json!({"val": ""})).unwrap();
+        assert_eq!(v.val, None);
+    }
+
+    #[test]
+    fn vec_none() {
+        let v: TestVec = serde_json::from_value(json!({"val": null})).unwrap();
+        assert_eq!(v.val, None);
+    }
+
+    #[test]
+    fn vec_invalid_string_errors() {
+        let err = serde_json::from_value::<TestVec>(json!({"val": "not-json"}))
+            .expect_err("should fail on invalid JSON");
+        let msg = err.to_string();
+        assert!(msg.contains("invalid string value for array field"));
+        assert!(msg.contains("expected a JSON array"));
+    }
+
+    // option_map_from_string
+    #[test]
+    fn map_native_object() {
+        let v: TestMap = serde_json::from_value(json!({"val": {"key": "value"}})).unwrap();
+        assert!(v.val.is_some());
+        let m = v.val.unwrap();
+        assert_eq!(
+            m.get("key"),
+            Some(&serde_json::Value::String("value".to_string()))
+        );
+    }
+
+    #[test]
+    fn map_json_string_object() {
+        let v: TestMap = serde_json::from_value(json!({"val": "{\"a\":1}"})).unwrap();
+        assert!(v.val.is_some());
+        let m = v.val.unwrap();
+        assert_eq!(m.get("a"), Some(&serde_json::Value::Number(1.into())));
+    }
+
+    #[test]
+    fn map_empty_string() {
+        let v: TestMap = serde_json::from_value(json!({"val": ""})).unwrap();
+        assert_eq!(v.val, None);
+    }
+
+    #[test]
+    fn map_none() {
+        let v: TestMap = serde_json::from_value(json!({"val": null})).unwrap();
+        assert_eq!(v.val, None);
+    }
+
+    #[test]
+    fn map_invalid_string_errors() {
+        let err = serde_json::from_value::<TestMap>(json!({"val": "not-json"}))
+            .expect_err("should fail on invalid JSON");
+        let msg = err.to_string();
+        assert!(msg.contains("invalid string value for object field"));
+        assert!(msg.contains("expected a JSON object"));
+    }
+}

crates/secrets-mcp/src/web.rs

@@ -17,11 +17,12 @@ use uuid::Uuid;
 
 use secrets_core::audit::log_login;
 use secrets_core::crypto::hex;
+use secrets_core::error::AppError;
 use secrets_core::service::{
     api_key::{ensure_api_key, regenerate_api_key},
     audit_log::list_for_user,
     delete::delete_by_id,
-    search::{SearchParams, count_entries, fetch_secret_schemas, list_entries},
+    search::{SearchParams, fetch_secret_schemas, ilike_pattern, list_entries},
     update::{UpdateEntryFieldsByIdParams, update_fields_by_id},
     user::{
         OAuthProfile, bind_oauth_account, find_or_create_user, get_user_by_id,
@@ -88,15 +89,16 @@ struct EntriesPageTemplate {
     user_name: String,
     user_email: String,
     entries: Vec<EntryListItemView>,
-    total_count: i64,
+    folder_tabs: Vec<FolderTabView>,
-    shown_count: usize,
+    type_options: Vec<String>,
-    limit: u32,
+    secret_type_options_json: String,
     filter_folder: String,
+    filter_name: String,
     filter_type: String,
     version: &'static str,
 }
 
-/// Non-sensitive fields only (no `secrets` / ciphertext).
+/// Non-sensitive entry fields; `secrets` lists field names/types only (no ciphertext).
 struct EntryListItemView {
     id: String,
     folder: String,
@@ -104,24 +106,37 @@ struct EntryListItemView {
     name: String,
     notes: String,
     tags: String,
-    metadata: String,
+    /// Compact JSON for `data-entry-metadata` (dialog editor).
+    metadata_json: String,
+    /// Secret field summaries for table + dialog chips.
     secrets: Vec<SecretSummaryView>,
-    /// RFC3339 UTC for `<time datetime>`; localized in entries.html.
+    /// JSON array of `{ id, name, secret_type }` for dialog secret chips.
+    secrets_json: String,
+    /// RFC3339 UTC; shown in edit dialog.
     updated_at_iso: String,
 }
 
+#[derive(Serialize)]
 struct SecretSummaryView {
     id: String,
     name: String,
     secret_type: String,
 }
 
+struct FolderTabView {
+    name: String,
+    count: i64,
+    href: String,
+    active: bool,
+}
+
 /// Cap for HTML list (avoids loading unbounded rows into memory).
 const ENTRIES_PAGE_LIMIT: u32 = 5_000;
 
 #[derive(Deserialize)]
 struct EntriesQuery {
     folder: Option<String>,
+    name: Option<String>,
     /// URL query key is `type` (maps to DB column `entries.type`).
     #[serde(rename = "type")]
     entry_type: Option<String>,
@@ -183,6 +198,7 @@ pub fn web_router() -> Router<AppState> {
         .route("/robots.txt", get(robots_txt))
         .route("/llms.txt", get(llms_txt))
         .route("/ai.txt", get(ai_txt))
+        .route("/static/i18n.js", get(i18n_js))
         .route("/favicon.svg", get(favicon_svg))
         .route(
             "/favicon.ico",
@@ -218,6 +234,8 @@ pub fn web_router() -> Router<AppState> {
             "/api/entries/{entry_id}/secrets/{secret_id}",
             axum::routing::delete(api_entry_secret_unlink),
         )
+        .route("/api/secrets/{secret_id}", patch(api_secret_patch))
+        .route("/api/secrets/check-name", get(api_secret_check_name))
 }
 
 fn text_asset_response(content: &'static str, content_type: &'static str) -> Response {
@@ -247,6 +265,13 @@ async fn ai_txt() -> Response {
     llms_txt().await
 }
 
+async fn i18n_js() -> Response {
+    text_asset_response(
+        include_str!("../templates/i18n.js"),
+        "application/javascript; charset=utf-8",
+    )
+}
+
 async fn favicon_svg() -> Response {
     Response::builder()
         .status(StatusCode::OK)
@@ -565,11 +590,17 @@ async fn entries_page(
         .map(|s| s.trim())
         .filter(|s| !s.is_empty())
         .map(|s| s.to_string());
-
+    let name_filter = q
+        .name
+        .as_ref()
+        .map(|s| s.trim())
+        .filter(|s| !s.is_empty())
+        .map(|s| s.to_string());
     let params = SearchParams {
         folder: folder_filter.as_deref(),
         entry_type: type_filter.as_deref(),
         name: None,
+        name_query: name_filter.as_deref(),
         tags: &[],
         query: None,
         sort: "updated",
@@ -578,16 +609,10 @@ async fn entries_page(
         user_id: Some(user_id),
     };
 
-    let total_count = count_entries(&state.pool, &params).await.map_err(|e| {
-        tracing::error!(error = %e, "failed to count entries for web");
-        StatusCode::INTERNAL_SERVER_ERROR
-    })?;
-
     let rows = list_entries(&state.pool, params).await.map_err(|e| {
         tracing::error!(error = %e, "failed to load entries list for web");
         StatusCode::INTERNAL_SERVER_ERROR
     })?;
-    let shown_count = rows.len();
     let entry_ids: Vec<Uuid> = rows.iter().map(|e| e.id).collect();
     let secret_schemas = fetch_secret_schemas(&state.pool, &entry_ids)
         .await
@@ -596,18 +621,112 @@ async fn entries_page(
             StatusCode::INTERNAL_SERVER_ERROR
         })?;
 
