Refactor: move encrypt/decrypt to one file (#10203)

### What problem does this PR solve?

Move base64 related function to api/common/base64.py

### Type of change

- [x] Refactoring

---------

Signed-off-by: jinhai <haijin.chn@gmail.com>

.github/workflows/release.yml

@@ -88,7 +88,9 @@ jobs:
         with:
           context: .
           push: true
-          tags: infiniflow/ragflow:${{ env.RELEASE_TAG }}
+          tags: |
+            infiniflow/ragflow:${{ env.RELEASE_TAG }}
+            infiniflow/ragflow:latest-full
           file: Dockerfile
           platforms: linux/amd64
 
@@ -98,7 +100,9 @@ jobs:
         with:
           context: .
           push: true
-          tags: infiniflow/ragflow:${{ env.RELEASE_TAG }}-slim
+          tags: |
+            infiniflow/ragflow:${{ env.RELEASE_TAG }}-slim
+            infiniflow/ragflow:latest-slim
           file: Dockerfile
           build-args: LIGHTEN=1
           platforms: linux/amd64

admin/README.md

@@ -0,0 +1,101 @@
+# RAGFlow Admin Service & CLI
+
+### Introduction
+
+Admin Service is a dedicated management component designed to monitor, maintain, and administrate the RAGFlow system. It provides comprehensive tools for ensuring system stability, performing operational tasks, and managing users and permissions efficiently.
+
+The service offers real-time monitoring of critical components, including the RAGFlow server, Task Executor processes, and dependent services such as MySQL, Elasticsearch, Redis, and MinIO. It automatically checks their health status, resource usage, and uptime, and performs restarts in case of failures to minimize downtime.
+
+For user and system management, it supports listing, creating, modifying, and deleting users and their associated resources like knowledge bases and Agents.
+
+Built with scalability and reliability in mind, the Admin Service ensures smooth system operation and simplifies maintenance workflows.
+
+It consists of a server-side Service and a command-line client (CLI), both implemented in Python. User commands are parsed using the Lark parsing toolkit.
+
+- **Admin Service**: A backend service that interfaces with the RAGFlow system to execute administrative operations and monitor its status.
+- **Admin CLI**: A command-line interface that allows users to connect to the Admin Service and issue commands for system management.
+
+### Starting the Admin Service
+
+1.  Before start Admin Service, please make sure RAGFlow system is already started.
+
+2.  Run the service script:
+    ```bash
+    python admin/admin_server.py
+    ```
+    The service will start and listen for incoming connections from the CLI on the configured port.
+
+### Using the Admin CLI
+
+1.  Ensure the Admin Service is running.
+2.  Launch the CLI client:
+    ```bash
+    python admin/admin_client.py -h 0.0.0.0 -p 9381
+
+## Supported Commands
+
+Commands are case-insensitive and must be terminated with a semicolon (`;`).
+
+### Service Management Commands
+
+-   `LIST SERVICES;`
+    -   Lists all available services within the RAGFlow system.
+-   `SHOW SERVICE <id>;`
+    -   Shows detailed status information for the service identified by `<id>`.
+-   `STARTUP SERVICE <id>;`
+    -   Attempts to start the service identified by `<id>`.
+-   `SHUTDOWN SERVICE <id>;`
+    -   Attempts to gracefully shut down the service identified by `<id>`.
+-   `RESTART SERVICE <id>;`
+    -   Attempts to restart the service identified by `<id>`.
+
+### User Management Commands
+
+-   `LIST USERS;`
+    -   Lists all users known to the system.
+-   `SHOW USER '<username>';`
+    -   Shows details and permissions for the specified user. The username must be enclosed in single or double quotes.
+-   `DROP USER '<username>';`
+    -   Removes the specified user from the system. Use with caution.
+-   `ALTER USER PASSWORD '<username>' '<new_password>';`
+    -   Changes the password for the specified user.
+
+### Data and Agent Commands
+
+-   `LIST DATASETS OF '<username>';`
+    -   Lists the datasets associated with the specified user.
+-   `LIST AGENTS OF '<username>';`
+    -   Lists the agents associated with the specified user.
+
+### Meta-Commands
+
+Meta-commands are prefixed with a backslash (`\`).
+
+-   `\?` or `\help`
+    -   Shows help information for the available commands.
+-   `\q` or `\quit`
+    -   Exits the CLI application.
+
+## Examples
+
+```commandline
+admin> list users;
++-------------------------------+------------------------+-----------+-------------+
+| create_date                   | email                  | is_active | nickname    |
++-------------------------------+------------------------+-----------+-------------+
+| Fri, 22 Nov 2024 16:03:41 GMT | jeffery@infiniflow.org | 1         | Jeffery     |
+| Fri, 22 Nov 2024 16:10:55 GMT | aya@infiniflow.org     | 1         | Waterdancer |
++-------------------------------+------------------------+-----------+-------------+
+
+admin> list services;
++-------------------------------------------------------------------------------------------+-----------+----+---------------+-------+----------------+
+| extra                                                                                     | host      | id | name          | port  | service_type   |
++-------------------------------------------------------------------------------------------+-----------+----+---------------+-------+----------------+
+| {}                                                                                        | 0.0.0.0   | 0  | ragflow_0     | 9380  | ragflow_server |
+| {'meta_type': 'mysql', 'password': 'infini_rag_flow', 'username': 'root'}                 | localhost | 1  | mysql         | 5455  | meta_data      |
+| {'password': 'infini_rag_flow', 'store_type': 'minio', 'user': 'rag_flow'}                | localhost | 2  | minio         | 9000  | file_store     |
+| {'password': 'infini_rag_flow', 'retrieval_type': 'elasticsearch', 'username': 'elastic'} | localhost | 3  | elasticsearch | 1200  | retrieval      |
+| {'db_name': 'default_db', 'retrieval_type': 'infinity'}                                   | localhost | 4  | infinity      | 23817 | retrieval      |
+| {'database': 1, 'mq_type': 'redis', 'password': 'infini_rag_flow'}                        | localhost | 5  | redis         | 6379  | message_queue  |
++-------------------------------------------------------------------------------------------+-----------+----+---------------+-------+----------------+
+```

admin/admin_client.py

@@ -0,0 +1,471 @@
+import argparse
+import base64
+from typing import Dict, List, Any
+from lark import Lark, Transformer, Tree
+import requests
+from requests.auth import HTTPBasicAuth
+
+GRAMMAR = r"""
+start: command
+
+command: sql_command | meta_command
+
+sql_command: list_services
+           | show_service
+           | startup_service
+           | shutdown_service
+           | restart_service
+           | list_users
+           | show_user
+           | drop_user
+           | alter_user
+           | list_datasets
+           | list_agents
+
+// meta command definition
+meta_command: "\\" meta_command_name [meta_args]
+
+meta_command_name: /[a-zA-Z?]+/
+meta_args: (meta_arg)+
+
+meta_arg: /[^\\s"']+/ | quoted_string
+
+// command definition
+
+LIST: "LIST"i
+SERVICES: "SERVICES"i
+SHOW: "SHOW"i
+SERVICE: "SERVICE"i
+SHUTDOWN: "SHUTDOWN"i
+STARTUP: "STARTUP"i
+RESTART: "RESTART"i
+USERS: "USERS"i
+DROP: "DROP"i
+USER: "USER"i
+ALTER: "ALTER"i
+PASSWORD: "PASSWORD"i
+DATASETS: "DATASETS"i
+OF: "OF"i
+AGENTS: "AGENTS"i
+
+list_services: LIST SERVICES ";"
+show_service: SHOW SERVICE NUMBER ";"
+startup_service: STARTUP SERVICE NUMBER ";"
+shutdown_service: SHUTDOWN SERVICE NUMBER ";"
+restart_service: RESTART SERVICE NUMBER ";"
+
+list_users: LIST USERS ";"
+drop_user: DROP USER quoted_string ";"
+alter_user: ALTER USER PASSWORD quoted_string quoted_string ";"
+show_user: SHOW USER quoted_string ";"
+
+list_datasets: LIST DATASETS OF quoted_string ";"
+list_agents: LIST AGENTS OF quoted_string ";"
+
+identifier: WORD
+quoted_string: QUOTED_STRING
+
+QUOTED_STRING: /'[^']+'/ | /"[^"]+"/
+WORD: /[a-zA-Z0-9_\-\.]+/
+NUMBER: /[0-9]+/
+
+%import common.WS
+%ignore WS
+"""
+
+
+class AdminTransformer(Transformer):
+
+    def start(self, items):
+        return items[0]
+
+    def command(self, items):
+        return items[0]
+
+    def list_services(self, items):
+        result = {'type': 'list_services'}
+        return result
+
+    def show_service(self, items):
+        service_id = int(items[2])
+        return {"type": "show_service", "number": service_id}
+
+    def startup_service(self, items):
+        service_id = int(items[2])
+        return {"type": "startup_service", "number": service_id}
+
+    def shutdown_service(self, items):
+        service_id = int(items[2])
+        return {"type": "shutdown_service", "number": service_id}
+
+    def restart_service(self, items):
+        service_id = int(items[2])
+        return {"type": "restart_service", "number": service_id}
+
+    def list_users(self, items):
+        return {"type": "list_users"}
+
+    def show_user(self, items):
+        user_name = items[2]
+        return {"type": "show_user", "username": user_name}
+
+    def drop_user(self, items):
+        user_name = items[2]
+        return {"type": "drop_user", "username": user_name}
+
+    def alter_user(self, items):
+        user_name = items[3]
+        new_password = items[4]
+        return {"type": "alter_user", "username": user_name, "password": new_password}
+
+    def list_datasets(self, items):
+        user_name = items[3]
+        return {"type": "list_datasets", "username": user_name}
+
+    def list_agents(self, items):
+        user_name = items[3]
+        return {"type": "list_agents", "username": user_name}
+
+    def meta_command(self, items):
+        command_name = str(items[0]).lower()
+        args = items[1:] if len(items) > 1 else []
+
+        # handle quoted parameter
+        parsed_args = []
+        for arg in args:
+            if hasattr(arg, 'value'):
+                parsed_args.append(arg.value)
+            else:
+                parsed_args.append(str(arg))
+
+        return {'type': 'meta', 'command': command_name, 'args': parsed_args}
+
+    def meta_command_name(self, items):
+        return items[0]
+
+    def meta_args(self, items):
+        return items
+
+
+def encode_to_base64(input_string):
+    base64_encoded = base64.b64encode(input_string.encode('utf-8'))
+    return base64_encoded.decode('utf-8')
+
+
+class AdminCommandParser:
+    def __init__(self):
+        self.parser = Lark(GRAMMAR, start='start', parser='lalr', transformer=AdminTransformer())
+        self.command_history = []
+
+    def parse_command(self, command_str: str) -> Dict[str, Any]:
+        if not command_str.strip():
+            return {'type': 'empty'}
+
+        self.command_history.append(command_str)
+
+        try:
+            result = self.parser.parse(command_str)
+            return result
+        except Exception as e:
+            return {'type': 'error', 'message': f'Parse error: {str(e)}'}
+
+
+class AdminCLI:
+    def __init__(self):
+        self.parser = AdminCommandParser()
+        self.is_interactive = False
+        self.admin_account = "admin@ragflow.io"
+        self.admin_password: str = "admin"
+        self.host: str = ""
+        self.port: int = 0
+
+    def verify_admin(self, args):
+
+        conn_info = self._parse_connection_args(args)
+        if 'error' in conn_info:
+            print(f"Error: {conn_info['error']}")
+            return
+
+        self.host = conn_info['host']
+        self.port = conn_info['port']
+        print(f"Attempt to access ip: {self.host}, port: {self.port}")
+        url = f'http://{self.host}:{self.port}/api/v1/admin/auth'
+
+        try_count = 0
+        while True:
+            try_count += 1
+            if try_count > 3:
+                return False
+
+            admin_passwd = input(f"password for {self.admin_account}: ").strip()
+            try:
+                self.admin_password = encode_to_base64(admin_passwd)
+                response = requests.get(url, auth=HTTPBasicAuth(self.admin_account, self.admin_password))
+                if response.status_code == 200:
+                    res_json = response.json()
+                    error_code = res_json.get('code', -1)
+                    if error_code == 0:
+                        print("Authentication successful.")
