feat: 知识库支持目录来源——多个目录批量导入、后缀过滤（默认全部）、可开关递归子目录，只收录文本/图片视频/音频三类其余忽略

crates/core/src/app.rs

@@ -123,6 +123,8 @@ fn open_db(path: &Path) -> Result<Connection> {
     ensure_column(&conn, "scheduled_tasks", "email_mode", "TEXT NOT NULL DEFAULT 'fixed'")?;
     ensure_column(&conn, "scheduled_tasks", "email_subject", "TEXT NOT NULL DEFAULT ''")?;
     ensure_column(&conn, "scheduled_tasks", "email_body", "TEXT NOT NULL DEFAULT ''")?;
+    ensure_column(&conn, "kb_documents", "file_path", "TEXT NOT NULL DEFAULT ''")?;
+    ensure_column(&conn, "kb_documents", "source_id", "TEXT")?;
     Ok(conn)
 }
 

crates/core/src/knowledge_base.rs

@@ -24,10 +24,22 @@ pub struct KbDocument {
     pub file_size: i64,
     pub char_count: i64,
     pub chunk_count: i64,
+    pub file_path: String,
+    pub source_id: Option<String>,
     pub created_at: String,
     pub updated_at: String,
 }
 
+#[derive(Debug, Clone, Serialize, Deserialize)]
+pub struct KbSource {
+    pub id: String,
+    pub kb_id: String,
+    pub path: String,
+    pub extensions: String,
+    pub recursive: bool,
+    pub created_at: String,
+}
+
 #[derive(Debug, Clone, Serialize, Deserialize)]
 pub struct KbDocumentDetail {
     #[serde(flatten)]
@@ -45,7 +57,8 @@ pub struct KbSearchHit {
 }
 
 const KB_COLUMNS: &str = "id, name, description, chunk_size, chunk_overlap, created_at, updated_at";
-const DOC_COLUMNS: &str = "id, kb_id, name, file_type, file_size, char_count, chunk_count, created_at, updated_at";
+const DOC_COLUMNS: &str = "id, kb_id, name, file_type, file_size, char_count, chunk_count, file_path, source_id, created_at, updated_at";
+const SOURCE_COLUMNS: &str = "id, kb_id, path, extensions, recursive, created_at";
 
 // ---------- 知识库 ----------
 
@@ -149,7 +162,7 @@ pub fn get_document_detail(db: &Connection, id: &str) -> Result<Option<KbDocumen
     ))?;
     let mut rows = stmt.query_map(params![id], |row| {
         let doc = row_to_doc(row)?;
-        let content: String = row.get(9)?;
+        let content: String = row.get(11)?;
         Ok(KbDocumentDetail { doc, content })
     })?;
     match rows.next() {
@@ -168,8 +181,8 @@ pub fn insert_document_with_chunks(
     let tx = db.unchecked_transaction()?;
     tx.execute(
         "INSERT INTO kb_documents
-           (id, kb_id, name, file_type, file_size, char_count, chunk_count, content)
-         VALUES (?1, ?2, ?3, ?4, ?5, ?6, ?7, ?8)",
+           (id, kb_id, name, file_type, file_size, char_count, chunk_count, content, file_path, source_id)
+         VALUES (?1, ?2, ?3, ?4, ?5, ?6, ?7, ?8, ?9, ?10)",
         params![
             doc.id,
             doc.kb_id,
@@ -179,6 +192,8 @@ pub fn insert_document_with_chunks(
             doc.char_count,
             chunks.len() as i64,
             content,
+            doc.file_path,
+            doc.source_id,
         ],
     )?;
     for (i, chunk) in chunks.iter().enumerate() {
@@ -203,6 +218,75 @@ pub fn delete_document(db: &Connection, id: &str) -> Result<()> {
     Ok(())
 }
 