+    #[derive(sqlx::FromRow)]
+    struct FolderCountRow {
+        folder: String,
+        count: i64,
+    }
+
+    let mut folder_sql =
+        "SELECT folder, COUNT(*)::bigint AS count FROM entries WHERE user_id = $1".to_string();
+    let mut bind_idx = 2;
+    if type_filter.is_some() {
+        folder_sql.push_str(&format!(" AND type = ${bind_idx}"));
+        bind_idx += 1;
+    }
+    if name_filter.is_some() {
+        folder_sql.push_str(&format!(" AND name ILIKE ${bind_idx} ESCAPE '\\'"));
+        bind_idx += 1;
+    }
+    let _ = bind_idx;
+    folder_sql.push_str(" GROUP BY folder ORDER BY folder");
+
+    let mut folder_query = sqlx::query_as::<_, FolderCountRow>(&folder_sql).bind(user_id);
+    if let Some(t) = type_filter.as_deref() {
+        folder_query = folder_query.bind(t);
+    }
+    if let Some(n) = name_filter.as_deref() {
+        folder_query = folder_query.bind(ilike_pattern(n));
+    }
+    let folder_rows: Vec<FolderCountRow> =
+        folder_query.fetch_all(&state.pool).await.map_err(|e| {
+            tracing::error!(error = %e, "failed to load folder tabs for web");
+            StatusCode::INTERNAL_SERVER_ERROR
+        })?;
+
+    #[derive(sqlx::FromRow)]
+    struct TypeOptionRow {
+        #[sqlx(rename = "type")]
+        entry_type: String,
+    }
+    let mut type_options: Vec<String> = sqlx::query_as::<_, TypeOptionRow>(
+        "SELECT DISTINCT type FROM entries WHERE user_id = $1 ORDER BY type",
+    )
+    .bind(user_id)
+    .fetch_all(&state.pool)
+    .await
+    .map_err(|e| {
+        tracing::error!(error = %e, "failed to load type options for web");
+        StatusCode::INTERNAL_SERVER_ERROR
+    })?
+    .into_iter()
+    .map(|r| r.entry_type)
+    .filter(|t| !t.is_empty())
+    .collect();
+    if let Some(current) = type_filter.as_ref()
+        && !current.is_empty()
+        && !type_options.iter().any(|t| t == current)
+    {
+        type_options.push(current.clone());
+        type_options.sort_unstable();
+    }
+
+    fn entries_href(folder: Option<&str>, entry_type: Option<&str>, name: Option<&str>) -> String {
+        let mut pairs: Vec<String> = Vec::new();
+        if let Some(f) = folder
+            && !f.is_empty()
+        {
+            pairs.push(format!("folder={}", urlencoding::encode(f)));
+        }
+        if let Some(t) = entry_type
+            && !t.is_empty()
+        {
+            pairs.push(format!("type={}", urlencoding::encode(t)));
+        }
+        if let Some(n) = name
+            && !n.is_empty()
+        {
+            pairs.push(format!("name={}", urlencoding::encode(n)));
+        }
+        if pairs.is_empty() {
+            "/entries".to_string()
+        } else {
+            format!("/entries?{}", pairs.join("&"))
+        }
+    }
+
+    let all_count: i64 = folder_rows.iter().map(|r| r.count).sum();
+    let mut folder_tabs: Vec<FolderTabView> = Vec::with_capacity(folder_rows.len() + 1);
+    folder_tabs.push(FolderTabView {
+        name: "全部".to_string(),
+        count: all_count,
+        href: entries_href(None, type_filter.as_deref(), name_filter.as_deref()),
+        active: folder_filter.is_none(),
+    });
+    for r in folder_rows {
+        let name = r.folder;
+        folder_tabs.push(FolderTabView {
+            href: entries_href(Some(&name), type_filter.as_deref(), name_filter.as_deref()),
+            active: folder_filter.as_deref() == Some(name.as_str()),
+            name,
+            count: r.count,
+        });
+    }
+
     let entries = rows
         .into_iter()
-        .map(|e| EntryListItemView {
+        .map(|e| {
-            id: e.id.to_string(),
+            let secrets: Vec<SecretSummaryView> = secret_schemas
-            folder: e.folder,
-            entry_type: e.entry_type,
-            name: e.name,
-            notes: e.notes,
-            tags: e.tags.join(", "),
-            metadata: serde_json::to_string_pretty(&e.metadata)
-                .unwrap_or_else(|_| "{}".to_string()),
-            secrets: secret_schemas
                 .get(&e.id)
                 .map(|fields| {
                     fields
@@ -619,8 +738,22 @@ async fn entries_page(
                         })
                         .collect()
                 })
-                .unwrap_or_default(),
+                .unwrap_or_default();
+            let secrets_json = serde_json::to_string(&secrets).unwrap_or_else(|_| "[]".to_string());
+            let metadata_json =
+                serde_json::to_string(&e.metadata).unwrap_or_else(|_| "{}".to_string());
+            EntryListItemView {
+                id: e.id.to_string(),
+                folder: e.folder,
+                entry_type: e.entry_type,
+                name: e.name,
+                notes: e.notes,
+                tags: e.tags.join(", "),
+                metadata_json,
+                secrets,
+                secrets_json,
                 updated_at_iso: e.updated_at.to_rfc3339_opts(SecondsFormat::Secs, true),
+            }
         })
         .collect();
 
@@ -628,10 +761,17 @@ async fn entries_page(
         user_name: user.name.clone(),
         user_email: user.email.clone().unwrap_or_default(),
         entries,
-        total_count,
+        folder_tabs,
-        shown_count,
+        type_options,
-        limit: ENTRIES_PAGE_LIMIT,
+        secret_type_options_json: serde_json::to_string(
+            &secrets_core::taxonomy::SECRET_TYPE_OPTIONS
+                .iter()
+                .map(|s| s.to_string())
+                .collect::<Vec<_>>(),
+        )
+        .unwrap_or_default(),
         filter_folder: folder_filter.unwrap_or_default(),
+        filter_name: name_filter.unwrap_or_default(),
         filter_type: type_filter.unwrap_or_default(),
         version: env!("CARGO_PKG_VERSION"),
     };
@@ -927,24 +1067,53 @@ struct EntryPatchBody {
 
 type EntryApiError = (StatusCode, Json<serde_json::Value>);
 
-fn map_entry_mutation_err(e: anyhow::Error) -> EntryApiError {
+#[derive(Clone, Copy)]
-    let msg = e.to_string();
+enum UiLang {
-    if msg.contains("Entry not found") {
+    ZhCn,
-        return (
+    ZhTw,
-            StatusCode::NOT_FOUND,
+    En,
-            Json(json!({ "error": "条目不存在或无权访问" })),
-        );
 }
+
+fn request_ui_lang(headers: &HeaderMap) -> UiLang {
+    let Some(raw) = headers
+        .get(header::ACCEPT_LANGUAGE)
+        .and_then(|v| v.to_str().ok())
+    else {
+        return UiLang::ZhCn;
+    };
+    let lower = raw.to_ascii_lowercase();
+    if lower.contains("zh-tw") || lower.contains("zh-hk") || lower.contains("zh-hant") {
+        UiLang::ZhTw
+    } else if lower.contains("zh") {
+        UiLang::ZhCn
+    } else if lower.contains("en") {
+        UiLang::En
+    } else {
+        UiLang::ZhCn
+    }
+}
+
+fn tr(lang: UiLang, zh_cn: &'static str, zh_tw: &'static str, en: &'static str) -> &'static str {
+    match lang {
+        UiLang::ZhCn => zh_cn,
+        UiLang::ZhTw => zh_tw,
+        UiLang::En => en,
+    }
+}
+
+fn map_entry_mutation_err(e: anyhow::Error, lang: UiLang) -> EntryApiError {
+    if let Some(app_err) = e.downcast_ref::<AppError>() {
+        return map_app_error(app_err, lang);
+    }
+
+    // Fallback for legacy string-based errors and raw sqlx errors
+    let msg = e.to_string();
     if msg.contains("already exists") {
         return (
             StatusCode::CONFLICT,
-            Json(json!({ "error": "该账号下已存在相同 folder + name 的条目" })),
+            Json(
-        );
+                json!({ "error": tr(lang, "该账号下已存在相同 folder + name 的条目", "此帳號下已存在相同 folder + name 的條目", "An entry with the same folder + name already exists for this account") }),
-    }
+            ),
-    if msg.contains("Concurrent modification") {
-        return (
-            StatusCode::CONFLICT,
-            Json(json!({ "error": "条目已被修改，请刷新后重试" })),
         );
     }
     if msg.contains("must be at most") {
@@ -953,19 +1122,63 @@ fn map_entry_mutation_err(e: anyhow::Error) -> EntryApiError {
     tracing::error!(error = %e, "entry mutation failed");
     (
         StatusCode::INTERNAL_SERVER_ERROR,
-        Json(json!({ "error": "操作失败，请稍后重试" })),
+        Json(
+            json!({ "error": tr(lang, "操作失败，请稍后重试", "操作失敗，請稍後重試", "Operation failed, please try again later") }),
+        ),
     )
 }
 