+                        return True
+                    else:
+                        error_message = res_json.get('message', 'Unknown error')
+                        print(f"Authentication failed: {error_message}, try again")
+                        continue
+                else:
+                    print(f"Bad response，status: {response.status_code}, try again")
+            except Exception:
+                print(f"Can't access {self.host}, port: {self.port}")
+
+    def _print_table_simple(self, data):
+        if not data:
+            print("No data to print")
+            return
+
+        columns = list(data[0].keys())
+        col_widths = {}
+
+        for col in columns:
+            max_width = len(str(col))
+            for item in data:
+                value_len = len(str(item.get(col, '')))
+                if value_len > max_width:
+                    max_width = value_len
+            col_widths[col] = max(2, max_width)
+
+        # Generate delimiter
+        separator = "+" + "+".join(["-" * (col_widths[col] + 2) for col in columns]) + "+"
+
+        # Print header
+        print(separator)
+        header = "|" + "|".join([f" {col:<{col_widths[col]}} " for col in columns]) + "|"
+        print(header)
+        print(separator)
+
+        # Print data
+        for item in data:
+            row = "|"
+            for col in columns:
+                value = str(item.get(col, ''))
+                if len(value) > col_widths[col]:
+                    value = value[:col_widths[col] - 3] + "..."
+                row += f" {value:<{col_widths[col]}} |"
+            print(row)
+
+        print(separator)
+
+    def run_interactive(self):
+
+        self.is_interactive = True
+        print("RAGFlow Admin command line interface - Type '\\?' for help, '\\q' to quit")
+
+        while True:
+            try:
+                command = input("admin> ").strip()
+                if not command:
+                    continue
+
+                print(f"command: {command}")
+                result = self.parser.parse_command(command)
+                self.execute_command(result)
+
+                if isinstance(result, Tree):
+                    continue
+
+                if result.get('type') == 'meta' and result.get('command') in ['q', 'quit', 'exit']:
+                    break
+
+            except KeyboardInterrupt:
+                print("\nUse '\\q' to quit")
+            except EOFError:
+                print("\nGoodbye!")
+                break
+
+    def run_single_command(self, args):
+        conn_info = self._parse_connection_args(args)
+        if 'error' in conn_info:
+            print(f"Error: {conn_info['error']}")
+            return
+
+    def _parse_connection_args(self, args: List[str]) -> Dict[str, Any]:
+        parser = argparse.ArgumentParser(description='Admin CLI Client', add_help=False)
+        parser.add_argument('-h', '--host', default='localhost', help='Admin service host')
+        parser.add_argument('-p', '--port', type=int, default=8080, help='Admin service port')
+
+        try:
+            parsed_args, remaining_args = parser.parse_known_args(args)
+            return {
+                'host': parsed_args.host,
+                'port': parsed_args.port,
+            }
+        except SystemExit:
+            return {'error': 'Invalid connection arguments'}
+
+    def execute_command(self, parsed_command: Dict[str, Any]):
+
+        command_dict: dict
+        if isinstance(parsed_command, Tree):
+            command_dict = parsed_command.children[0]
+        else:
+            if parsed_command['type'] == 'error':
+                print(f"Error: {parsed_command['message']}")
+                return
+            else:
+                command_dict = parsed_command
+
+        # print(f"Parsed command: {command_dict}")
+
+        command_type = command_dict['type']
+
+        match command_type:
+            case 'list_services':
+                self._handle_list_services(command_dict)
+            case 'show_service':
+                self._handle_show_service(command_dict)
+            case 'restart_service':
+                self._handle_restart_service(command_dict)
+            case 'shutdown_service':
+                self._handle_shutdown_service(command_dict)
+            case 'startup_service':
+                self._handle_startup_service(command_dict)
+            case 'list_users':
+                self._handle_list_users(command_dict)
+            case 'show_user':
+                self._handle_show_user(command_dict)
+            case 'drop_user':
+                self._handle_drop_user(command_dict)
+            case 'alter_user':
+                self._handle_alter_user(command_dict)
+            case 'list_datasets':
+                self._handle_list_datasets(command_dict)
+            case 'list_agents':
+                self._handle_list_agents(command_dict)
+            case 'meta':
+                self._handle_meta_command(command_dict)
+            case _:
+                print(f"Command '{command_type}' would be executed with API")
+
+    def _handle_list_services(self, command):
+        print("Listing all services")
+
+        url = f'http://{self.host}:{self.port}/api/v1/admin/services'
+        response = requests.get(url, auth=HTTPBasicAuth(self.admin_account, self.admin_password))
+        res_json = dict
+        if response.status_code == 200:
+            res_json = response.json()
+            self._print_table_simple(res_json['data'])
+        else:
+            print(f"Fail to get all users, code: {res_json['code']}, message: {res_json['message']}")
+
+    def _handle_show_service(self, command):
+        service_id: int = command['number']
+        print(f"Showing service: {service_id}")
+
+    def _handle_restart_service(self, command):
+        service_id: int = command['number']
+        print(f"Restart service {service_id}")
+
+    def _handle_shutdown_service(self, command):
+        service_id: int = command['number']
+        print(f"Shutdown service {service_id}")
+
+    def _handle_startup_service(self, command):
+        service_id: int = command['number']
+        print(f"Startup service {service_id}")
+
+    def _handle_list_users(self, command):
+        print("Listing all users")
+
+        url = f'http://{self.host}:{self.port}/api/v1/admin/users'
+        response = requests.get(url, auth=HTTPBasicAuth(self.admin_account, self.admin_password))
+        res_json = dict
+        if response.status_code == 200:
+            res_json = response.json()
+            self._print_table_simple(res_json['data'])
+        else:
+            print(f"Fail to get all users, code: {res_json['code']}, message: {res_json['message']}")
+
+    def _handle_show_user(self, command):
+        username_tree: Tree = command['username']
+        username: str = username_tree.children[0].strip("'\"")
+        print(f"Showing user: {username}")
+
+    def _handle_drop_user(self, command):
+        username_tree: Tree = command['username']
+        username: str = username_tree.children[0].strip("'\"")
+        print(f"Drop user: {username}")
+
+    def _handle_alter_user(self, command):
+        username_tree: Tree = command['username']
+        username: str = username_tree.children[0].strip("'\"")
+        password_tree: Tree = command['password']
+        password: str = password_tree.children[0].strip("'\"")
+        print(f"Alter user: {username}, password: {password}")
+
+    def _handle_list_datasets(self, command):
+        username_tree: Tree = command['username']
+        username: str = username_tree.children[0].strip("'\"")
+        print(f"Listing all datasets of user: {username}")
+
+    def _handle_list_agents(self, command):
+        username_tree: Tree = command['username']
+        username: str = username_tree.children[0].strip("'\"")
+        print(f"Listing all agents of user: {username}")
+
+    def _handle_meta_command(self, command):
+        meta_command = command['command']
+        args = command.get('args', [])
+
+        if meta_command in ['?', 'h', 'help']:
+            self.show_help()
+        elif meta_command in ['q', 'quit', 'exit']:
+            print("Goodbye!")
+        else:
+            print(f"Meta command '{meta_command}' with args {args}")
+
+    def show_help(self):
+        """Help info"""
+        help_text = """
+Commands:
+  LIST SERVICES
+  SHOW SERVICE <service>
+  STARTUP SERVICE <service>
+  SHUTDOWN SERVICE <service>
+  RESTART SERVICE <service>
+  LIST USERS
+  SHOW USER <user>
+  DROP USER <user>
+  CREATE USER <user> <password>
+  ALTER USER PASSWORD <user> <new_password>
+  LIST DATASETS OF <user>
+  LIST AGENTS OF <user>
+
+Meta Commands:
+  \\?, \\h, \\help     Show this help
+  \\q, \\quit, \\exit   Quit the CLI
+        """
+        print(help_text)
+
+
+def main():
+    import sys
+
+    cli = AdminCLI()
+
+    if len(sys.argv) == 1 or (len(sys.argv) > 1 and sys.argv[1] == '-'):
+        print(r"""
+            ____  ___   ______________                 ___       __          _     
+           / __ \/   | / ____/ ____/ /___ _      __   /   | ____/ /___ ___  (_)___ 
+          / /_/ / /| |/ / __/ /_  / / __ \ | /| / /  / /| |/ __  / __ `__ \/ / __ \
+         / _, _/ ___ / /_/ / __/ / / /_/ / |/ |/ /  / ___ / /_/ / / / / / / / / / /
+        /_/ |_/_/  |_\____/_/   /_/\____/|__/|__/  /_/  |_\__,_/_/ /_/ /_/_/_/ /_/ 
+        """)
+        if cli.verify_admin(sys.argv):
+            cli.run_interactive()
+    else:
+        if cli.verify_admin(sys.argv):
+            cli.run_interactive()
+            # cli.run_single_command(sys.argv[1:])
+
+
+if __name__ == '__main__':
+    main()

admin/admin_server.py

@@ -0,0 +1,46 @@
+
+import os
+import signal
+import logging
+import time
+import threading
+import traceback
+from werkzeug.serving import run_simple
+from flask import Flask
+from routes import admin_bp
+from api.utils.log_utils import init_root_logger
+from api.constants import SERVICE_CONF
+from config import load_configurations, SERVICE_CONFIGS
+
+stop_event = threading.Event()
+
+if __name__ == '__main__':
+    init_root_logger("admin_service")
+    logging.info(r"""
+        ____  ___   ______________                 ___       __          _     
+       / __ \/   | / ____/ ____/ /___ _      __   /   | ____/ /___ ___  (_)___ 
+      / /_/ / /| |/ / __/ /_  / / __ \ | /| / /  / /| |/ __  / __ `__ \/ / __ \
+     / _, _/ ___ / /_/ / __/ / / /_/ / |/ |/ /  / ___ / /_/ / / / / / / / / / /
+    /_/ |_/_/  |_\____/_/   /_/\____/|__/|__/  /_/  |_\__,_/_/ /_/ /_/_/_/ /_/ 
+    """)
+
+    app = Flask(__name__)
+    app.register_blueprint(admin_bp)
+
+    SERVICE_CONFIGS.configs = load_configurations(SERVICE_CONF)
+
+    try:
+        logging.info("RAGFlow Admin service start...")
