Fix lỗi upload file không được lưu trữ lại lâu dài

backend/admin_services.py

@@ -16,6 +16,7 @@ if ROOT_DIR not in sys.path:
 from backend import db
 from backend.auth import is_admin
 from backend.runtime_paths import PDF_DIR
+from backend.db_sync import schedule_pdf_upload, schedule_pdf_delete, schedule_vector_sync
 
 
 # ======================== UserService ========================
@@ -108,6 +109,7 @@ def create_system_doc(file_path: str, filename: str = None) -> tuple[bool, str]:
         else:
             shutil.copy2(file_path, dest)
         db.insert_tai_lieu_he_thong(ma_tai_lieu=ma, ten_file=filename, duong_dan=os.path.abspath(dest))
+        schedule_pdf_upload(dest, filename)
         return True, "Đã thêm tài liệu. Hãy chạy 'Tái lập chỉ mục' để cập nhật RAG."
     except Exception as e:
         return False, str(e)
@@ -131,6 +133,7 @@ def update_system_doc(ma_tai_lieu: str, file_path: str = None, ten_file: str = N
             return False, str(e)
     # Cập nhật tên trong DB nếu đổi tên (cần hàm update trong db - hiện chỉ có insert upsert)
     db.insert_tai_lieu_he_thong(ma_tai_lieu=ma_tai_lieu, ten_file=new_name, duong_dan=old_path)
+    schedule_pdf_upload(old_path, new_name)
     return True, "Đã cập nhật. Chạy 'Tái lập chỉ mục' nếu đã thay file."
 
 
@@ -140,12 +143,15 @@ def delete_system_doc(ma_tai_lieu: str) -> tuple[bool, str]:
     if not docs:
         return False, "Không tìm thấy tài liệu"
     path = docs[0].get("duong_dan")
+    filename = docs[0].get("ten_file") or (os.path.basename(path) if path else "")
     try:
         if path and os.path.isfile(path):
             os.remove(path)
     except OSError:
         pass
     db.delete_tai_lieu_he_thong(ma_tai_lieu)
+    if filename:
+        schedule_pdf_delete(filename)
     return True, "Đã xóa tài liệu. Chạy 'Tái lập chỉ mục' để cập nhật RAG."
 
 
@@ -170,6 +176,7 @@ def reindex_all() -> tuple[bool, str]:
         except Exception:
             pass
         create_vector_database(chunks)
+        schedule_vector_sync()
         return True, f"Đã tái lập chỉ mục: {len(documents)} tài liệu, {len(chunks)} chunks."
     except Exception as e:
         return False, f"Lỗi: {str(e)}"

backend/db.py

@@ -482,6 +482,7 @@ def save_tai_lieu(ma_tai_lieu: str, ma_nguoi_dung: str, ten_file: str,
         conn.commit()
     finally:
         conn.close()
+    _schedule_sync()
 
 
 def update_trang_thai_tai_lieu(ma_tai_lieu: str, trang_thai: str):
@@ -495,6 +496,7 @@ def update_trang_thai_tai_lieu(ma_tai_lieu: str, trang_thai: str):
         conn.commit()
     finally:
         conn.close()
+    _schedule_sync()
 
 
 def load_tai_lieu_by_user(ma_nguoi_dung: str) -> list:
@@ -520,6 +522,7 @@ def delete_tai_lieu(ma_tai_lieu: str):
         conn.commit()
     finally:
         conn.close()
+    _schedule_sync()
 
 
 # ======================== PHẢN HỒI NGƯỜI DÙNG ========================
@@ -725,6 +728,7 @@ def insert_tai_lieu_he_thong(ma_tai_lieu: str, ten_file: str, duong_dan: str):
         conn.commit()
     finally:
         conn.close()
+    _schedule_sync()
 
 
 def delete_tai_lieu_he_thong(ma_tai_lieu: str):
@@ -735,6 +739,7 @@ def delete_tai_lieu_he_thong(ma_tai_lieu: str):
         conn.commit()
     finally:
         conn.close()
+    _schedule_sync()
 
 
 # Khởi tạo DB khi import module

backend/db_sync.py

@@ -1,5 +1,5 @@
 """
-Đồng bộ chatbot.db lên Hugging Face Dataset repo để dữ liệu bền vững.
+Đồng bộ dữ liệu runtime lên Hugging Face Dataset repo để dữ liệu bền vững.
 