+fn map_app_error(err: &AppError, lang: UiLang) -> EntryApiError {
+    match err {
+        AppError::ConflictEntryName { .. } | AppError::ConflictSecretName { .. } => (
+            StatusCode::CONFLICT,
+            Json(json!({ "error": err.to_string() })),
+        ),
+        AppError::NotFoundEntry => (
+            StatusCode::NOT_FOUND,
+            Json(
+                json!({ "error": tr(lang, "条目不存在或无权访问", "條目不存在或無權存取", "Entry not found or no access") }),
+            ),
+        ),
+        AppError::Validation { message } => {
+            (StatusCode::BAD_REQUEST, Json(json!({ "error": message })))
+        }
+        AppError::ConcurrentModification => (
+            StatusCode::CONFLICT,
+            Json(
+                json!({ "error": tr(lang, "条目已被修改，请刷新后重试", "條目已被修改，請重新整理後重試", "Entry was modified, please refresh and try again") }),
+            ),
+        ),
+        AppError::DecryptionFailed => (
+            StatusCode::BAD_REQUEST,
+            Json(
+                json!({ "error": tr(lang, "解密失败，请检查密码短语", "解密失敗，請檢查密碼短語", "Decryption failed — please check your passphrase") }),
+            ),
+        ),
+        AppError::Internal(_) => {
+            tracing::error!(error = %err, "internal error in entry mutation");
+            (
+                StatusCode::INTERNAL_SERVER_ERROR,
+                Json(
+                    json!({ "error": tr(lang, "操作失败，请稍后重试", "操作失敗，請稍後重試", "Operation failed, please try again later") }),
+                ),
+            )
+        }
+    }
+}
+
 async fn api_entry_patch(
     State(state): State<AppState>,
     session: Session,
+    headers: HeaderMap,
     Path(entry_id): Path<Uuid>,
     Json(body): Json<EntryPatchBody>,
 ) -> Result<Json<serde_json::Value>, EntryApiError> {
-    let user_id = current_user_id(&session)
+    let lang = request_ui_lang(&headers);
-        .await
+    let user_id = current_user_id(&session).await.ok_or((
-        .ok_or((StatusCode::UNAUTHORIZED, Json(json!({ "error": "未登录" }))))?;
+        StatusCode::UNAUTHORIZED,
+        Json(json!({ "error": tr(lang, "未登录", "尚未登入", "Not logged in") })),
+    ))?;
 
     let folder = body.folder.trim();
     let entry_type = body.entry_type.trim();
@@ -975,7 +1188,9 @@ async fn api_entry_patch(
     if name.is_empty() {
         return Err((
             StatusCode::BAD_REQUEST,
-            Json(json!({ "error": "name 不能为空" })),
+            Json(
+                json!({ "error": tr(lang, "name 不能为空", "name 不能為空", "name cannot be empty") }),
+            ),
         ));
     }
 
@@ -989,7 +1204,9 @@ async fn api_entry_patch(
     if !body.metadata.is_object() {
         return Err((
             StatusCode::BAD_REQUEST,
-            Json(json!({ "error": "metadata 必须是 JSON 对象" })),
+            Json(
+                json!({ "error": tr(lang, "metadata 必须是 JSON 对象", "metadata 必須是 JSON 物件", "metadata must be a JSON object") }),
+            ),
         ));
     }
 
@@ -1007,7 +1224,7 @@ async fn api_entry_patch(
         },
     )
     .await
-    .map_err(map_entry_mutation_err)?;
+    .map_err(|e| map_entry_mutation_err(e, lang))?;
 
     Ok(Json(json!({ "ok": true })))
 }
@@ -1015,25 +1232,291 @@ async fn api_entry_patch(
 async fn api_entry_delete(
     State(state): State<AppState>,
     session: Session,
+    headers: HeaderMap,
     Path(entry_id): Path<Uuid>,
 ) -> Result<Json<serde_json::Value>, EntryApiError> {
-    let user_id = current_user_id(&session)
+    let lang = request_ui_lang(&headers);
-        .await
+    let user_id = current_user_id(&session).await.ok_or((
-        .ok_or((StatusCode::UNAUTHORIZED, Json(json!({ "error": "未登录" }))))?;
+        StatusCode::UNAUTHORIZED,
+        Json(json!({ "error": tr(lang, "未登录", "尚未登入", "Not logged in") })),
+    ))?;
 
-    let result = delete_by_id(&state.pool, entry_id, user_id)
+    delete_by_id(&state.pool, entry_id, user_id)
         .await
-        .map_err(map_entry_mutation_err)?;
+        .map_err(|e| map_entry_mutation_err(e, lang))?;
 
     Ok(Json(json!({
         "ok": true,
-        "migrated": result.migrated,
     })))
 }
 
+#[derive(Deserialize)]
+struct SecretCheckNameQuery {
+    name: String,
+    exclude_secret_id: Option<Uuid>,
+}
+
+#[derive(Serialize)]
+struct SecretCheckNameResponse {
+    ok: bool,
+    available: bool,
+    #[serde(skip_serializing_if = "Option::is_none")]
+    error: Option<String>,
+}
+
+async fn api_secret_check_name(
+    State(state): State<AppState>,
+    session: Session,
+    headers: HeaderMap,
+    Query(params): Query<SecretCheckNameQuery>,
+) -> Result<Json<SecretCheckNameResponse>, EntryApiError> {
+    let lang = request_ui_lang(&headers);
+    let user_id = current_user_id(&session).await.ok_or((
+        StatusCode::UNAUTHORIZED,
+        Json(json!({ "error": tr(lang, "未登录", "尚未登入", "Not logged in") })),
+    ))?;
+
+    let name = params.name.trim();
+    if name.is_empty() {
+        return Err((
+            StatusCode::BAD_REQUEST,
+            Json(
+                json!({ "error": tr(lang, "secret name 不能为空", "secret name 不能為空", "secret name cannot be empty") }),
+            ),
+        ));
+    }
+    if name.chars().count() > 256 {
+        return Err((
+            StatusCode::BAD_REQUEST,
+            Json(
+                json!({ "error": tr(lang, "secret name 长度不能超过 256 个字符", "secret name 長度不能超過 256 個字元", "secret name must be at most 256 characters") }),
+            ),
+        ));
+    }
+
+    let count: i64 = if let Some(exclude_id) = params.exclude_secret_id {
+        sqlx::query_scalar::<_, i64>(
+            "SELECT COUNT(*) FROM secrets WHERE user_id = $1 AND name = $2 AND id != $3",
+        )
+        .bind(user_id)
+        .bind(name)
+        .bind(exclude_id)
+        .fetch_one(&state.pool)
+        .await
+    } else {
+        sqlx::query_scalar::<_, i64>(
+            "SELECT COUNT(*) FROM secrets WHERE user_id = $1 AND name = $2",
+        )
+        .bind(user_id)
+        .bind(name)
+        .fetch_one(&state.pool)
+        .await
+    }.map_err(|e| {
+        tracing::error!(error = %e, "failed to check secret name availability");
+        (
+            StatusCode::INTERNAL_SERVER_ERROR,
+            Json(
+                json!({ "error": tr(lang, "操作失败，请稍后重试", "操作失敗，請稍後重試", "Operation failed, please try again later") }),
+            ),
+        )
+    })?;
+
+    let available = count == 0;
+    let error = if available {
+        None
+    } else {
+        Some(
+            tr(
+                lang,
+                "该用户下已存在相同 name 的密文",
+                "該用戶下已存在相同 name 的密文",
+                "A secret with the same name already exists for this user",
+            )
+            .to_string(),
+        )
+    };
+
+    Ok(Json(SecretCheckNameResponse {
+        ok: true,
+        available,
+        error,
+    }))
+}
+
+#[derive(Deserialize)]
+struct SecretPatchBody {
+    name: Option<String>,
+    #[serde(rename = "type")]
+    secret_type: Option<String>,
+}
+
+async fn api_secret_patch(
+    State(state): State<AppState>,
+    session: Session,
+    headers: HeaderMap,
+    Path(secret_id): Path<Uuid>,
+    Json(body): Json<SecretPatchBody>,
+) -> Result<Json<serde_json::Value>, EntryApiError> {
+    #[derive(Serialize, sqlx::FromRow)]
+    struct LinkedEntryAuditRow {
+        folder: String,
+        #[sqlx(rename = "type")]
+        entry_type: String,
+        name: String,
+    }
+
+    let lang = request_ui_lang(&headers);
+    let user_id = current_user_id(&session).await.ok_or((
+        StatusCode::UNAUTHORIZED,
+        Json(json!({ "error": tr(lang, "未登录", "尚未登入", "Not logged in") })),
+    ))?;
+
+    let name = body.name.as_ref().map(|s| s.trim());
+    let secret_type = body.secret_type.as_ref().map(|s| s.trim());
+
+    if let Some(n) = name {
+        if n.is_empty() {
+            return Err((
+                StatusCode::BAD_REQUEST,
+                Json(
+                    json!({ "error": tr(lang, "secret name 不能为空", "secret name 不能為空", "secret name cannot be empty") }),
+                ),
+            ));
+        }
+        if n.chars().count() > 256 {
+            return Err((
+                StatusCode::BAD_REQUEST,
+                Json(
+                    json!({ "error": tr(lang, "secret name 长度不能超过 256 个字符", "secret name 長度不能超過 256 個字元", "secret name must be at most 256 characters") }),
+                ),
+            ));
+        }
+    }
+
+    if let Some(t) = secret_type {
+        if t.is_empty() {
+            return Err((
+                StatusCode::BAD_REQUEST,
+                Json(
+                    json!({ "error": tr(lang, "secret type 不能为空", "secret type 不能為空", "secret type cannot be empty") }),
+                ),
+            ));
+        }
+        if t.chars().count() > 64 {
+            return Err((
+                StatusCode::BAD_REQUEST,
+                Json(
+                    json!({ "error": tr(lang, "secret type 长度不能超过 64 个字符", "secret type 長度不能超過 64 個字元", "secret type must be at most 64 characters") }),
+                ),
+            ));
+        }
+    }
+
+    if name.is_none() && secret_type.is_none() {
+        return Err((
+            StatusCode::BAD_REQUEST,
+            Json(
+                json!({ "error": tr(lang, "至少需要提供 name 或 type 之一", "至少需要提供 name 或 type 之一", "At least one of name or type is required") }),
+            ),
+        ));
+    }
+
+    let mut tx = state
+        .pool
+        .begin()
+        .await
+        .map_err(|e| map_entry_mutation_err(e.into(), lang))?;
+
+    let secret_row: Option<(String, String)> =
+        sqlx::query_as("SELECT name, type FROM secrets WHERE id = $1 AND user_id = $2 FOR UPDATE")
+            .bind(secret_id)
+            .bind(user_id)
+            .fetch_optional(&mut *tx)
+            .await
+            .map_err(|e| map_entry_mutation_err(e.into(), lang))?;
+
+    let Some((old_name, old_type)) = secret_row else {
+        let _ = tx.rollback().await;
+        return Err((
+            StatusCode::NOT_FOUND,
+            Json(
+                json!({ "error": tr(lang, "密文不存在或无权访问", "密文不存在或無權存取", "Secret not found or no access") }),
+            ),
+        ));
+    };
+
+    let linked_entries: Vec<LinkedEntryAuditRow> = sqlx::query_as(
+        "SELECT e.folder, e.type, e.name \
+         FROM entry_secrets es \
+         JOIN entries e ON e.id = es.entry_id \
+         WHERE es.secret_id = $1 AND e.user_id = $2 \
+         ORDER BY e.folder, e.type, e.name",
+    )
+    .bind(secret_id)
+    .bind(user_id)
+    .fetch_all(&mut *tx)
+    .await
+    .map_err(|e| map_entry_mutation_err(e.into(), lang))?;
+
+    let new_name = name.unwrap_or(&old_name).to_string();
+    let new_type = secret_type.unwrap_or(&old_type).to_string();
+
+    let result = sqlx::query(
+        "UPDATE secrets SET name = $1, type = $2, version = version + 1, updated_at = NOW() \
+         WHERE id = $3",
+    )
+    .bind(&new_name)
+    .bind(&new_type)
+    .bind(secret_id)
+    .execute(&mut *tx)
+    .await;
+
+    if let Err(e) = result {
+        if let Some(db_err) = e.as_database_error()
+            && db_err.code() == Some("23505".into())
+        {
+            let _ = tx.rollback().await;
+            return Err(map_app_error(
+                &AppError::ConflictSecretName {
+                    secret_name: new_name.clone(),
+                },
+                lang,
+            ));
+        }
+        let _ = tx.rollback().await;
+        return Err(map_entry_mutation_err(e.into(), lang));
+    }
+
+    secrets_core::audit::log_tx(
+        &mut tx,
+        Some(user_id),
+        "rename_secret",
+        "",
+        "",
+        &old_name,
+        json!({
+            "source": "web",
+            "secret_id": secret_id,
+            "old_name": old_name,
+            "new_name": new_name,
+            "old_type": old_type,
+            "new_type": new_type,
+            "linked_entries": linked_entries,
+        }),
+    )
+    .await;
+
+    tx.commit()
+        .await
+        .map_err(|e| map_entry_mutation_err(e.into(), lang))?;
+
+    Ok(Json(json!({ "ok": true })))
+}
+
 async fn api_entry_secret_unlink(
     State(state): State<AppState>,
     session: Session,
+    headers: HeaderMap,
     Path((entry_id, secret_id)): Path<(Uuid, Uuid)>,
 ) -> Result<Json<serde_json::Value>, EntryApiError> {
     #[derive(sqlx::FromRow)]
@@ -1044,15 +1527,17 @@ async fn api_entry_secret_unlink(
         name: String,
     }
 