+        run_simple(
+            hostname="0.0.0.0",
+            port=9381,
+            application=app,
+            threaded=True,
+            use_reloader=True,
+            use_debugger=True,
+        )
+    except Exception:
+        traceback.print_exc()
+        stop_event.set()
+        time.sleep(1)
+        os.kill(os.getpid(), signal.SIGKILL)

admin/auth.py

@@ -0,0 +1,57 @@
+import logging
+import uuid
+from functools import wraps
+from flask import request, jsonify
+
+from exceptions import AdminException
+from api.db.init_data import encode_to_base64
+from api.db.services import UserService
+
+
+def check_admin(username: str, password: str):
+    users = UserService.query(email=username)
+    if not users:
+        logging.info(f"Username: {username} is not registered!")
+        user_info = {
+            "id": uuid.uuid1().hex,
+            "password": encode_to_base64("admin"),
+            "nickname": "admin",
+            "is_superuser": True,
+            "email": "admin@ragflow.io",
+            "creator": "system",
+            "status": "1",
+        }
+        if not UserService.save(**user_info):
+            raise AdminException("Can't init admin.", 500)
+
+    user = UserService.query_user(username, password)
+    if user:
+        return True
+    else:
+        return False
+
+
+def login_verify(f):
+    @wraps(f)
+    def decorated(*args, **kwargs):
+        auth = request.authorization
+        if not auth or 'username' not in auth.parameters or 'password' not in auth.parameters:
+            return jsonify({
+                "code": 401,
+                "message": "Authentication required",
+                "data": None
+            }), 200
+
+        username = auth.parameters['username']
+        password = auth.parameters['password']
+        # TODO: to check the username and password from DB
+        if check_admin(username, password) is False:
+            return jsonify({
+                "code": 403,
+                "message": "Access denied",
+                "data": None
+            }), 200
+
+        return f(*args, **kwargs)
+
+    return decorated

admin/config.py

@@ -0,0 +1,280 @@
+import logging
+import threading
+from enum import Enum
+
+from pydantic import BaseModel
+from typing import Any
+from api.utils import read_config
+from urllib.parse import urlparse
+
+
+class ServiceConfigs:
+    def __init__(self):
+        self.configs = []
+        self.lock = threading.Lock()
+
+
+SERVICE_CONFIGS = ServiceConfigs
+
+
+class ServiceType(Enum):
+    METADATA = "metadata"
+    RETRIEVAL = "retrieval"
+    MESSAGE_QUEUE = "message_queue"
+    RAGFLOW_SERVER = "ragflow_server"
+    TASK_EXECUTOR = "task_executor"
+    FILE_STORE = "file_store"
+
+
+class BaseConfig(BaseModel):
+    id: int
+    name: str
+    host: str
+    port: int
+    service_type: str
+
+    def to_dict(self) -> dict[str, Any]:
+        return {'id': self.id, 'name': self.name, 'host': self.host, 'port': self.port, 'service_type': self.service_type}
+
+
+class MetaConfig(BaseConfig):
+    meta_type: str
+
+    def to_dict(self) -> dict[str, Any]:
+        result = super().to_dict()
+        if 'extra' not in result:
+            result['extra'] = dict()
+        extra_dict = result['extra'].copy()
+        extra_dict['meta_type'] = self.meta_type
+        result['extra'] = extra_dict
+        return result
+
+
+class MySQLConfig(MetaConfig):
+    username: str
+    password: str
+
+    def to_dict(self) -> dict[str, Any]:
+        result = super().to_dict()
+        if 'extra' not in result:
+            result['extra'] = dict()
+        extra_dict = result['extra'].copy()
+        extra_dict['username'] = self.username
+        extra_dict['password'] = self.password
+        result['extra'] = extra_dict
+        return result
+
+
+class PostgresConfig(MetaConfig):
+
+    def to_dict(self) -> dict[str, Any]:
+        result = super().to_dict()
+        if 'extra' not in result:
+            result['extra'] = dict()
+        return result
+
+
+class RetrievalConfig(BaseConfig):
+    retrieval_type: str
+
+    def to_dict(self) -> dict[str, Any]:
+        result = super().to_dict()
+        if 'extra' not in result:
+            result['extra'] = dict()
+        extra_dict = result['extra'].copy()
+        extra_dict['retrieval_type'] = self.retrieval_type
+        result['extra'] = extra_dict
+        return result
+
+
+class InfinityConfig(RetrievalConfig):
+    db_name: str
+
+    def to_dict(self) -> dict[str, Any]:
+        result = super().to_dict()
+        if 'extra' not in result:
+            result['extra'] = dict()
+        extra_dict = result['extra'].copy()
+        extra_dict['db_name'] = self.db_name
+        result['extra'] = extra_dict
+        return result
+
+
+class ElasticsearchConfig(RetrievalConfig):
+    username: str
+    password: str
+
+    def to_dict(self) -> dict[str, Any]:
+        result = super().to_dict()
+        if 'extra' not in result:
+            result['extra'] = dict()
+        extra_dict = result['extra'].copy()
+        extra_dict['username'] = self.username
+        extra_dict['password'] = self.password
+        result['extra'] = extra_dict
+        return result
+
+
+class MessageQueueConfig(BaseConfig):
+    mq_type: str
+
+    def to_dict(self) -> dict[str, Any]:
+        result = super().to_dict()
+        if 'extra' not in result:
+            result['extra'] = dict()
+        extra_dict = result['extra'].copy()
+        extra_dict['mq_type'] = self.mq_type
+        result['extra'] = extra_dict
+        return result
+
+
+class RedisConfig(MessageQueueConfig):
+    database: int
+    password: str
+
+    def to_dict(self) -> dict[str, Any]:
+        result = super().to_dict()
+        if 'extra' not in result:
+            result['extra'] = dict()
+        extra_dict = result['extra'].copy()
+        extra_dict['database'] = self.database
+        extra_dict['password'] = self.password
+        result['extra'] = extra_dict
+        return result
+
+
+class RabbitMQConfig(MessageQueueConfig):
+
+    def to_dict(self) -> dict[str, Any]:
+        result = super().to_dict()
+        if 'extra' not in result:
+            result['extra'] = dict()
+        return result
+
+
+class RAGFlowServerConfig(BaseConfig):
+
+    def to_dict(self) -> dict[str, Any]:
+        result = super().to_dict()
+        if 'extra' not in result:
+            result['extra'] = dict()
+        return result
+
+
+class TaskExecutorConfig(BaseConfig):
+
+    def to_dict(self) -> dict[str, Any]:
+        result = super().to_dict()
+        if 'extra' not in result:
+            result['extra'] = dict()
+        return result
+
+
+class FileStoreConfig(BaseConfig):
+    store_type: str
+
+    def to_dict(self) -> dict[str, Any]:
+        result = super().to_dict()
+        if 'extra' not in result:
+            result['extra'] = dict()
+        extra_dict = result['extra'].copy()
+        extra_dict['store_type'] = self.store_type
+        result['extra'] = extra_dict
+        return result
+
+
+class MinioConfig(FileStoreConfig):
+    user: str
+    password: str
+
+    def to_dict(self) -> dict[str, Any]:
+        result = super().to_dict()
+        if 'extra' not in result:
+            result['extra'] = dict()
+        extra_dict = result['extra'].copy()
+        extra_dict['user'] = self.user
+        extra_dict['password'] = self.password
+        result['extra'] = extra_dict
+        return result
+
+
+def load_configurations(config_path: str) -> list[BaseConfig]:
+    raw_configs = read_config(config_path)
+    configurations = []
+    ragflow_count = 0
+    id_count = 0
+    for k, v in raw_configs.items():
+        match (k):
+            case "ragflow":
+                name: str = f'ragflow_{ragflow_count}'
+                host: str = v['host']
+                http_port: int = v['http_port']
+                config = RAGFlowServerConfig(id=id_count, name=name, host=host, port=http_port, service_type="ragflow_server")
+                configurations.append(config)
+                id_count += 1
+            case "es":
+                name: str = 'elasticsearch'
+                url = v['hosts']
+                parsed = urlparse(url)
+                host: str = parsed.hostname
+                port: int = parsed.port
+                username: str = v.get('username')
+                password: str = v.get('password')
+                config = ElasticsearchConfig(id=id_count, name=name, host=host, port=port, service_type="retrieval",
+                                             retrieval_type="elasticsearch",
+                                             username=username, password=password)
+                configurations.append(config)
+                id_count += 1
+
+            case "infinity":
+                name: str = 'infinity'
+                url = v['uri']
+                parts = url.split(':', 1)
+                host = parts[0]
+                port = int(parts[1])
+                database: str = v.get('db_name', 'default_db')
+                config = InfinityConfig(id=id_count, name=name, host=host, port=port, service_type="retrieval", retrieval_type="infinity",
+                                        db_name=database)
+                configurations.append(config)
+                id_count += 1
+            case "minio":
+                name: str = 'minio'
+                url = v['host']
+                parts = url.split(':', 1)
+                host = parts[0]
+                port = int(parts[1])
+                user = v.get('user')
+                password = v.get('password')
+                config = MinioConfig(id=id_count, name=name, host=host, port=port, user=user, password=password, service_type="file_store",
+                                     store_type="minio")
+                configurations.append(config)
+                id_count += 1
+            case "redis":
+                name: str = 'redis'
+                url = v['host']
+                parts = url.split(':', 1)
+                host = parts[0]
+                port = int(parts[1])
+                password = v.get('password')
+                db: int = v.get('db')
+                config = RedisConfig(id=id_count, name=name, host=host, port=port, password=password, database=db,