+/// 同一知识库下是否已存在来自该文件路径的文档（目录导入去重）。
+pub fn document_exists(db: &Connection, kb_id: &str, file_path: &str) -> Result<bool> {
+    if file_path.is_empty() {
+        return Ok(false);
+    }
+    let exists: bool = db.query_row(
+        "SELECT EXISTS(SELECT 1 FROM kb_documents WHERE kb_id = ?1 AND file_path = ?2)",
+        params![kb_id, file_path],
+        |row| row.get(0),
+    )?;
+    Ok(exists)
+}
+
+// ---------- 目录来源 ----------
+
+pub fn list_sources(db: &Connection, kb_id: &str) -> Result<Vec<KbSource>> {
+    let mut stmt = db.prepare(&format!(
+        "SELECT {SOURCE_COLUMNS} FROM kb_sources WHERE kb_id = ?1 ORDER BY created_at ASC"
+    ))?;
+    let rows = stmt.query_map(params![kb_id], row_to_source)?;
+    let mut out = Vec::new();
+    for row in rows {
+        out.push(row?);
+    }
+    Ok(out)
+}
+
+pub fn get_source(db: &Connection, id: &str) -> Result<Option<KbSource>> {
+    let mut stmt = db.prepare(&format!(
+        "SELECT {SOURCE_COLUMNS} FROM kb_sources WHERE id = ?1"
+    ))?;
+    let mut rows = stmt.query_map(params![id], row_to_source)?;
+    match rows.next() {
+        Some(row) => Ok(Some(row?)),
+        None => Ok(None),
+    }
+}
+
+pub fn upsert_source(db: &Connection, source: &KbSource) -> Result<()> {
+    db.execute(
+        "INSERT INTO kb_sources (id, kb_id, path, extensions, recursive)
+         VALUES (?1, ?2, ?3, ?4, ?5)
+         ON CONFLICT(kb_id, path) DO UPDATE SET
+           extensions = excluded.extensions, recursive = excluded.recursive",
+        params![
+            source.id,
+            source.kb_id,
+            source.path,
+            source.extensions,
+            source.recursive as i32,
+        ],
+    )?;
+    Ok(())
+}
+
+pub fn delete_source(db: &Connection, id: &str) -> Result<()> {
+    db.execute("DELETE FROM kb_sources WHERE id = ?1", params![id])?;
+    Ok(())
+}
+
+/// 删除某个目录来源导入的全部文档（分块随外键级联）。
+pub fn delete_docs_by_source(db: &Connection, source_id: &str) -> Result<usize> {
+    let n = db.execute(
+        "DELETE FROM kb_documents WHERE source_id = ?1",
+        params![source_id],
+    )?;
+    Ok(n)
+}
+
 /// 重新切分文档：删除旧分块后按新参数重分并重建索引。
 pub fn rechunk_document(
     db: &Connection,
@@ -328,8 +412,22 @@ fn row_to_doc(row: &rusqlite::Row<'_>) -> rusqlite::Result<KbDocument> {
         file_size: row.get(4)?,
         char_count: row.get(5)?,
         chunk_count: row.get(6)?,
-        created_at: row.get(7)?,
-        updated_at: row.get(8)?,
+        file_path: row.get(7)?,
+        source_id: row.get(8)?,
+        created_at: row.get(9)?,
+        updated_at: row.get(10)?,
+    })
+}
+
+fn row_to_source(row: &rusqlite::Row<'_>) -> rusqlite::Result<KbSource> {
+    let recursive: i32 = row.get(4)?;
+    Ok(KbSource {
+        id: row.get(0)?,
+        kb_id: row.get(1)?,
+        path: row.get(2)?,
+        extensions: row.get(3)?,
+        recursive: recursive != 0,
+        created_at: row.get(5)?,
     })
 }
 
@@ -363,6 +461,8 @@ mod tests {
             file_size: content.len() as i64,
             char_count: content.chars().count() as i64,
             chunk_count: 0,
+            file_path: String::new(),
+            source_id: None,
             created_at: String::new(),
             updated_at: String::new(),
         };

crates/core/src/lib.rs

@@ -12,7 +12,7 @@ pub mod workflows;
 
 pub use app::CoreApp;
 pub use error::{CoreError, Result};
-pub use knowledge_base::{KbDocument, KbDocumentDetail, KbSearchHit, KnowledgeBase};
+pub use knowledge_base::{KbDocument, KbDocumentDetail, KbSearchHit, KbSource, KnowledgeBase};
 pub use models::ModelInfo;
 pub use agents::Agent;
 pub use scheduled_tasks::ScheduledTask;

crates/core/src/schema.sql

@@ -147,6 +147,8 @@ CREATE TABLE IF NOT EXISTS kb_documents (
     char_count  INTEGER NOT NULL DEFAULT 0,
     chunk_count INTEGER NOT NULL DEFAULT 0,
     content     TEXT NOT NULL DEFAULT '',
+    file_path   TEXT NOT NULL DEFAULT '',
+    source_id   TEXT,
     created_at  TEXT NOT NULL DEFAULT (datetime('now')),
     updated_at  TEXT NOT NULL DEFAULT (datetime('now'))
 );
@@ -160,6 +162,17 @@ CREATE TABLE IF NOT EXISTS kb_chunks (
     created_at  TEXT NOT NULL DEFAULT (datetime('now'))
 );
 
+CREATE TABLE IF NOT EXISTS kb_sources (
+    id         TEXT PRIMARY KEY,
+    kb_id      TEXT NOT NULL REFERENCES knowledge_bases(id) ON DELETE CASCADE,
+    path       TEXT NOT NULL,
+    extensions TEXT NOT NULL DEFAULT '',
+    recursive  INTEGER NOT NULL DEFAULT 1,
+    created_at TEXT NOT NULL DEFAULT (datetime('now'))
+);
+
+CREATE UNIQUE INDEX IF NOT EXISTS idx_kb_sources_kb_path ON kb_sources(kb_id, path);
+
 -- 全文检索索引（trigram 分词，对中文/短文本更友好；content='' 为 contentless-delete 模式）
 CREATE VIRTUAL TABLE IF NOT EXISTS kb_chunks_fts USING fts5(content, content = '', tokenize = 'trigram');