-    let user_id = current_user_id(&session)
+    let lang = request_ui_lang(&headers);
-        .await
+    let user_id = current_user_id(&session).await.ok_or((
-        .ok_or((StatusCode::UNAUTHORIZED, Json(json!({ "error": "未登录" }))))?;
+        StatusCode::UNAUTHORIZED,
+        Json(json!({ "error": tr(lang, "未登录", "尚未登入", "Not logged in") })),
+    ))?;
 
     let mut tx = state
         .pool
         .begin()
         .await
-        .map_err(|e| map_entry_mutation_err(e.into()))?;
+        .map_err(|e| map_entry_mutation_err(e.into(), lang))?;
 
     let entry_row: Option<EntryAuditRow> =
         sqlx::query_as("SELECT folder, type, name FROM entries WHERE id = $1 AND user_id = $2")
@@ -1060,15 +1545,15 @@ async fn api_entry_secret_unlink(
             .bind(user_id)
             .fetch_optional(&mut *tx)
             .await
-            .map_err(|e| map_entry_mutation_err(e.into()))?;
+            .map_err(|e| map_entry_mutation_err(e.into(), lang))?;
 
     let Some(entry_row) = entry_row else {
-        tx.rollback()
+        let _ = tx.rollback().await;
-            .await
-            .map_err(|e| map_entry_mutation_err(e.into()))?;
         return Err((
             StatusCode::NOT_FOUND,
-            Json(json!({ "error": "条目不存在或无权访问" })),
+            Json(
+                json!({ "error": tr(lang, "条目不存在或无权访问", "條目不存在或無權存取", "Entry not found or no access") }),
+            ),
         ));
     };
 
@@ -1077,16 +1562,14 @@ async fn api_entry_secret_unlink(
         .bind(secret_id)
         .execute(&mut *tx)
         .await
-        .map_err(|e| map_entry_mutation_err(e.into()))?
+        .map_err(|e| map_entry_mutation_err(e.into(), lang))?
         .rows_affected();
 
     if deleted == 0 {
-        tx.rollback()
+        let _ = tx.rollback().await;
-            .await
-            .map_err(|e| map_entry_mutation_err(e.into()))?;
         return Err((
             StatusCode::NOT_FOUND,
-            Json(json!({ "error": "关联不存在" })),
+            Json(json!({ "error": tr(lang, "关联不存在", "關聯不存在", "Relation not found") })),
         ));
     }
 
@@ -1098,7 +1581,7 @@ async fn api_entry_secret_unlink(
     .bind(secret_id)
     .execute(&mut *tx)
     .await
-    .map_err(|e| map_entry_mutation_err(e.into()))?
+    .map_err(|e| map_entry_mutation_err(e.into(), lang))?
     .rows_affected()
         > 0;
 
@@ -1120,7 +1603,7 @@ async fn api_entry_secret_unlink(
 
     tx.commit()
         .await
-        .map_err(|e| map_entry_mutation_err(e.into()))?;
+        .map_err(|e| map_entry_mutation_err(e.into(), lang))?;
 
     Ok(Json(json!({
         "ok": true,
@@ -1173,3 +1656,27 @@ fn format_audit_target(folder: &str, entry_type: &str, name: &str) -> String {
         name.to_string()
     }
 }
+
+#[cfg(test)]
+mod tests {
+    use super::*;
+
+    #[test]
+    fn request_ui_lang_prefers_zh_cn_over_en_fallback() {
+        let mut headers = HeaderMap::new();
+        headers.insert(header::ACCEPT_LANGUAGE, "zh-CN, en;q=0.5".parse().unwrap());
+
+        assert!(matches!(request_ui_lang(&headers), UiLang::ZhCn));
+    }
+
+    #[test]
+    fn request_ui_lang_detects_traditional_chinese_variants() {
+        let mut headers = HeaderMap::new();
+        headers.insert(
+            header::ACCEPT_LANGUAGE,
+            "zh-Hant, en;q=0.5".parse().unwrap(),
+        );
+
+        assert!(matches!(request_ui_lang(&headers), UiLang::ZhTw));
+    }
+}

crates/secrets-mcp/templates/audit.html

@@ -38,6 +38,10 @@
         }
         .topbar-spacer { flex: 1; }
         .nav-user { font-size: 13px; color: var(--text-muted); }
+        .lang-bar { display: flex; gap: 2px; background: var(--surface2); border-radius: 6px; padding: 2px; }
+        .lang-btn { padding: 3px 9px; border: none; background: none; color: var(--text-muted);
+                    font-size: 12px; cursor: pointer; border-radius: 4px; }
+        .lang-btn.active { background: var(--border); color: var(--text); }
         .btn-sign-out {
             padding: 5px 12px; border-radius: 6px; border: 1px solid var(--border);
             background: none; color: var(--text); font-size: 12px; text-decoration: none; cursor: pointer;
@@ -77,11 +81,8 @@
             td::before {
                 display: block; color: var(--text-muted); font-size: 11px;
                 margin-bottom: 4px; text-transform: uppercase;
+                content: attr(data-label);
             }
-            td.col-time::before { content: "Time"; }
-            td.col-action::before { content: "Action"; }
-            td.col-target::before { content: "Target"; }
-            td.col-detail::before { content: "Detail"; }
             .detail { max-width: none; }
         }
     </style>
@@ -91,9 +92,9 @@
 <aside class="sidebar">
     <a href="/dashboard" class="sidebar-logo"><span>secrets</span></a>
     <nav class="sidebar-menu">
-        <a href="/dashboard" class="sidebar-link">MCP</a>
+        <a href="/dashboard" class="sidebar-link" data-i18n="navMcp">MCP</a>
-        <a href="/entries" class="sidebar-link">条目</a>
+        <a href="/entries" class="sidebar-link" data-i18n="navEntries">条目</a>
-        <a href="/audit" class="sidebar-link active">审计</a>
+        <a href="/audit" class="sidebar-link active" data-i18n="navAudit">审计</a>
     </nav>
 </aside>
 