+                                     service_type="message_queue", mq_type="redis")
+                configurations.append(config)
+                id_count += 1
+            case "mysql":
+                name: str = 'mysql'
+                host: str = v.get('host')
+                port: int = v.get('port')
+                username = v.get('user')
+                password = v.get('password')
+                config = MySQLConfig(id=id_count, name=name, host=host, port=port, username=username, password=password,
+                                     service_type="meta_data", meta_type="mysql")
+                configurations.append(config)
+                id_count += 1
+            case "admin":
+                pass
+            case _:
+                logging.warning(f"Unknown configuration key: {k}")
+                continue
+
+    return configurations

admin/exceptions.py

@@ -0,0 +1,17 @@
+class AdminException(Exception):
+    def __init__(self, message, code=400):
+        super().__init__(message)
+        self.code = code
+        self.message = message
+
+class UserNotFoundError(AdminException):
+    def __init__(self, username):
+        super().__init__(f"User '{username}' not found", 404)
+
+class UserAlreadyExistsError(AdminException):
+    def __init__(self, username):
+        super().__init__(f"User '{username}' already exists", 409)
+
+class CannotDeleteAdminError(AdminException):
+    def __init__(self):
+        super().__init__("Cannot delete admin account", 403)

admin/responses.py

@@ -0,0 +1,15 @@
+from flask import jsonify
+
+def success_response(data=None, message="Success", code = 0):
+    return jsonify({
+        "code": code,
+        "message": message,
+        "data": data
+    }), 200
+
+def error_response(message="Error", code=-1, data=None):
+    return jsonify({
+        "code": code,
+        "message": message,
+        "data": data
+    }), 400

admin/routes.py

@@ -0,0 +1,141 @@
+from flask import Blueprint, request
+from auth import login_verify
+from responses import success_response, error_response
+from services import UserMgr, ServiceMgr
+from exceptions import AdminException
+
+admin_bp = Blueprint('admin', __name__, url_prefix='/api/v1/admin')
+
+
+@admin_bp.route('/auth', methods=['GET'])
+@login_verify
+def auth_admin():
+    try:
+        return success_response(None, "Admin is authorized", 0)
+    except Exception as e:
+        return error_response(str(e), 500)
+
+
+@admin_bp.route('/users', methods=['GET'])
+@login_verify
+def list_users():
+    try:
+        users = UserMgr.get_all_users()
+        return success_response(users, "Get all users", 0)
+    except Exception as e:
+        return error_response(str(e), 500)
+
+
+@admin_bp.route('/users', methods=['POST'])
+@login_verify
+def create_user():
+    try:
+        data = request.get_json()
+        if not data or 'username' not in data or 'password' not in data:
+            return error_response("Username and password are required", 400)
+
+        username = data['username']
+        password = data['password']
+        role = data.get('role', 'user')
+
+        user = UserMgr.create_user(username, password, role)
+        return success_response(user, "User created successfully", 201)
+
+    except AdminException as e:
+        return error_response(e.message, e.code)
+    except Exception as e:
+        return error_response(str(e), 500)
+
+
+@admin_bp.route('/users/<username>', methods=['DELETE'])
+@login_verify
+def delete_user(username):
+    try:
+        UserMgr.delete_user(username)
+        return success_response(None, "User and all data deleted successfully")
+
+    except AdminException as e:
+        return error_response(e.message, e.code)
+    except Exception as e:
+        return error_response(str(e), 500)
+
+
+@admin_bp.route('/users/<username>/password', methods=['PUT'])
+@login_verify
+def change_password(username):
+    try:
+        data = request.get_json()
+        if not data or 'new_password' not in data:
+            return error_response("New password is required", 400)
+
+        new_password = data['new_password']
+        UserMgr.update_user_password(username, new_password)
+        return success_response(None, "Password updated successfully")
+
+    except AdminException as e:
+        return error_response(e.message, e.code)
+    except Exception as e:
+        return error_response(str(e), 500)
+
+
+@admin_bp.route('/users/<username>', methods=['GET'])
+@login_verify
+def get_user_details(username):
+    try:
+        user_details = UserMgr.get_user_details(username)
+        return success_response(user_details)
+
+    except AdminException as e:
+        return error_response(e.message, e.code)
+    except Exception as e:
+        return error_response(str(e), 500)
+
+
+@admin_bp.route('/services', methods=['GET'])
+@login_verify
+def get_services():
+    try:
+        services = ServiceMgr.get_all_services()
+        return success_response(services, "Get all services", 0)
+    except Exception as e:
+        return error_response(str(e), 500)
+
+
+@admin_bp.route('/service_types/<service_type>', methods=['GET'])
+@login_verify
+def get_services_by_type(service_type_str):
+    try:
+        services = ServiceMgr.get_services_by_type(service_type_str)
+        return success_response(services)
+    except Exception as e:
+        return error_response(str(e), 500)
+
+
+@admin_bp.route('/services/<service_id>', methods=['GET'])
+@login_verify
+def get_service(service_id):
+    try:
+        services = ServiceMgr.get_service_details(service_id)
+        return success_response(services)
+    except Exception as e:
+        return error_response(str(e), 500)
+
+
+@admin_bp.route('/services/<service_id>', methods=['DELETE'])
+@login_verify
+def shutdown_service(service_id):
+    try:
+        services = ServiceMgr.shutdown_service(service_id)
+        return success_response(services)
+    except Exception as e:
+        return error_response(str(e), 500)
+
+
+@admin_bp.route('/services/<service_id>', methods=['PUT'])
+@login_verify
+def restart_service(service_id):
+    try:
+        services = ServiceMgr.restart_service(service_id)
+        return success_response(services)
+    except Exception as e:
+        return error_response(str(e), 500)

admin/services.py

@@ -0,0 +1,54 @@
+from api.db.services import UserService
+from exceptions import AdminException
+from config import SERVICE_CONFIGS
+
+class UserMgr:
+    @staticmethod
+    def get_all_users():
+        users = UserService.get_all_users()
+        result = []
+        for user in users:
+            result.append({'email': user.email, 'nickname': user.nickname, 'create_date': user.create_date, 'is_active': user.is_active})
+        return result
+
+    @staticmethod
+    def get_user_details(username):
+        raise AdminException("get_user_details: not implemented")
+
+    @staticmethod
+    def create_user(username, password, role="user"):
+        raise AdminException("create_user: not implemented")
+
+    @staticmethod
+    def delete_user(username):
+        raise AdminException("delete_user: not implemented")
+
+    @staticmethod
+    def update_user_password(username, new_password):
+        raise AdminException("update_user_password: not implemented")
+
+class ServiceMgr:
+
+    @staticmethod
+    def get_all_services():
+        result = []
+        configs = SERVICE_CONFIGS.configs
+        for config in configs:
+            result.append(config.to_dict())
+        return result
+
+    @staticmethod
+    def get_services_by_type(service_type_str: str):
+        raise AdminException("get_services_by_type: not implemented")
+
+    @staticmethod
+    def get_service_details(service_id: int):
+        raise AdminException("get_service_details: not implemented")
+
+    @staticmethod
+    def shutdown_service(service_id: int):
+        raise AdminException("shutdown_service: not implemented")
+
+    @staticmethod
+    def restart_service(service_id: int):
+        raise AdminException("restart_service: not implemented")

agent/canvas.py

@@ -27,7 +27,7 @@ from agent.component import component_class
 from agent.component.base import ComponentBase
 from api.db.services.file_service import FileService
 from api.utils import get_uuid, hash_str2int
-from rag.prompts.prompts import chunks_format
+from rag.prompts.generator import chunks_format
 from rag.utils.redis_conn import REDIS_CONN
 
 class Graph:
@@ -490,7 +490,8 @@ class Canvas(Graph):
 
         r = self.retrieval[-1]
         for ck in chunks_format({"chunks": chunks}):
-            cid = hash_str2int(ck["id"], 100)
+            cid = hash_str2int(ck["id"], 500)
+            # cid = uuid.uuid5(uuid.NAMESPACE_DNS, ck["id"])
             if cid not in r:
                 r["chunks"][cid] = ck
 

agent/component/agent_with_tools.py

@@ -28,9 +28,8 @@ from api.db.services.llm_service import LLMBundle
 from api.db.services.tenant_llm_service import TenantLLMService
 from api.db.services.mcp_server_service import MCPServerService
 from api.utils.api_utils import timeout
-from rag.prompts import message_fit_in
-from rag.prompts.prompts import next_step, COMPLETE_TASK, analyze_task, \
-    citation_prompt, reflect, rank_memories, kb_prompt, citation_plus, full_question
+from rag.prompts.generator import next_step, COMPLETE_TASK, analyze_task, \
+    citation_prompt, reflect, rank_memories, kb_prompt, citation_plus, full_question, message_fit_in
 from rag.utils.mcp_tool_call_conn import MCPToolCallSession, mcp_tool_metadata_to_openai_tool
 from agent.component.llm import LLMParam, LLM
 

agent/component/base.py

@@ -244,7 +244,7 @@ class ComponentParamBase(ABC):
 
                 if not value_legal:
                     raise ValueError(
-                        "Plase check runtime conf, {} = {} does not match user-parameter restriction".format(
+                        "Please check runtime conf, {} = {} does not match user-parameter restriction".format(
                             variable, value
                         )
                     )

agent/component/categorize.py

@@ -28,7 +28,7 @@ from rag.llm.chat_model import ERROR_PREFIX
 class CategorizeParam(LLMParam):
 
     """
-    Define the Categorize component parameters.
+    Define the categorize component parameters.