 Chỉ hoạt động khi chạy trên HF Space và có cấu hình:
   - HF_DATASET_REPO
@@ -13,6 +13,7 @@ import shutil
 import sqlite3
 import threading
 import time
+from datetime import datetime, timezone
 
 _lock = threading.Lock()
 _last_sync: float = 0
@@ -77,6 +78,142 @@ def _do_upload(db_path: str, repo: str, token: str, revision: str) -> bool:
         return False
 
 
+def _spawn_sync(task) -> None:
+    """Chạy tác vụ sync trong background thread."""
+    thread = threading.Thread(target=task, daemon=True)
+    thread.start()
+
+
+def _upload_manifest(repo: str, token: str, revision: str, include_pdf: bool, include_vector: bool) -> None:
+    """Cập nhật manifest.json cơ bản trong dataset repo."""
+    try:
+        from huggingface_hub import HfApi
+        import json
+        import tempfile
+
+        payload = {
+            "generated_at": datetime.now(timezone.utc).isoformat(),
+            "includes": {
+                "chatbot_db": True,
+                "pdf": include_pdf,
+                "csdl_vector": include_vector,
+            },
+        }
+        with tempfile.NamedTemporaryFile("w", suffix=".json", delete=False, encoding="utf-8") as f:
+            json.dump(payload, f, ensure_ascii=False, indent=2)
+            temp_manifest = f.name
+
+        try:
+            api = HfApi(token=token)
+            api.upload_file(
+                path_or_fileobj=temp_manifest,
+                path_in_repo="manifest.json",
+                repo_id=repo,
+                repo_type="dataset",
+                revision=revision,
+                commit_message="Auto-update manifest",
+            )
+        finally:
+            try:
+                os.remove(temp_manifest)
+            except OSError:
+                pass
+    except Exception as exc:
+        print(f"[DB_SYNC] Manifest update failed: {exc}")
+
+
+def schedule_pdf_upload(local_pdf_path: str, remote_filename: str | None = None) -> None:
+    """Upload 1 file PDF mới/chỉnh sửa lên dataset repo tại pdf/<filename>."""
+    config = _get_config()
+    if not config:
+        return
+    if not os.path.exists(local_pdf_path):
+        return
+
+    repo, token, revision = config
+    remote_name = remote_filename or os.path.basename(local_pdf_path)
+
+    def _task():
+        with _lock:
+            try:
+                from huggingface_hub import HfApi
+                api = HfApi(token=token)
+                api.upload_file(
+                    path_or_fileobj=local_pdf_path,
+                    path_in_repo=f"pdf/{remote_name}",
+                    repo_id=repo,
+                    repo_type="dataset",
+                    revision=revision,
+                    commit_message=f"Auto-sync PDF: {remote_name}",
+                )
+                _upload_manifest(repo, token, revision, include_pdf=True, include_vector=False)
+                print(f"[DB_SYNC] Uploaded pdf/{remote_name} to {repo}")
+            except Exception as exc:
+                print(f"[DB_SYNC] PDF upload failed: {exc}")
+
+    _spawn_sync(_task)
+
+
+def schedule_pdf_delete(remote_filename: str) -> None:
+    """Xóa 1 file PDF từ dataset repo tại pdf/<filename>."""
+    config = _get_config()
+    if not config or not remote_filename:
+        return
+    repo, token, revision = config
+
+    def _task():
+        with _lock:
+            try:
+                from huggingface_hub import HfApi
+                api = HfApi(token=token)
+                api.delete_file(
+                    path_in_repo=f"pdf/{remote_filename}",
+                    repo_id=repo,
+                    repo_type="dataset",
+                    revision=revision,
+                    commit_message=f"Auto-remove PDF: {remote_filename}",
+                )
+                _upload_manifest(repo, token, revision, include_pdf=True, include_vector=False)
+                print(f"[DB_SYNC] Deleted pdf/{remote_filename} from {repo}")
+            except Exception as exc:
+                print(f"[DB_SYNC] PDF delete failed: {exc}")
+
+    _spawn_sync(_task)
+
+
+def schedule_vector_sync() -> None:
+    """Upload toàn bộ thư mục csdl_vector lên dataset repo."""
+    config = _get_config()
+    if not config:
+        return
+    from backend.runtime_paths import VECTOR_DIR
+
+    if not os.path.isdir(VECTOR_DIR):
+        return
+
+    repo, token, revision = config
+
+    def _task():
+        with _lock:
+            try:
+                from huggingface_hub import HfApi
+                api = HfApi(token=token)
+                api.upload_folder(
+                    folder_path=VECTOR_DIR,
+                    path_in_repo="csdl_vector",
+                    repo_id=repo,
+                    repo_type="dataset",
+                    revision=revision,
+                    commit_message="Auto-sync vector store",
+                )
+                _upload_manifest(repo, token, revision, include_pdf=True, include_vector=True)
+                print(f"[DB_SYNC] Uploaded csdl_vector to {repo}")
+            except Exception as exc:
+                print(f"[DB_SYNC] Vector sync failed: {exc}")
+
+    _spawn_sync(_task)
+
+
 def schedule_sync(db_path: str | None = None) -> None:
     """
     Lên lịch đồng bộ DB lên dataset repo (chạy background, có rate-limit).
@@ -109,5 +246,4 @@ def schedule_sync(db_path: str | None = None) -> None:
             _last_sync = time.time()
             _do_upload(db_path, repo, token, revision)
 
-    thread = threading.Thread(target=_sync, daemon=True)
-    thread.start()
+    _spawn_sync(_sync)