@@ -101,35 +102,40 @@
 <div class="topbar">
     <span class="topbar-spacer"></span>
     <span class="nav-user">{{ user_name }}{% if !user_email.is_empty() %} · {{ user_email }}{% endif %}</span>
+    <div class="lang-bar">
+        <button class="lang-btn" onclick="setLang('zh-CN')">简</button>
+        <button class="lang-btn" onclick="setLang('zh-TW')">繁</button>
+        <button class="lang-btn" onclick="setLang('en')">EN</button>
+    </div>
     <form action="/auth/logout" method="post" style="display:inline">
-        <button type="submit" class="btn-sign-out">退出</button>
+        <button type="submit" class="btn-sign-out" data-i18n="signOut">退出</button>
     </form>
 </div>
 
 <main class="main">
     <section class="card">
-        <div class="card-title">我的审计</div>
+        <div class="card-title" data-i18n="auditTitle">我的审计</div>
-        <div class="card-subtitle">展示最近 100 条与当前用户相关的新审计记录。时间为浏览器本地时区。</div>
+        <div class="card-subtitle" data-i18n="auditSubtitle">展示最近 100 条与当前用户相关的新审计记录。时间为浏览器本地时区。</div>
 
         {% if entries.is_empty() %}
-        <div class="empty">暂无审计记录。</div>
+        <div class="empty" data-i18n="emptyAudit">暂无审计记录。</div>
         {% else %}
         <table>
             <thead>
                 <tr>
-                    <th>时间</th>
+                    <th data-i18n="colTime">时间</th>
-                    <th>动作</th>
+                    <th data-i18n="colAction">动作</th>
-                    <th>目标</th>
+                    <th data-i18n="colTarget">目标</th>
-                    <th>详情</th>
+                    <th data-i18n="colDetail">详情</th>
                 </tr>
             </thead>
             <tbody>
                 {% for entry in entries %}
                 <tr>
-                    <td class="col-time mono"><time class="audit-local-time" datetime="{{ entry.created_at_iso }}">{{ entry.created_at_iso }}</time></td>
+                    <td class="col-time mono" data-label="时间"><time class="audit-local-time" datetime="{{ entry.created_at_iso }}">{{ entry.created_at_iso }}</time></td>
-                    <td class="col-action mono">{{ entry.action }}</td>
+                    <td class="col-action mono" data-label="动作">{{ entry.action }}</td>
-                    <td class="col-target mono">{{ entry.target }}</td>
+                    <td class="col-target mono" data-label="目标">{{ entry.target }}</td>
-                    <td class="col-detail"><pre class="detail">{{ entry.detail }}</pre></td>
+                    <td class="col-detail" data-label="详情"><pre class="detail">{{ entry.detail }}</pre></td>
                 </tr>
                 {% endfor %}
             </tbody>
@@ -139,8 +145,28 @@
 </main>
 </div>
 </div>
+<script src="/static/i18n.js"></script>
 <script>
 (function () {
+  I18N_PAGE = {
+    'zh-CN': { pageTitle: 'Secrets — 审计', auditTitle: '我的审计', auditSubtitle: '展示最近 100 条与当前用户相关的新审计记录。时间为浏览器本地时区。', emptyAudit: '暂无审计记录。', colTime: '时间', colAction: '动作', colTarget: '目标', colDetail: '详情' },
+    'zh-TW': { pageTitle: 'Secrets — 審計', auditTitle: '我的審計', auditSubtitle: '顯示最近 100 筆與目前使用者相關的新審計記錄。時間為瀏覽器本地時區。', emptyAudit: '暫無審計記錄。', colTime: '時間', colAction: '動作', colTarget: '目標', colDetail: '詳情' },
+    en: { pageTitle: 'Secrets — Audit', auditTitle: 'My audit', auditSubtitle: 'Shows the latest 100 audit records related to the current user. Time is in browser local timezone.', emptyAudit: 'No audit records.', colTime: 'Time', colAction: 'Action', colTarget: 'Target', colDetail: 'Detail' }
+  };
+
+  window.applyPageLang = function () {
+    document.querySelectorAll('tbody tr').forEach(function (tr) {
+      var time = tr.querySelector('.col-time');
+      var action = tr.querySelector('.col-action');
+      var target = tr.querySelector('.col-target');
+      var detail = tr.querySelector('.col-detail');
+      if (time) time.setAttribute('data-label', t('mobileLabelTime'));
+      if (action) action.setAttribute('data-label', t('mobileLabelAction'));
+      if (target) target.setAttribute('data-label', t('mobileLabelTarget'));
+      if (detail) detail.setAttribute('data-label', t('mobileLabelDetail'));
+    });
+  };
+
   document.querySelectorAll('time.audit-local-time[datetime]').forEach(function (el) {
     var raw = el.getAttribute('datetime');
     var d = raw ? new Date(raw) : null;
@@ -149,6 +175,7 @@
       el.title = raw + ' (UTC)';
     }
   });
+  applyLang();
 })();
 </script>
 </body>

crates/secrets-mcp/templates/i18n.js

@@ -0,0 +1,76 @@
+var I18N_SHARED = {
+  'zh-CN': {
+    pageTitleBase: 'Secrets',
+    navMcp: 'MCP',
+    navEntries: '条目',
+    navAudit: '审计',
+    signOut: '退出',
+    mobileLabelTime: '时间',
+    mobileLabelAction: '动作',
+    mobileLabelTarget: '目标',
+    mobileLabelDetail: '详情'
+  },
+  'zh-TW': {
+    pageTitleBase: 'Secrets',
+    navMcp: 'MCP',
+    navEntries: '條目',
+    navAudit: '審計',
+    signOut: '登出',
+    mobileLabelTime: '時間',
+    mobileLabelAction: '動作',
+    mobileLabelTarget: '目標',
+    mobileLabelDetail: '詳情'
+  },
+  en: {
+    pageTitleBase: 'Secrets',
+    navMcp: 'MCP',
+    navEntries: 'Entries',
+    navAudit: 'Audit',
+    signOut: 'Sign out',
+    mobileLabelTime: 'Time',
+    mobileLabelAction: 'Action',
+    mobileLabelTarget: 'Target',
+    mobileLabelDetail: 'Detail'
+  }
+};
+
+var currentLang = localStorage.getItem('lang') || 'zh-CN';
+var I18N_PAGE = {};
+
+function t(key) {
+  var dict = I18N_PAGE[currentLang] || I18N_PAGE['en'] || {};
+  var val = dict[key] || (I18N_SHARED[currentLang] && I18N_SHARED[currentLang][key]) || (I18N_SHARED.en && I18N_SHARED.en[key]) || key;
+  return val;
+}
+
+function tf(key, vars) {
+  var tpl = t(key);
+  return Object.keys(vars || {}).reduce(function (acc, k) {
+    return acc.replace(new RegExp('\\{' + k + '\\}', 'g'), String(vars[k]));
+  }, tpl);
+}
+
+function applyLang() {
+  document.documentElement.lang = currentLang;
+  var title = t('pageTitle');
+  if (title) document.title = title;
+  document.querySelectorAll('[data-i18n]').forEach(function (el) {
+    var key = el.getAttribute('data-i18n');
+    el.textContent = t(key);
+  });
+  document.querySelectorAll('[data-i18n-ph]').forEach(function (el) {
+    var key = el.getAttribute('data-i18n-ph');
+    el.placeholder = t(key);
+  });
+  document.querySelectorAll('.lang-btn').forEach(function (btn) {
+    var map = { 'zh-CN': '简', 'zh-TW': '繁', en: 'EN' };
+    btn.classList.toggle('active', btn.textContent === map[currentLang]);
+  });
+  if (typeof applyPageLang === 'function') applyPageLang();
+}
+
+window.setLang = function (lang) {
+  currentLang = lang;
+  localStorage.setItem('lang', lang);
+  applyLang();
+};