     """
     def __init__(self):
         super().__init__()

agent/component/llm.py

@@ -26,8 +26,7 @@ from api.db.services.llm_service import LLMBundle
 from api.db.services.tenant_llm_service import TenantLLMService
 from agent.component.base import ComponentBase, ComponentParamBase
 from api.utils.api_utils import timeout
-from rag.prompts import message_fit_in, citation_prompt
-from rag.prompts.prompts import tool_call_summary
+from rag.prompts.generator import tool_call_summary, message_fit_in, citation_prompt
 
 
 class LLMParam(ComponentParamBase):
@@ -82,9 +81,9 @@ class LLMParam(ComponentParamBase):
 
 class LLM(ComponentBase):
     component_name = "LLM"
-    
-    def __init__(self, canvas, id, param: ComponentParamBase):
-        super().__init__(canvas, id, param)
+
+    def __init__(self, canvas, component_id, param: ComponentParamBase):
+        super().__init__(canvas, component_id, param)
         self.chat_mdl = LLMBundle(self._canvas.get_tenant_id(), TenantLLMService.llm_id2llm_type(self._param.llm_id),
                                   self._param.llm_id, max_retries=self._param.max_retries,
                                   retry_interval=self._param.delay_after_error
@@ -210,7 +209,7 @@ class LLM(ComponentBase):
             return re.sub(r"```\n*$", "", ans, flags=re.DOTALL)
 
         prompt, msg, _ = self._prepare_prompt_variables()
-        error = ""
+        error: str = ""
 
         if self._param.output_structure:
             prompt += "\nThe output MUST follow this JSON format:\n"+json.dumps(self._param.output_structure, ensure_ascii=False, indent=2)

agent/component/message.py

@@ -49,7 +49,7 @@ class MessageParam(ComponentParamBase):
 class Message(ComponentBase):
     component_name = "Message"
 
-    def get_kwargs(self, script:str, kwargs:dict = {}, delimeter:str=None) -> tuple[str, dict[str, str | list | Any]]:
+    def get_kwargs(self, script:str, kwargs:dict = {}, delimiter:str=None) -> tuple[str, dict[str, str | list | Any]]:
         for k,v in self.get_input_elements_from_text(script).items():
             if k in kwargs:
                 continue
@@ -60,8 +60,8 @@ class Message(ComponentBase):
             if isinstance(v, partial):
                 for t in v():
                     ans += t
-            elif isinstance(v, list) and delimeter:
-                ans = delimeter.join([str(vv) for vv in v])
+            elif isinstance(v, list) and delimiter:
+                ans = delimiter.join([str(vv) for vv in v])
             elif not isinstance(v, str):
                 try:
                     ans = json.dumps(v, ensure_ascii=False)

agent/component/string_transform.py

@@ -90,7 +90,7 @@ class StringTransform(Message, ABC):
         for k,v in kwargs.items():
             if not v:
                 v = ""
-            script = re.sub(k, v, script)
+            script = re.sub(k, lambda match: v, script)
 
         self.set_output("result", script)
 

agent/templates/sql_assistant.json

@@ -83,7 +83,7 @@
                             },
                             "password": "20010812Yy!",
                             "port": 3306,
-                            "sql": "Agent:WickedGoatsDivide@content",
+                            "sql": "{Agent:WickedGoatsDivide@content}",
                             "username": "13637682833@163.com"
                         }
                     },
@@ -114,9 +114,7 @@
                         "params": {
                             "cross_languages": [],
                             "empty_response": "",
-                            "kb_ids": [
-                                "ed31364c727211f0bdb2bafe6e7908e6"
-                            ],
+                            "kb_ids": [],
                             "keywords_similarity_weight": 0.7,
                             "outputs": {
                                 "formalized_content": {
@@ -124,7 +122,7 @@
                                     "value": ""
                                 }
                             },
-                            "query": "sys.query",
+                            "query": "{sys.query}",
                             "rerank_id": "",
                             "similarity_threshold": 0.2,
                             "top_k": 1024,
@@ -145,9 +143,7 @@
                         "params": {
                             "cross_languages": [],
                             "empty_response": "",
-                            "kb_ids": [
-                                "0f968106727311f08357bafe6e7908e6"
-                            ],
+                            "kb_ids": [],
                             "keywords_similarity_weight": 0.7,
                             "outputs": {
                                 "formalized_content": {
@@ -155,7 +151,7 @@
                                     "value": ""
                                 }
                             },
-                            "query": "sys.query",
+                            "query": "{sys.query}",
                             "rerank_id": "",
                             "similarity_threshold": 0.2,
                             "top_k": 1024,
@@ -176,9 +172,7 @@
                         "params": {
                             "cross_languages": [],
                             "empty_response": "",
-                            "kb_ids": [
-                                "4ad1f9d0727311f0827dbafe6e7908e6"
-                            ],
+                            "kb_ids": [],
                             "keywords_similarity_weight": 0.7,
                             "outputs": {
                                 "formalized_content": {
@@ -186,7 +180,7 @@
                                     "value": ""
                                 }
                             },
-                            "query": "sys.query",
+                            "query": "{sys.query}",
                             "rerank_id": "",
                             "similarity_threshold": 0.2,
                             "top_k": 1024,
@@ -347,9 +341,7 @@
                             "form": {
                                 "cross_languages": [],
                                 "empty_response": "",
-                                "kb_ids": [
-                                    "ed31364c727211f0bdb2bafe6e7908e6"
-                                ],
+                                "kb_ids": [],
                                 "keywords_similarity_weight": 0.7,
                                 "outputs": {
                                     "formalized_content": {
@@ -357,7 +349,7 @@
                                         "value": ""
                                     }
                                 },
-                                "query": "sys.query",
+                                "query": "{sys.query}",
                                 "rerank_id": "",
                                 "similarity_threshold": 0.2,
                                 "top_k": 1024,
@@ -387,9 +379,7 @@
                             "form": {
                                 "cross_languages": [],
                                 "empty_response": "",
-                                "kb_ids": [
-                                    "0f968106727311f08357bafe6e7908e6"
-                                ],
+                                "kb_ids": [],
                                 "keywords_similarity_weight": 0.7,
                                 "outputs": {
                                     "formalized_content": {
@@ -397,7 +387,7 @@
                                         "value": ""
                                     }
                                 },
-                                "query": "sys.query",
+                                "query": "{sys.query}",
                                 "rerank_id": "",
                                 "similarity_threshold": 0.2,
                                 "top_k": 1024,
@@ -427,9 +417,7 @@
                             "form": {
                                 "cross_languages": [],
                                 "empty_response": "",
-                                "kb_ids": [
-                                    "4ad1f9d0727311f0827dbafe6e7908e6"
-                                ],
+                                "kb_ids": [],
                                 "keywords_similarity_weight": 0.7,
                                 "outputs": {
                                     "formalized_content": {
@@ -437,7 +425,7 @@
                                         "value": ""
                                     }
                                 },
-                                "query": "sys.query",
+                                "query": "{sys.query}",
                                 "rerank_id": "",
                                 "similarity_threshold": 0.2,
                                 "top_k": 1024,
@@ -539,7 +527,7 @@
                                 },
                                 "password": "20010812Yy!",
                                 "port": 3306,
-                                "sql": "Agent:WickedGoatsDivide@content",
+                                "sql": "{Agent:WickedGoatsDivide@content}",
                                 "username": "13637682833@163.com"
                             },
                             "label": "ExeSQL",

agent/tools/base.py

@@ -22,7 +22,7 @@ from typing import TypedDict, List, Any
 from agent.component.base import ComponentParamBase, ComponentBase
 from api.utils import hash_str2int
 from rag.llm.chat_model import ToolCallSession
-from rag.prompts.prompts import kb_prompt
+from rag.prompts.generator import kb_prompt
 from rag.utils.mcp_tool_call_conn import MCPToolCallSession
 from timeit import default_timer as timer
 

agent/tools/code_exec.py

@@ -157,7 +157,7 @@ class CodeExec(ToolBase, ABC):
 
         try:
             resp = requests.post(url=f"http://{settings.SANDBOX_HOST}:9385/run", json=code_req, timeout=os.environ.get("COMPONENT_EXEC_TIMEOUT", 10*60))
-            logging.info(f"http://{settings.SANDBOX_HOST}:9385/run", code_req, resp.status_code)
+            logging.info(f"http://{settings.SANDBOX_HOST}:9385/run,  code_req: {code_req}, resp.status_code {resp.status_code}:")
             if resp.status_code != 200:
                 resp.raise_for_status()
             body = resp.json()

agent/tools/exesql.py

@@ -53,7 +53,7 @@ class ExeSQLParam(ToolParamBase):
         self.max_records = 1024
 
     def check(self):
-        self.check_valid_value(self.db_type, "Choose DB type", ['mysql', 'postgresql', 'mariadb', 'mssql'])
+        self.check_valid_value(self.db_type, "Choose DB type", ['mysql', 'postgres', 'mariadb', 'mssql'])
         self.check_empty(self.database, "Database name")
         self.check_empty(self.username, "database username")
         self.check_empty(self.host, "IP Address")
@@ -111,7 +111,7 @@ class ExeSQL(ToolBase, ABC):
         if self._param.db_type in ["mysql", "mariadb"]:
             db = pymysql.connect(db=self._param.database, user=self._param.username, host=self._param.host,
                                  port=self._param.port, password=self._param.password)
-        elif self._param.db_type == 'postgresql':
+        elif self._param.db_type == 'postgres':
             db = psycopg2.connect(dbname=self._param.database, user=self._param.username, host=self._param.host,
                                   port=self._param.port, password=self._param.password)
         elif self._param.db_type == 'mssql':

agent/tools/retrieval.py

@@ -23,8 +23,7 @@ from api.db.services.llm_service import LLMBundle
 from api import settings
 from api.utils.api_utils import timeout
 from rag.app.tag import label_question
-from rag.prompts import kb_prompt
-from rag.prompts.prompts import cross_languages
+from rag.prompts.generator import cross_languages, kb_prompt
 
 
 class RetrievalParam(ToolParamBase):
@@ -163,9 +162,16 @@ class Retrieval(ToolBase, ABC):
             self.set_output("formalized_content", self._param.empty_response)
             return
 
+        # Format the chunks for JSON output (similar to how other tools do it)
+        json_output = kbinfos["chunks"].copy()
+        
         self._canvas.add_reference(kbinfos["chunks"], kbinfos["doc_aggs"])
         form_cnt = "\n".join(kb_prompt(kbinfos, 200000, True))
+        
+        # Set both formalized content and JSON output
         self.set_output("formalized_content", form_cnt)
+        self.set_output("json", json_output)
+        
         return form_cnt
 
     def thoughts(self) -> str:

api/apps/api_app.py

@@ -39,7 +39,7 @@ from api.utils.api_utils import server_error_response, get_data_error_result, ge
 
 from api.utils.file_utils import filename_type, thumbnail
 from rag.app.tag import label_question
-from rag.prompts import keyword_extraction
+from rag.prompts.generator import keyword_extraction
 from rag.utils.storage_factory import STORAGE_IMPL
 
 from api.db.services.canvas_service import UserCanvasService

api/apps/canvas_app.py

@@ -23,7 +23,7 @@ import trio
 from flask import request, Response
 from flask_login import login_required, current_user
 
-from agent.component import LLM