migrations/001_nn_schema.sql

@@ -1,126 +0,0 @@
--- Entry-Secret N:N migration (manual SQL)
--- Safe to re-run: uses IF EXISTS/IF NOT EXISTS guards.
-
-BEGIN;
-
--- 1) secrets: add new columns
-ALTER TABLE secrets
-    ADD COLUMN IF NOT EXISTS user_id UUID REFERENCES users(id) ON DELETE SET NULL;
-ALTER TABLE secrets
-    ADD COLUMN IF NOT EXISTS type VARCHAR(64) NOT NULL DEFAULT 'text';
-
--- 2) rename field_name -> name (idempotent)
-DO $$ BEGIN
-    IF EXISTS (
-        SELECT 1
-        FROM information_schema.columns
-        WHERE table_name = 'secrets' AND column_name = 'field_name'
-    ) THEN
-        ALTER TABLE secrets RENAME COLUMN field_name TO name;
-    END IF;
-END $$;
-
--- 3) create join table
-CREATE TABLE IF NOT EXISTS entry_secrets (
-    entry_id UUID NOT NULL REFERENCES entries(id) ON DELETE CASCADE,
-    secret_id UUID NOT NULL REFERENCES secrets(id) ON DELETE CASCADE,
-    sort_order INT NOT NULL DEFAULT 0,
-    created_at TIMESTAMPTZ NOT NULL DEFAULT NOW(),
-    PRIMARY KEY (entry_id, secret_id)
-);
-CREATE INDEX IF NOT EXISTS idx_entry_secrets_secret_id ON entry_secrets(secret_id);
-
--- 4) backfill user_id and relationship from old secrets.entry_id
-DO $$ BEGIN
-    IF EXISTS (
-        SELECT 1
-        FROM information_schema.columns
-        WHERE table_name = 'secrets' AND column_name = 'entry_id'
-    ) THEN
-        UPDATE secrets s
-        SET user_id = e.user_id
-        FROM entries e
-        WHERE s.entry_id = e.id AND s.user_id IS NULL;
-
-        INSERT INTO entry_secrets(entry_id, secret_id, sort_order)
-        SELECT entry_id, id, 0
-        FROM secrets
-        WHERE entry_id IS NOT NULL
-        ON CONFLICT DO NOTHING;
-    END IF;
-END $$;
-
--- 5) backfill secret types
-UPDATE secrets SET type = 'pem' WHERE name IN ('ssh_key');
-UPDATE secrets SET type = 'password' WHERE name IN ('password');
-UPDATE secrets SET type = 'phone' WHERE name LIKE 'phone%';
-UPDATE secrets SET type = 'url' WHERE name IN ('webhook_url', 'address');
-UPDATE secrets
-SET type = 'token'
-WHERE name IN (
-    'access_key_id',
-    'access_key_secret',
-    'global_api_key',
-    'api_key',
-    'secret_key',
-    'personal_access_token',
-    'runner_token',
-    'GOOGLE_CLIENT_ID',
-    'GOOGLE_CLIENT_SECRET'
-);
-
--- 6) drop old entry_id path
-ALTER TABLE secrets DROP CONSTRAINT IF EXISTS secrets_entry_id_fkey;
-DROP INDEX IF EXISTS idx_secrets_entry_id;
-ALTER TABLE secrets DROP CONSTRAINT IF EXISTS secrets_entry_id_field_name_key;
-ALTER TABLE secrets DROP CONSTRAINT IF EXISTS secrets_entry_id_name_key;
-ALTER TABLE secrets DROP COLUMN IF EXISTS entry_id;
-
--- 7) add indexes for new access paths
-CREATE INDEX IF NOT EXISTS idx_secrets_user_id
-    ON secrets(user_id) WHERE user_id IS NOT NULL;
-DO $$
-DECLARE
-    duplicate_samples TEXT;
-BEGIN
-    SELECT string_agg(
-               format('user_id=%s, name=%s, count=%s', t.user_id, t.name, t.cnt),
-               E'\n'
-           )
-    INTO duplicate_samples
-    FROM (
-        SELECT user_id::TEXT AS user_id, name, COUNT(*) AS cnt
-        FROM secrets
-        WHERE user_id IS NOT NULL
-        GROUP BY user_id, name
-        HAVING COUNT(*) > 1
-        ORDER BY cnt DESC, user_id, name
-        LIMIT 20
-    ) t;
-
-    IF duplicate_samples IS NOT NULL THEN
-        RAISE EXCEPTION
-            'Cannot enforce unique constraint on secrets(user_id, name). Duplicates found:%',
-            E'\n' || duplicate_samples
-            USING HINT = 'Please deduplicate conflicting rows, then rerun migration.';
-    END IF;
-END $$;
-CREATE UNIQUE INDEX IF NOT EXISTS idx_secrets_unique_user_name
-    ON secrets(user_id, name) WHERE user_id IS NOT NULL;
-CREATE INDEX IF NOT EXISTS idx_secrets_name ON secrets(name);
-CREATE INDEX IF NOT EXISTS idx_secrets_type ON secrets(type);
-
--- 8) secrets_history: rename and remove entry-scoped columns
-DO $$ BEGIN
-    IF EXISTS (
-        SELECT 1
-        FROM information_schema.columns
-        WHERE table_name = 'secrets_history' AND column_name = 'field_name'
-    ) THEN
-        ALTER TABLE secrets_history RENAME COLUMN field_name TO name;
-    END IF;
-END $$;
-ALTER TABLE secrets_history DROP COLUMN IF EXISTS entry_id;
-ALTER TABLE secrets_history DROP COLUMN IF EXISTS entry_version;
-
-COMMIT;

migrations/002_data_cleanup.sql

@@ -1,67 +0,0 @@
--- Metadata cleanup migration (manual SQL)
--- Keep tags/type as dedicated columns; remove duplicated metadata keys.
-
-BEGIN;
-
--- 1) Promote metadata.type -> entries.type when present.
-UPDATE entries
-SET type = metadata->>'type'
-WHERE metadata->>'type' IS NOT NULL
-  AND metadata->>'type' <> '';
-
--- 2) Remove metadata.type.
-UPDATE entries
-SET metadata = metadata - 'type'
-WHERE metadata ? 'type';
-
--- 3) Remove metadata.environment (duplicated by tags prod/dev).
-UPDATE entries
-SET metadata = metadata - 'environment'
-WHERE metadata ? 'environment';
-
--- 4) Remove metadata.account when equal to folder.
-UPDATE entries
-SET metadata = metadata - 'account'
-WHERE metadata->>'account' = folder;
-
--- 5) Normalize manufacturer -> provider.
-UPDATE entries
-SET metadata = (metadata - 'manufacturer')
-    || jsonb_build_object('provider', metadata->>'manufacturer')
-WHERE metadata ? 'manufacturer'
-  AND NOT metadata ? 'provider';
-
-UPDATE entries
-SET metadata = metadata - 'manufacturer'
-WHERE metadata ? 'manufacturer'
-  AND metadata ? 'provider';
-
--- 6) Drop ssh_key_format (moved to secrets.type).
-UPDATE entries
-SET metadata = metadata - 'ssh_key_format'
-WHERE metadata ? 'ssh_key_format';
-
--- 7) Remove display_name when duplicated by name.
-UPDATE entries
-SET metadata = metadata - 'display_name'
-WHERE metadata->>'display_name' = name;
-
--- 8) Condense server_* metadata into server_ref.
-UPDATE entries
-SET metadata = metadata
-    - 'server_account'
-    - 'server_hostname'
-    - 'server_location'
-    - 'server_public_ip'
-    || CASE
-        WHEN metadata ? 'server_entry_name'
-            THEN jsonb_build_object('server_ref', metadata->>'server_entry_name')
-        ELSE '{}'::jsonb
-    END
-WHERE metadata ? 'server_entry_name' OR metadata ? 'server_account';
-
-UPDATE entries
-SET metadata = metadata - 'server_entry_name'
-WHERE metadata ? 'server_entry_name';
-
-COMMIT;

scripts/cleanup-orphan-user-ids.sql

@@ -1,22 +0,0 @@
--- Run against prod BEFORE deploying secrets-mcp with FK migration.
--- Requires: write access to SECRETS_DATABASE_URL.
--- Example: psql "$SECRETS_DATABASE_URL" -v ON_ERROR_STOP=1 -f scripts/cleanup-orphan-user-ids.sql
-
-BEGIN;
-
-UPDATE entries
-SET user_id = NULL
-WHERE user_id IS NOT NULL
-  AND NOT EXISTS (SELECT 1 FROM users u WHERE u.id = entries.user_id);
-
-UPDATE entries_history
-SET user_id = NULL
-WHERE user_id IS NOT NULL
-  AND NOT EXISTS (SELECT 1 FROM users u WHERE u.id = entries_history.user_id);
-
-UPDATE audit_log
-SET user_id = NULL
-WHERE user_id IS NOT NULL
-  AND NOT EXISTS (SELECT 1 FROM users u WHERE u.id = audit_log.user_id);
-
-COMMIT;

scripts/migrate-db-prod-to-nn-test.sh

@@ -1,81 +0,0 @@
-#!/usr/bin/env bash
-# Migrate PostgreSQL data from secrets-mcp-prod to secrets-nn-test.
-#
-# Prereqs: pg_dump and pg_restore (PostgreSQL client tools) on PATH.
-# TLS: Use the same connection parameters as your MCP / app (e.g. sslmode=verify-full
-#      and PGSSLROOTCERT if needed). If local psql fails with "certificate verify failed",
-#      run this script from a host that trusts the server CA, or set PGSSLROOTCERT.
-#
-# Usage:
-#   export SOURCE_DATABASE_URL='postgres://USER:PASS@host:5432/secrets-mcp-prod?sslmode=verify-full'
-#   export TARGET_DATABASE_URL='postgres://USER:PASS@host:5432/secrets-nn-test?sslmode=verify-full'
-#   ./scripts/migrate-db-prod-to-nn-test.sh
-#
-# Options (env):
-#   BACKUP_TARGET_FIRST=1     # default: dump target to ./backup-secrets-nn-test-<timestamp>.dump before restore
-#   RUN_NN_SQL=1               # default: run migrations/001_nn_schema.sql then 002_data_cleanup.sql on target after restore
-#   SKIP_TARGET_BACKUP=1       # skip target backup
-#
-# WARNINGS:
-# - pg_restore with --clean --if-exists drops objects that exist in the dump; target DB is replaced
-#   to match the logical content of the source dump (same as typical full restore).
-# - Optionally keep a manual dump of the target before proceeding.
-# - 001_nn_schema.sql will fail if secrets has duplicate (user_id, name) after backfill; fix data first.
-
-set -euo pipefail
-
-ROOT="$(cd "$(dirname "${BASH_SOURCE[0]}")/.." && pwd)"
-cd "$ROOT"
-
-SOURCE_URL="${SOURCE_DATABASE_URL:-}"
-TARGET_URL="${TARGET_DATABASE_URL:-}"
-
-if [[ -z "$SOURCE_URL" || -z "$TARGET_URL" ]]; then
-  echo "Set SOURCE_DATABASE_URL and TARGET_DATABASE_URL (postgres URLs)." >&2
-  exit 1
-fi
-
-if ! command -v pg_dump >/dev/null || ! command -v pg_restore >/dev/null; then
-  echo "pg_dump and pg_restore are required." >&2
-  exit 1
-fi
-
-TS="$(date +%Y%m%d%H%M%S)"
-DUMP_FILE="${DUMP_FILE:-$ROOT/tmp/secrets-mcp-prod-${TS}.dump}"
-mkdir -p "$(dirname "$DUMP_FILE")"
-
-if [[ "${EXCLUDE_TOWER_SESSIONS:-}" == "1" ]]; then
-  echo "==> Excluding schema tower_sessions from dump"
-  pg_dump "$SOURCE_URL" -Fc --no-owner --no-acl --exclude-schema=tower_sessions -f "$DUMP_FILE"
-else
-  echo "==> Dumping source (custom format) -> $DUMP_FILE"
-  pg_dump "$SOURCE_URL" -Fc --no-owner --no-acl -f "$DUMP_FILE"
-fi
-
-if [[ "${SKIP_TARGET_BACKUP:-}" != "1" && "${BACKUP_TARGET_FIRST:-1}" == "1" ]]; then
-  BACKUP_FILE="$ROOT/tmp/secrets-nn-test-before-${TS}.dump"
-  echo "==> Backing up target -> $BACKUP_FILE"
-  pg_dump "$TARGET_URL" -Fc --no-owner --no-acl -f "$BACKUP_FILE" || {
-    echo "Target backup failed (empty DB is OK). Continuing." >&2
-  }
-fi
-
-echo "==> Restoring into target (--clean --if-exists)"
-pg_restore -d "$TARGET_URL" --no-owner --no-acl --clean --if-exists --verbose "$DUMP_FILE"
-
-if [[ "${RUN_NN_SQL:-1}" == "1" ]]; then
-  if [[ ! -f "$ROOT/migrations/001_nn_schema.sql" ]]; then
-    echo "migrations/001_nn_schema.sql not found; skip NN SQL." >&2
-  else
-    echo "==> Applying migrations/001_nn_schema.sql on target"
-    psql "$TARGET_URL" -v ON_ERROR_STOP=1 -f "$ROOT/migrations/001_nn_schema.sql"
-  fi
-  if [[ -f "$ROOT/migrations/002_data_cleanup.sql" ]]; then
-    echo "==> Applying migrations/002_data_cleanup.sql on target"
-    psql "$TARGET_URL" -v ON_ERROR_STOP=1 -f "$ROOT/migrations/002_data_cleanup.sql"
-  fi
-fi
-
-echo "==> Done. Suggested verification:"
-echo "    psql \"\$TARGET_DATABASE_URL\" -c \"SELECT COUNT(*) FROM entries; SELECT COUNT(*) FROM secrets; SELECT COUNT(*) FROM entry_secrets;\""
-echo "    ./scripts/release-check.sh   # optional app-side sanity"