+from agent.component.llm import LLM
 from api.db import CanvasCategory, FileType
 from api.db.services.canvas_service import CanvasTemplateService, UserCanvasService, API4ConversationService
 from api.db.services.document_service import DocumentService
@@ -332,7 +332,7 @@ def test_db_connect():
         if req["db_type"] in ["mysql", "mariadb"]:
             db = MySQLDatabase(req["database"], user=req["username"], host=req["host"], port=req["port"],
                                password=req["password"])
-        elif req["db_type"] == 'postgresql':
+        elif req["db_type"] == 'postgres':
             db = PostgresqlDatabase(req["database"], user=req["username"], host=req["host"], port=req["port"],
                                     password=req["password"])
         elif req["db_type"] == 'mssql':
@@ -474,7 +474,7 @@ def sessions(canvas_id):
 @manager.route('/prompts', methods=['GET'])  # noqa: F821
 @login_required
 def prompts():
-    from rag.prompts.prompts import ANALYZE_TASK_SYSTEM, ANALYZE_TASK_USER, NEXT_STEP, REFLECT, CITATION_PROMPT_TEMPLATE
+    from rag.prompts.generator import ANALYZE_TASK_SYSTEM, ANALYZE_TASK_USER, NEXT_STEP, REFLECT, CITATION_PROMPT_TEMPLATE
     return get_json_result(data={
         "task_analysis": ANALYZE_TASK_SYSTEM +"\n\n"+ ANALYZE_TASK_USER,
         "plan_generation": NEXT_STEP,

api/apps/chunk_app.py

@@ -33,8 +33,7 @@ from api.utils.api_utils import get_data_error_result, get_json_result, server_e
 from rag.app.qa import beAdoc, rmPrefix
 from rag.app.tag import label_question
 from rag.nlp import rag_tokenizer, search
-from rag.prompts import cross_languages, keyword_extraction
-from rag.prompts.prompts import gen_meta_filter
+from rag.prompts.generator import gen_meta_filter, cross_languages, keyword_extraction
 from rag.settings import PAGERANK_FLD
 from rag.utils import rmSpace
 

api/apps/conversation_app.py

@@ -15,7 +15,7 @@
 #
 import json
 import re
-import traceback
+import logging
 from copy import deepcopy
 from flask import Response, request
 from flask_login import current_user, login_required
@@ -29,8 +29,8 @@ from api.db.services.search_service import SearchService
 from api.db.services.tenant_llm_service import TenantLLMService
 from api.db.services.user_service import TenantService, UserTenantService
 from api.utils.api_utils import get_data_error_result, get_json_result, server_error_response, validate_request
-from rag.prompts.prompt_template import load_prompt
-from rag.prompts.prompts import chunks_format
+from rag.prompts.template import load_prompt
+from rag.prompts.generator import chunks_format
 
 
 @manager.route("/set", methods=["POST"])  # noqa: F821
@@ -226,7 +226,7 @@ def completion():
                 if not is_embedded:
                     ConversationService.update_by_id(conv.id, conv.to_dict())
             except Exception as e:
-                traceback.print_exc()
+                logging.exception(e)
                 yield "data:" + json.dumps({"code": 500, "message": str(e), "data": {"answer": "**ERROR**: " + str(e), "reference": []}}, ensure_ascii=False) + "\n\n"
             yield "data:" + json.dumps({"code": 0, "message": "", "data": True}, ensure_ascii=False) + "\n\n"
 

api/apps/dataflow_app.py

@@ -24,7 +24,7 @@ from flask import request
 from flask_login import current_user, login_required
 
 from agent.canvas import Canvas
-from agent.component import LLM
+from agent.component.llm import LLM
 from api.db import CanvasCategory, FileType
 from api.db.services.canvas_service import CanvasTemplateService, UserCanvasService
 from api.db.services.document_service import DocumentService

api/apps/kb_app.py

@@ -379,3 +379,19 @@ def get_meta():
                 code=settings.RetCode.AUTHENTICATION_ERROR
             )
     return get_json_result(data=DocumentService.get_meta_by_kbs(kb_ids))
+
+
+@manager.route("/basic_info", methods=["GET"])  # noqa: F821
+@login_required
+def get_basic_info():
+    kb_id = request.args.get("kb_id", "")
+    if not KnowledgebaseService.accessible(kb_id, current_user.id):
+        return get_json_result(
+            data=False,
+            message='No authorization.',
+            code=settings.RetCode.AUTHENTICATION_ERROR
+        )
+
+    basic_info = DocumentService.knowledgebase_basic_info(kb_id)
+
+    return get_json_result(data=basic_info)

api/apps/sdk/doc.py

@@ -40,7 +40,7 @@ from api.utils.api_utils import check_duplicate_ids, construct_json_result, get_
 from rag.app.qa import beAdoc, rmPrefix
 from rag.app.tag import label_question
 from rag.nlp import rag_tokenizer, search
-from rag.prompts import cross_languages, keyword_extraction
+from rag.prompts.generator import cross_languages, keyword_extraction
 from rag.utils import rmSpace
 from rag.utils.storage_factory import STORAGE_IMPL
 

api/apps/sdk/files.py

@@ -3,9 +3,11 @@ import re
 
 import flask
 from flask import request
+from pathlib import Path
 
 from api.db.services.document_service import DocumentService
 from api.db.services.file2document_service import File2DocumentService
+from api.db.services.knowledgebase_service import KnowledgebaseService
 from api.utils.api_utils import server_error_response, token_required
 from api.utils import get_uuid
 from api.db import FileType
@@ -81,16 +83,16 @@ def upload(tenant_id):
             return get_json_result(data=False, message="Can't find this folder!", code=404)
 
         for file_obj in file_objs:
-            # 文件路径处理
+            # Handle file path
             full_path = '/' + file_obj.filename
             file_obj_names = full_path.split('/')
             file_len = len(file_obj_names)
 
-            # 获取文件夹路径ID
+            # Get folder path ID
             file_id_list = FileService.get_id_list_by_id(pf_id, file_obj_names, 1, [pf_id])
             len_id_list = len(file_id_list)
 
-            # 创建文件夹结构
+            # Crete file folder
             if file_len != len_id_list:
                 e, file = FileService.get_by_id(file_id_list[len_id_list - 1])
                 if not e:
@@ -666,3 +668,71 @@ def move(tenant_id):
         return get_json_result(data=True)
     except Exception as e:
         return server_error_response(e)
+
+@manager.route('/file/convert', methods=['POST'])  # noqa: F821
+@token_required
+def convert(tenant_id):
+    req = request.json
+    kb_ids = req["kb_ids"]
+    file_ids = req["file_ids"]
+    file2documents = []
+
+    try:
+        files = FileService.get_by_ids(file_ids)
+        files_set = dict({file.id: file for file in files})
+        for file_id in file_ids:
+            file = files_set[file_id]
+            if not file:
+                return get_json_result(message="File not found!", code=404)
+            file_ids_list = [file_id]
+            if file.type == FileType.FOLDER.value:
+                file_ids_list = FileService.get_all_innermost_file_ids(file_id, [])
+            for id in file_ids_list:
+                informs = File2DocumentService.get_by_file_id(id)
+                # delete
+                for inform in informs:
+                    doc_id = inform.document_id
+                    e, doc = DocumentService.get_by_id(doc_id)
+                    if not e:
+                        return get_json_result(message="Document not found!", code=404)
+                    tenant_id = DocumentService.get_tenant_id(doc_id)
+                    if not tenant_id:
+                        return get_json_result(message="Tenant not found!", code=404)
+                    if not DocumentService.remove_document(doc, tenant_id):
+                        return get_json_result(
+                            message="Database error (Document removal)!", code=404)
+                File2DocumentService.delete_by_file_id(id)
+
+                # insert
+                for kb_id in kb_ids:
+                    e, kb = KnowledgebaseService.get_by_id(kb_id)
+                    if not e:
+                        return get_json_result(
+                            message="Can't find this knowledgebase!", code=404)
+                    e, file = FileService.get_by_id(id)
+                    if not e:
+                        return get_json_result(
+                            message="Can't find this file!", code=404)
+
+                    doc = DocumentService.insert({
+                        "id": get_uuid(),
+                        "kb_id": kb.id,
+                        "parser_id": FileService.get_parser(file.type, file.name, kb.parser_id),
+                        "parser_config": kb.parser_config,
+                        "created_by": tenant_id,
+                        "type": file.type,
+                        "name": file.name,
+                        "suffix": Path(file.name).suffix.lstrip("."),
+                        "location": file.location,
+                        "size": file.size
+                    })
+                    file2document = File2DocumentService.insert({
+                        "id": get_uuid(),
+                        "file_id": id,
+                        "document_id": doc.id,
+                    })
+
+                    file2documents.append(file2document.to_json())
+        return get_json_result(data=file2documents)
+    except Exception as e:
+        return server_error_response(e)

api/apps/sdk/session.py

@@ -38,9 +38,8 @@ from api.db.services.user_service import UserTenantService
 from api.utils import get_uuid
 from api.utils.api_utils import check_duplicate_ids, get_data_openai, get_error_data_result, get_json_result, get_result, server_error_response, token_required, validate_request
 from rag.app.tag import label_question
-from rag.prompts import chunks_format
-from rag.prompts.prompt_template import load_prompt
-from rag.prompts.prompts import cross_languages, gen_meta_filter, keyword_extraction
+from rag.prompts.template import load_prompt
+from rag.prompts.generator import cross_languages, gen_meta_filter, keyword_extraction, chunks_format
 
 
 @manager.route("/chats/<chat_id>/sessions", methods=["POST"])  # noqa: F821
@@ -414,7 +413,7 @@ def agents_completion_openai_compatibility(tenant_id, agent_id):
                 tenant_id,
                 agent_id,
                 question,
-                session_id=req.get("session_id", req.get("id", "") or req.get("metadata", {}).get("id", "")),
+                session_id=req.pop("session_id", req.get("id", "")) or req.get("metadata", {}).get("id", ""),
                 stream=True,
                 **req,
             ),
@@ -432,7 +431,7 @@ def agents_completion_openai_compatibility(tenant_id, agent_id):
                 tenant_id,
                 agent_id,
                 question,
-                session_id=req.get("session_id", req.get("id", "") or req.get("metadata", {}).get("id", "")),
+                session_id=req.pop("session_id", req.get("id", "")) or req.get("metadata", {}).get("id", ""),
                 stream=False,
                 **req,
             )

api/apps/system_app.py

@@ -36,6 +36,8 @@ from rag.utils.storage_factory import STORAGE_IMPL, STORAGE_IMPL_TYPE
 from timeit import default_timer as timer
 
 from rag.utils.redis_conn import REDIS_CONN
+from flask import jsonify
+from api.utils.health_utils import run_health_checks
 
 @manager.route("/version", methods=["GET"])  # noqa: F821
 @login_required
@@ -169,6 +171,12 @@ def status():
     return get_json_result(data=res)
 
 
+@manager.route("/healthz", methods=["GET"])  # noqa: F821
+def healthz():
+    result, all_ok = run_health_checks()
+    return jsonify(result), (200 if all_ok else 500)
+
+
 @manager.route("/new_token", methods=["POST"])  # noqa: F821
 @login_required
 def new_token():

api/apps/user_app.py

@@ -34,7 +34,6 @@ from api.db.services.user_service import TenantService, UserService, UserTenantS
 from api.utils import (
     current_timestamp,
     datetime_format,
-    decrypt,
     download_img,
     get_format_time,
     get_uuid,
@@ -46,6 +45,7 @@ from api.utils.api_utils import (
     server_error_response,
     validate_request,
 )
+from api.utils.crypt import decrypt
 
 
 @manager.route("/login", methods=["POST", "GET"])  # noqa: F821

api/db/__init__.py

@@ -111,7 +111,7 @@ class CanvasCategory(StrEnum):
     Agent = "agent_canvas"
     DataFlow = "dataflow_canvas"
 
-VALID_CAVAS_CATEGORIES = {CanvasCategory.Agent, CanvasCategory.DataFlow}
+VALID_CANVAS_CATEGORIES = {CanvasCategory.Agent, CanvasCategory.DataFlow}
 
 
 class MCPServerType(StrEnum):

api/db/init_data.py

@@ -144,8 +144,9 @@ def init_llm_factory():
             except Exception:
                 pass
             break
+    doc_count = DocumentService.get_all_kb_doc_count()
     for kb_id in KnowledgebaseService.get_all_ids():
-        KnowledgebaseService.update_document_number_in_init(kb_id=kb_id, doc_num=DocumentService.get_kb_doc_count(kb_id))
+        KnowledgebaseService.update_document_number_in_init(kb_id=kb_id, doc_num=doc_count.get(kb_id, 0))
 
 
 

api/db/services/__init__.py

@@ -19,7 +19,7 @@ from pathlib import PurePath
 from .user_service import UserService as UserService
 
 
-def split_name_counter(filename: str) -> tuple[str, int | None]:
+def _split_name_counter(filename: str) -> tuple[str, int | None]:
     """
     Splits a filename into main part and counter (if present in parentheses).
 
@@ -87,7 +87,7 @@ def duplicate_name(query_func, **kwargs) -> str:
         stem = path.stem
         suffix = path.suffix
 
-        main_part, counter = split_name_counter(stem)
+        main_part, counter = _split_name_counter(stem)
         counter = counter + 1 if counter else 1
 
         new_name = f"{main_part}({counter}){suffix}"

api/db/services/conversation_service.py

@@ -23,7 +23,7 @@ from api.db.services.dialog_service import DialogService, chat
 from api.utils import get_uuid
 import json
 
-from rag.prompts import chunks_format
+from rag.prompts.generator import chunks_format
 
 
 class ConversationService(CommonService):

api/db/services/dialog_service.py

@@ -39,8 +39,8 @@ from graphrag.general.mind_map_extractor import MindMapExtractor
 from rag.app.resume import forbidden_select_fields4resume
 from rag.app.tag import label_question
 from rag.nlp.search import index_name
-from rag.prompts import chunks_format, citation_prompt, cross_languages, full_question, kb_prompt, keyword_extraction, message_fit_in
-from rag.prompts.prompts import gen_meta_filter, PROMPT_JINJA_ENV, ASK_SUMMARY
+from rag.prompts.generator import chunks_format, citation_prompt, cross_languages, full_question, kb_prompt, keyword_extraction, message_fit_in, \
+    gen_meta_filter, PROMPT_JINJA_ENV, ASK_SUMMARY
 from rag.utils import num_tokens_from_string, rmSpace
 from rag.utils.tavily_conn import Tavily
 
@@ -176,7 +176,7 @@ def chat_solo(dialog, messages, stream=True):
         delta_ans = ""
         for ans in chat_mdl.chat_streamly(prompt_config.get("system", ""), msg, dialog.llm_setting):
             answer = ans
-            delta_ans = ans[len(last_ans) :]
+            delta_ans = ans[len(last_ans):]
             if num_tokens_from_string(delta_ans) < 16:
                 continue
             last_ans = answer
@@ -261,13 +261,13 @@ def convert_conditions(metadata_condition):
         "not is": "≠"
     }
     return [
-    {
-        "op": op_mapping.get(cond["comparison_operator"], cond["comparison_operator"]),
-        "key": cond["name"],
-        "value": cond["value"]
-    }
-    for cond in metadata_condition.get("conditions", [])
-]
+        {
+            "op": op_mapping.get(cond["comparison_operator"], cond["comparison_operator"]),
+            "key": cond["name"],
+            "value": cond["value"]
+        }
+        for cond in metadata_condition.get("conditions", [])
+    ]
 
 
 def meta_filter(metas: dict, filters: list[dict]):
@@ -284,19 +284,19 @@ def meta_filter(metas: dict, filters: list[dict]):
                 value = str(value)
 
             for conds in [
-                    (operator == "contains", str(value).lower() in str(input).lower()),
-                    (operator == "not contains", str(value).lower() not in str(input).lower()),
-                    (operator == "start with", str(input).lower().startswith(str(value).lower())),
-                    (operator == "end with", str(input).lower().endswith(str(value).lower())),
-                    (operator == "empty", not input),
-                    (operator == "not empty", input),
-                    (operator == "=", input == value),
-                    (operator == "≠", input != value),
-                    (operator == ">", input > value),
-                    (operator == "<", input < value),
-                    (operator == "≥", input >= value),
-                    (operator == "≤", input <= value),
-                ]:
+                (operator == "contains", str(value).lower() in str(input).lower()),
+                (operator == "not contains", str(value).lower() not in str(input).lower()),
+                (operator == "start with", str(input).lower().startswith(str(value).lower())),
+                (operator == "end with", str(input).lower().endswith(str(value).lower())),
+                (operator == "empty", not input),
+                (operator == "not empty", input),
+                (operator == "=", input == value),
+                (operator == "≠", input != value),
+                (operator == ">", input > value),
+                (operator == "<", input < value),
+                (operator == "≥", input >= value),
+                (operator == "≤", input <= value),
+            ]:
                 try:
                     if all(conds):
                         ids.extend(docids)
@@ -456,7 +456,8 @@ def chat(dialog, messages, stream=True, **kwargs):
                 kbinfos["chunks"].extend(tav_res["chunks"])
                 kbinfos["doc_aggs"].extend(tav_res["doc_aggs"])
             if prompt_config.get("use_kg"):
-                ck = settings.kg_retrievaler.retrieval(" ".join(questions), tenant_ids, dialog.kb_ids, embd_mdl, LLMBundle(dialog.tenant_id, LLMType.CHAT))
+                ck = settings.kg_retrievaler.retrieval(" ".join(questions), tenant_ids, dialog.kb_ids, embd_mdl,
+                                                       LLMBundle(dialog.tenant_id, LLMType.CHAT))
                 if ck["content_with_weight"]:
                     kbinfos["chunks"].insert(0, ck)
 
@@ -467,7 +468,8 @@ def chat(dialog, messages, stream=True, **kwargs):
     retrieval_ts = timer()
     if not knowledges and prompt_config.get("empty_response"):
         empty_res = prompt_config["empty_response"]
-        yield {"answer": empty_res, "reference": kbinfos, "prompt": "\n\n### Query:\n%s" % " ".join(questions), "audio_binary": tts(tts_mdl, empty_res)}
+        yield {"answer": empty_res, "reference": kbinfos, "prompt": "\n\n### Query:\n%s" % " ".join(questions),
+               "audio_binary": tts(tts_mdl, empty_res)}
         return {"answer": prompt_config["empty_response"], "reference": kbinfos}
 
     kwargs["knowledge"] = "\n------\n" + "\n\n------\n\n".join(knowledges)
@@ -565,7 +567,8 @@ def chat(dialog, messages, stream=True, **kwargs):
 
     if langfuse_tracer:
         langfuse_generation = langfuse_tracer.start_generation(
-            trace_context=trace_context, name="chat", model=llm_model_config["llm_name"], input={"prompt": prompt, "prompt4citation": prompt4citation, "messages": msg}
+            trace_context=trace_context, name="chat", model=llm_model_config["llm_name"],
+            input={"prompt": prompt, "prompt4citation": prompt4citation, "messages": msg}
         )
 
     if stream:
@@ -575,12 +578,12 @@ def chat(dialog, messages, stream=True, **kwargs):
             if thought:
                 ans = re.sub(r"^.*</think>", "", ans, flags=re.DOTALL)
             answer = ans
-            delta_ans = ans[len(last_ans) :]
+            delta_ans = ans[len(last_ans):]
             if num_tokens_from_string(delta_ans) < 16:
                 continue
             last_ans = answer
             yield {"answer": thought + answer, "reference": {}, "audio_binary": tts(tts_mdl, delta_ans)}
-        delta_ans = answer[len(last_ans) :]
+        delta_ans = answer[len(last_ans):]
         if delta_ans:
             yield {"answer": thought + answer, "reference": {}, "audio_binary": tts(tts_mdl, delta_ans)}
         yield decorate_answer(thought + answer)
@@ -676,7 +679,9 @@ Please write the SQL, only SQL, without any other explanations or text.
 
     # compose Markdown table
     columns = (
-        "|" + "|".join([re.sub(r"(/.*|（[^（）]+）)", "", field_map.get(tbl["columns"][i]["name"], tbl["columns"][i]["name"])) for i in column_idx]) + ("|Source|" if docid_idx and docid_idx else "|")
+            "|" + "|".join(
+        [re.sub(r"(/.*|（[^（）]+）)", "", field_map.get(tbl["columns"][i]["name"], tbl["columns"][i]["name"])) for i in column_idx]) + (
+                "|Source|" if docid_idx and docid_idx else "|")
     )
 
     line = "|" + "|".join(["------" for _ in range(len(column_idx))]) + ("|------|" if docid_idx and docid_idx else "")
@@ -753,7 +758,7 @@ def ask(question, kb_ids, tenant_id, chat_llm_name=None, search_config={}):
                 doc_ids = None
 
     kbinfos = retriever.retrieval(
-        question = question,
+        question=question,
         embd_mdl=embd_mdl,
         tenant_ids=tenant_ids,
         kb_ids=kb_ids,
@@ -775,7 +780,8 @@ def ask(question, kb_ids, tenant_id, chat_llm_name=None, search_config={}):
 
     def decorate_answer(answer):
         nonlocal knowledges, kbinfos, sys_prompt
-        answer, idx = retriever.insert_citations(answer, [ck["content_ltks"] for ck in kbinfos["chunks"]], [ck["vector"] for ck in kbinfos["chunks"]], embd_mdl, tkweight=0.7, vtweight=0.3)
+        answer, idx = retriever.insert_citations(answer, [ck["content_ltks"] for ck in kbinfos["chunks"]], [ck["vector"] for ck in kbinfos["chunks"]],
+                                                 embd_mdl, tkweight=0.7, vtweight=0.3)
         idx = set([kbinfos["chunks"][int(i)]["doc_id"] for i in idx])
         recall_docs = [d for d in kbinfos["doc_aggs"] if d["doc_id"] in idx]
         if not recall_docs:

api/db/services/document_service.py

@@ -24,7 +24,7 @@ from io import BytesIO
 
 import trio
 import xxhash
-from peewee import fn
+from peewee import fn, Case
 
 from api import settings
 from api.constants import IMG_BASE64_PREFIX, FILE_NAME_LEN_LIMIT
@@ -660,8 +660,16 @@ class DocumentService(CommonService):
     @classmethod
     @DB.connection_context()
     def get_kb_doc_count(cls, kb_id):
-        return len(cls.model.select(cls.model.id).where(
-            cls.model.kb_id == kb_id).dicts())
+        return cls.model.select().where(cls.model.kb_id == kb_id).count()
+
+    @classmethod
+    @DB.connection_context()
+    def get_all_kb_doc_count(cls):
+        result = {}
+        rows = cls.model.select(cls.model.kb_id, fn.COUNT(cls.model.id).alias('count')).group_by(cls.model.kb_id)
+        for row in rows:
+            result[row.kb_id] = row.count
+        return result
 
     @classmethod
     @DB.connection_context()
@@ -674,6 +682,53 @@ class DocumentService(CommonService):
         return False
 
 
+    @classmethod
+    @DB.connection_context()
+    def knowledgebase_basic_info(cls, kb_id: str) -> dict[str, int]:
+        # cancelled: run == "2" but progress can vary
+        cancelled = (
+            cls.model.select(fn.COUNT(1))
+            .where((cls.model.kb_id == kb_id) & (cls.model.run == TaskStatus.CANCEL))
+            .scalar()
+        )
+
+        row = (
+            cls.model.select(
+                # finished: progress == 1
+                fn.COALESCE(fn.SUM(Case(None, [(cls.model.progress == 1, 1)], 0)), 0).alias("finished"),
+
+                # failed: progress == -1
+                fn.COALESCE(fn.SUM(Case(None, [(cls.model.progress == -1, 1)], 0)), 0).alias("failed"),
+
+                # processing: 0 <= progress < 1
+                fn.COALESCE(
+                    fn.SUM(
+                        Case(
+                            None,
+                            [
+                                (((cls.model.progress == 0) | ((cls.model.progress > 0) & (cls.model.progress < 1))), 1),
+                            ],
+                            0,
+                        )
+                    ),
+                    0,
+                ).alias("processing"),
+            )
+            .where(
+                (cls.model.kb_id == kb_id)
+                & ((cls.model.run.is_null(True)) | (cls.model.run != TaskStatus.CANCEL))
+            )
+            .dicts()
+            .get()
+        )
+
+        return {
+            "processing": int(row["processing"]),
+            "finished": int(row["finished"]),
+            "failed": int(row["failed"]),
+            "cancelled": int(cancelled),
+        }
+
 def queue_raptor_o_graphrag_tasks(doc, ty, priority):
     chunking_config = DocumentService.get_chunking_config(doc["id"])
     hasher = xxhash.xxh64()
@@ -702,6 +757,8 @@ def queue_raptor_o_graphrag_tasks(doc, ty, priority):
 
 def get_queue_length(priority):
     group_info = REDIS_CONN.queue_info(get_svr_queue_name(priority), SVR_CONSUMER_GROUP_NAME)
+    if not group_info:
+        return 0
     return int(group_info.get("lag", 0) or 0)
 
 
@@ -847,3 +904,4 @@ def doc_upload_and_parse(conversation_id, file_objs, user_id):
             doc_id, kb.id, token_counts[doc_id], chunk_counts[doc_id], 0)
 
     return [d["id"] for d, _ in files]
+

api/db/services/user_service.py

@@ -45,22 +45,22 @@ class UserService(CommonService):
     def query(cls, cols=None, reverse=None, order_by=None, **kwargs):
         if 'access_token' in kwargs:
             access_token = kwargs['access_token']
-            
+
             # Reject empty, None, or whitespace-only access tokens
             if not access_token or not str(access_token).strip():
                 logging.warning("UserService.query: Rejecting empty access_token query")
                 return cls.model.select().where(cls.model.id == "INVALID_EMPTY_TOKEN")  # Returns empty result
-            
+
             # Reject tokens that are too short (should be UUID, 32+ chars)
             if len(str(access_token).strip()) < 32:
                 logging.warning(f"UserService.query: Rejecting short access_token query: {len(str(access_token))} chars")
                 return cls.model.select().where(cls.model.id == "INVALID_SHORT_TOKEN")  # Returns empty result
-            
+
             # Reject tokens that start with "INVALID_" (from logout)
             if str(access_token).startswith("INVALID_"):
                 logging.warning("UserService.query: Rejecting invalidated access_token")
                 return cls.model.select().where(cls.model.id == "INVALID_LOGOUT_TOKEN")  # Returns empty result
-        
+
         # Call parent query method for valid requests
         return super().query(cols=cols, reverse=reverse, order_by=order_by, **kwargs)
 
@@ -140,6 +140,12 @@ class UserService(CommonService):
             cls.model.id == user_id,
             cls.model.is_superuser == 1).count() > 0
 
+    @classmethod
+    @DB.connection_context()
+    def get_all_users(cls):
+        users = cls.model.select()
+        return list(users)
+
 
 class TenantService(CommonService):
     """Service class for managing tenant-related database operations.