scripts/migrate-v0.3.0.sql

@@ -1,194 +0,0 @@
--- ============================================================================
--- migrate-v0.3.0.sql
--- Schema migration from v0.2.x → v0.3.0
---
--- Changes:
---   • entries:         namespace → folder, kind → type; add notes column
---   • audit_log:       namespace → folder, kind → type
---   • entries_history: namespace → folder, kind → type; add user_id column
---   • Unique index: (user_id, name) → (user_id, folder, name)
---     Same name in different folders is now allowed; no rename needed.
---
--- Safe to run multiple times (fully idempotent).
--- Preserves all data in users, entries, secrets.
--- ============================================================================
-
-BEGIN;
-
--- ── entries: rename namespace→folder, kind→type ──────────────────────────────
-DO $$ BEGIN
-    IF EXISTS (
-        SELECT 1 FROM information_schema.columns
-        WHERE table_name = 'entries' AND column_name = 'namespace'
-    ) THEN
-        ALTER TABLE entries RENAME COLUMN namespace TO folder;
-    END IF;
-END $$;
-
-DO $$ BEGIN
-    IF EXISTS (
-        SELECT 1 FROM information_schema.columns
-        WHERE table_name = 'entries' AND column_name = 'kind'
-    ) THEN
-        ALTER TABLE entries RENAME COLUMN kind TO type;
-    END IF;
-END $$;
-
--- Set NOT NULL + default for folder/type in entries
-DO $$ BEGIN
-    IF EXISTS (
-        SELECT 1 FROM information_schema.columns
-        WHERE table_name = 'entries' AND column_name = 'folder'
-    ) THEN
-        UPDATE entries SET folder = '' WHERE folder IS NULL;
-        ALTER TABLE entries ALTER COLUMN folder SET NOT NULL;
-        ALTER TABLE entries ALTER COLUMN folder SET DEFAULT '';
-    END IF;
-END $$;
-
-DO $$ BEGIN
-    IF EXISTS (
-        SELECT 1 FROM information_schema.columns
-        WHERE table_name = 'entries' AND column_name = 'type'
-    ) THEN
-        UPDATE entries SET type = '' WHERE type IS NULL;
-        ALTER TABLE entries ALTER COLUMN type SET NOT NULL;
-        ALTER TABLE entries ALTER COLUMN type SET DEFAULT '';
-    END IF;
-END $$;
-
--- Add notes column to entries if missing
-ALTER TABLE entries ADD COLUMN IF NOT EXISTS notes TEXT NOT NULL DEFAULT '';
-
--- ── audit_log: rename namespace→folder, kind→type ────────────────────────────
-DO $$ BEGIN
-    IF EXISTS (
-        SELECT 1 FROM information_schema.columns
-        WHERE table_name = 'audit_log' AND column_name = 'namespace'
-    ) THEN
-        ALTER TABLE audit_log RENAME COLUMN namespace TO folder;
-    END IF;
-END $$;
-
-DO $$ BEGIN
-    IF EXISTS (
-        SELECT 1 FROM information_schema.columns
-        WHERE table_name = 'audit_log' AND column_name = 'kind'
-    ) THEN
-        ALTER TABLE audit_log RENAME COLUMN kind TO type;
-    END IF;
-END $$;
-
-DO $$ BEGIN
-    IF EXISTS (
-        SELECT 1 FROM information_schema.columns
-        WHERE table_name = 'audit_log' AND column_name = 'folder'
-    ) THEN
-        UPDATE audit_log SET folder = '' WHERE folder IS NULL;
-        ALTER TABLE audit_log ALTER COLUMN folder SET NOT NULL;
-        ALTER TABLE audit_log ALTER COLUMN folder SET DEFAULT '';
-    END IF;
-END $$;
-
-DO $$ BEGIN
-    IF EXISTS (
-        SELECT 1 FROM information_schema.columns
-        WHERE table_name = 'audit_log' AND column_name = 'type'
-    ) THEN
-        UPDATE audit_log SET type = '' WHERE type IS NULL;
-        ALTER TABLE audit_log ALTER COLUMN type SET NOT NULL;
-        ALTER TABLE audit_log ALTER COLUMN type SET DEFAULT '';
-    END IF;
-END $$;
-
-ALTER TABLE audit_log DROP COLUMN IF EXISTS actor;
-
--- ── entries_history: rename namespace→folder, kind→type; add user_id ─────────
-DO $$ BEGIN
-    IF EXISTS (
-        SELECT 1 FROM information_schema.columns
-        WHERE table_name = 'entries_history' AND column_name = 'namespace'
-    ) THEN
-        ALTER TABLE entries_history RENAME COLUMN namespace TO folder;
-    END IF;
-END $$;
-
-DO $$ BEGIN
-    IF EXISTS (
-        SELECT 1 FROM information_schema.columns
-        WHERE table_name = 'entries_history' AND column_name = 'kind'
-    ) THEN
-        ALTER TABLE entries_history RENAME COLUMN kind TO type;
-    END IF;
-END $$;
-
-DO $$ BEGIN
-    IF EXISTS (
-        SELECT 1 FROM information_schema.columns
-        WHERE table_name = 'entries_history' AND column_name = 'folder'
-    ) THEN
-        UPDATE entries_history SET folder = '' WHERE folder IS NULL;
-        ALTER TABLE entries_history ALTER COLUMN folder SET NOT NULL;
-        ALTER TABLE entries_history ALTER COLUMN folder SET DEFAULT '';
-    END IF;
-END $$;
-
-DO $$ BEGIN
-    IF EXISTS (
-        SELECT 1 FROM information_schema.columns
-        WHERE table_name = 'entries_history' AND column_name = 'type'
-    ) THEN
-        UPDATE entries_history SET type = '' WHERE type IS NULL;
-        ALTER TABLE entries_history ALTER COLUMN type SET NOT NULL;
-        ALTER TABLE entries_history ALTER COLUMN type SET DEFAULT '';
-    END IF;
-END $$;
-
-ALTER TABLE entries_history ADD COLUMN IF NOT EXISTS user_id UUID;
-ALTER TABLE entries_history DROP COLUMN IF EXISTS actor;
-
--- ── secrets_history: drop actor column ───────────────────────────────────────
-ALTER TABLE secrets_history DROP COLUMN IF EXISTS actor;
-
--- ── Rebuild unique indexes: (user_id, folder, name) ──────────────────────────
--- Note: folder is now part of the key, so same name in different folders is
--- naturally distinct — no rename of existing rows needed.
-DROP INDEX IF EXISTS idx_entries_unique_legacy;
-DROP INDEX IF EXISTS idx_entries_unique_user;
-
-CREATE UNIQUE INDEX IF NOT EXISTS idx_entries_unique_legacy
-    ON entries(folder, name)
-    WHERE user_id IS NULL;
-
-CREATE UNIQUE INDEX IF NOT EXISTS idx_entries_unique_user
-    ON entries(user_id, folder, name)
-    WHERE user_id IS NOT NULL;
-
--- ── Replace old namespace/kind indexes with folder/type ──────────────────────
-DROP INDEX IF EXISTS idx_entries_namespace;
-DROP INDEX IF EXISTS idx_entries_kind;
-DROP INDEX IF EXISTS idx_audit_log_ns_kind;
-DROP INDEX IF EXISTS idx_entries_history_ns_kind_name;
-
-CREATE INDEX IF NOT EXISTS idx_entries_folder
-    ON entries(folder) WHERE folder <> '';
-CREATE INDEX IF NOT EXISTS idx_entries_type
-    ON entries(type) WHERE type <> '';
-CREATE INDEX IF NOT EXISTS idx_entries_user_id
-    ON entries(user_id) WHERE user_id IS NOT NULL;
-CREATE INDEX IF NOT EXISTS idx_audit_log_folder_type
-    ON audit_log(folder, type);
-CREATE INDEX IF NOT EXISTS idx_entries_history_folder_type_name
-    ON entries_history(folder, type, name, version DESC);
-CREATE INDEX IF NOT EXISTS idx_entries_history_user_id
-    ON entries_history(user_id) WHERE user_id IS NOT NULL;
-
-COMMIT;
-
--- ── Verification queries (run these manually to confirm) ─────────────────────
--- SELECT column_name, data_type FROM information_schema.columns
---   WHERE table_name = 'entries' ORDER BY ordinal_position;
--- SELECT indexname, indexdef FROM pg_indexes WHERE tablename = 'entries';
--- SELECT COUNT(*) FROM entries;
--- SELECT COUNT(*) FROM users;
--- SELECT COUNT(*) FROM secrets;

scripts/sync-test-to-prod.sh

@@ -0,0 +1,95 @@
+#!/bin/bash
+# 同步测试环境数据到生产环境
+# 用法: ./scripts/sync-test-to-prod.sh
+
+set -euo pipefail
+
+# PostgreSQL 客户端工具路径 (Homebrew libpq)
+export PATH="/opt/homebrew/opt/libpq/bin:$PATH"
+
+# SSL 配置
+export PGSSLMODE=verify-full
+export PGSSLROOTCERT=/etc/ssl/cert.pem
+
+# 测试环境
+TEST_DB="postgres://postgres:Voson_2026_Pg18!@db.refining.ltd:5432/secrets-nn-test"
+
+# 生产环境
+PROD_DB="postgres://postgres:Voson_2026_Pg18!@db.refining.ltd:5432/secrets-nn-prod"
+
+echo "========================================="
+echo "  测试环境 -> 生产环境 数据同步"
+echo "========================================="
+echo ""
+
+# 确认操作
+read -p "⚠️  此操作将覆盖生产环境数据,确认继续? (yes/no): " confirm
+if [ "$confirm" != "yes" ]; then
+    echo "已取消"
+    exit 0
+fi
+
+echo ""
+echo "步骤 1/4: 导出测试环境数据..."
+TEMP_DIR=$(mktemp -d)
+trap "rm -rf $TEMP_DIR" EXIT
+
+# 导出测试环境数据(不含审计日志和历史记录)
+pg_dump "$TEST_DB" \
+    --table=entries \
+    --table=secrets \
+    --table=entry_secrets \
+    --table=users \
+    --table=oauth_accounts \
+    --data-only \
+    --column-inserts \
+    --no-owner \
+    --no-privileges \
+    > "$TEMP_DIR/test_data.sql"
+
+echo "✓ 测试数据已导出到临时文件"
+echo "  文件大小: $(du -h "$TEMP_DIR/test_data.sql" | cut -f1)"
+
+echo ""
+echo "步骤 2/4: 备份当前生产数据..."
+pg_dump "$PROD_DB" \
+    --table=entries \
+    --table=secrets \
+    --table=entry_secrets \
+    --table=users \
+    --table=oauth_accounts \
+    --data-only \
+    --column-inserts \
+    --no-owner \
+    --no-privileges \
+    > "$TEMP_DIR/prod_backup_$(date +%Y%m%d_%H%M%S).sql"
+
+echo "✓ 生产数据已备份"
+
+echo ""
+echo "步骤 3/4: 清空生产环境目标表..."
+psql "$PROD_DB" <<'SQL'
+TRUNCATE TABLE entry_secrets CASCADE;
+TRUNCATE TABLE secrets CASCADE;
+TRUNCATE TABLE entries CASCADE;
+SQL
+
+echo "✓ 生产环境目标表已清空"
+
+echo ""
+echo "步骤 4/4: 导入测试数据到生产环境..."
+psql "$PROD_DB" -f "$TEMP_DIR/test_data.sql" 2>&1 | tail -20
+
+echo ""
+echo "验证数据..."
+echo "生产环境数据统计:"
+psql "$PROD_DB" -c "SELECT 'users' as table_name, count(*) FROM users UNION ALL SELECT 'entries', count(*) FROM entries UNION ALL SELECT 'secrets', count(*) FROM secrets UNION ALL SELECT 'entry_secrets', count(*) FROM entry_secrets UNION ALL SELECT 'oauth_accounts', count(*) FROM oauth_accounts ORDER BY table_name;"
+
+echo ""
+echo "========================================="
+echo "  ✓ 数据同步完成!"
+echo "========================================="
+echo ""
+echo "提示:"
+echo "  - 生产数据备份已保存在临时目录"
+echo "  - 临时文件将在脚本退出后自动删除"