api/utils/__init__.py

@@ -28,8 +28,6 @@ import logging
 import copy
 from enum import Enum, IntEnum
 import importlib
-from Cryptodome.PublicKey import RSA
-from Cryptodome.Cipher import PKCS1_v1_5 as Cipher_pkcs1_v1_5
 from filelock import FileLock
 from api.constants import SERVICE_CONF
 
@@ -363,37 +361,6 @@ def elapsed2time(elapsed):
     return '%02d:%02d:%02d' % (hour, minuter, second)
 
 
-def decrypt(line):
-    file_path = os.path.join(
-        file_utils.get_project_base_directory(),
-        "conf",
-        "private.pem")
-    rsa_key = RSA.importKey(open(file_path).read(), "Welcome")
-    cipher = Cipher_pkcs1_v1_5.new(rsa_key)
-    return cipher.decrypt(base64.b64decode(
-        line), "Fail to decrypt password!").decode('utf-8')
-
-
-def decrypt2(crypt_text):
-    from base64 import b64decode, b16decode
-    from Crypto.Cipher import PKCS1_v1_5 as Cipher_PKCS1_v1_5
-    from Crypto.PublicKey import RSA
-    decode_data = b64decode(crypt_text)
-    if len(decode_data) == 127:
-        hex_fixed = '00' + decode_data.hex()
-        decode_data = b16decode(hex_fixed.upper())
-
-    file_path = os.path.join(
-        file_utils.get_project_base_directory(),
-        "conf",
-        "private.pem")
-    pem = open(file_path).read()
-    rsa_key = RSA.importKey(pem, "Welcome")
-    cipher = Cipher_PKCS1_v1_5.new(rsa_key)
-    decrypt_text = cipher.decrypt(decode_data, None)
-    return (b64decode(decrypt_text)).decode()
-
-
 def download_img(url):
     if not url:
         return ""

api/utils/crypt.py

@@ -0,0 +1,61 @@
+#
+#  Copyright 2025 The InfiniFlow Authors. All Rights Reserved.
+#
+#  Licensed under the Apache License, Version 2.0 (the "License");
+#  you may not use this file except in compliance with the License.
+#  You may obtain a copy of the License at
+#
+#      http://www.apache.org/licenses/LICENSE-2.0
+#
+#  Unless required by applicable law or agreed to in writing, software
+#  distributed under the License is distributed on an "AS IS" BASIS,
+#  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+#  See the License for the specific language governing permissions and
+#  limitations under the License.
+#
+
+import base64
+import os
+import sys
+from Cryptodome.PublicKey import RSA
+from Cryptodome.Cipher import PKCS1_v1_5 as Cipher_pkcs1_v1_5
+from api.utils import file_utils
+
+
+def crypt(line):
+    file_path = os.path.join(file_utils.get_project_base_directory(), "conf", "public.pem")
+    rsa_key = RSA.importKey(open(file_path).read(), "Welcome")
+    cipher = Cipher_pkcs1_v1_5.new(rsa_key)
+    password_base64 = base64.b64encode(line.encode('utf-8')).decode("utf-8")
+    encrypted_password = cipher.encrypt(password_base64.encode())
+    return base64.b64encode(encrypted_password).decode('utf-8')
+
+
+def decrypt(line):
+    file_path = os.path.join(file_utils.get_project_base_directory(), "conf", "private.pem")
+    rsa_key = RSA.importKey(open(file_path).read(), "Welcome")
+    cipher = Cipher_pkcs1_v1_5.new(rsa_key)
+    return cipher.decrypt(base64.b64decode(line), "Fail to decrypt password!").decode('utf-8')
+
+
+def decrypt2(crypt_text):
+    from base64 import b64decode, b16decode
+    from Crypto.Cipher import PKCS1_v1_5 as Cipher_PKCS1_v1_5
+    from Crypto.PublicKey import RSA
+    decode_data = b64decode(crypt_text)
+    if len(decode_data) == 127:
+        hex_fixed = '00' + decode_data.hex()
+        decode_data = b16decode(hex_fixed.upper())
+
+    file_path = os.path.join(file_utils.get_project_base_directory(), "conf", "private.pem")
+    pem = open(file_path).read()
+    rsa_key = RSA.importKey(pem, "Welcome")
+    cipher = Cipher_PKCS1_v1_5.new(rsa_key)
+    decrypt_text = cipher.decrypt(decode_data, None)
+    return (b64decode(decrypt_text)).decode()
+
+
+if __name__ == "__main__":
+    passwd = crypt(sys.argv[1])
+    print(passwd)
+    print(decrypt(passwd))

api/utils/health_utils.py

@@ -0,0 +1,107 @@
+#
+#  Copyright 2025 The InfiniFlow Authors. All Rights Reserved.
+#
+#  Licensed under the Apache License, Version 2.0 (the "License");
+#  you may not use this file except in compliance with the License.
+#  You may obtain a copy of the License at
+#
+#      http://www.apache.org/licenses/LICENSE-2.0
+#
+#  Unless required by applicable law or agreed to in writing, software
+#  distributed under the License is distributed on an "AS IS" BASIS,
+#  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+#  See the License for the specific language governing permissions and
+#  limitations under the License.
+#
+
+
+from timeit import default_timer as timer
+
+from api import settings
+from api.db.db_models import DB
+from rag.utils.redis_conn import REDIS_CONN
+from rag.utils.storage_factory import STORAGE_IMPL
+
+
+def _ok_nok(ok: bool) -> str:
+    return "ok" if ok else "nok"
+
+
+def check_db() -> tuple[bool, dict]:
+    st = timer()
+    try:
+        # lightweight probe; works for MySQL/Postgres
+        DB.execute_sql("SELECT 1")
+        return True, {"elapsed": f"{(timer() - st) * 1000.0:.1f}"}
+    except Exception as e:
+        return False, {"elapsed": f"{(timer() - st) * 1000.0:.1f}", "error": str(e)}
+
+
+def check_redis() -> tuple[bool, dict]:
+    st = timer()
+    try:
+        ok = bool(REDIS_CONN.health())
+        return ok, {"elapsed": f"{(timer() - st) * 1000.0:.1f}"}
+    except Exception as e:
+        return False, {"elapsed": f"{(timer() - st) * 1000.0:.1f}", "error": str(e)}
+
+
+def check_doc_engine() -> tuple[bool, dict]:
+    st = timer()
+    try:
+        meta = settings.docStoreConn.health()
+        # treat any successful call as ok
+        return True, {"elapsed": f"{(timer() - st) * 1000.0:.1f}", **(meta or {})}
+    except Exception as e:
+        return False, {"elapsed": f"{(timer() - st) * 1000.0:.1f}", "error": str(e)}
+
+
+def check_storage() -> tuple[bool, dict]:
+    st = timer()
+    try:
+        STORAGE_IMPL.health()
+        return True, {"elapsed": f"{(timer() - st) * 1000.0:.1f}"}
+    except Exception as e:
+        return False, {"elapsed": f"{(timer() - st) * 1000.0:.1f}", "error": str(e)}
+
+
+
+
+def run_health_checks() -> tuple[dict, bool]:
+    result: dict[str, str | dict] = {}
+
+    db_ok, db_meta = check_db()
+    result["db"] = _ok_nok(db_ok)
+    if not db_ok:
+        result.setdefault("_meta", {})["db"] = db_meta
+
+    try:
+        redis_ok, redis_meta = check_redis()
+        result["redis"] = _ok_nok(redis_ok)
+        if not redis_ok:
+            result.setdefault("_meta", {})["redis"] = redis_meta
+    except Exception:
+        result["redis"] = "nok"
+
+    try:
+        doc_ok, doc_meta = check_doc_engine()
+        result["doc_engine"] = _ok_nok(doc_ok)
+        if not doc_ok:
+            result.setdefault("_meta", {})["doc_engine"] = doc_meta
+    except Exception:
+        result["doc_engine"] = "nok"
+
+    try:
+        sto_ok, sto_meta = check_storage()
+        result["storage"] = _ok_nok(sto_ok)
+        if not sto_ok:
+            result.setdefault("_meta", {})["storage"] = sto_meta
+    except Exception:
+        result["storage"] = "nok"
+
+
+    all_ok = (result.get("db") == "ok") and (result.get("redis") == "ok") and (result.get("doc_engine") == "ok") and (result.get("storage") == "ok")
+    result["status"] = "ok" if all_ok else "nok"
+    return result, all_ok
+
+

api/utils/t_crypt.py

@@ -1,40 +0,0 @@
-#
-#  Copyright 2025 The InfiniFlow Authors. All Rights Reserved.
-#
-#  Licensed under the Apache License, Version 2.0 (the "License");
-#  you may not use this file except in compliance with the License.
-#  You may obtain a copy of the License at
-#
-#      http://www.apache.org/licenses/LICENSE-2.0
-#
-#  Unless required by applicable law or agreed to in writing, software
-#  distributed under the License is distributed on an "AS IS" BASIS,
-#  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-#  See the License for the specific language governing permissions and
-#  limitations under the License.
-#
-
-import base64
-import os
-import sys
-from Cryptodome.PublicKey import RSA
-from Cryptodome.Cipher import PKCS1_v1_5 as Cipher_pkcs1_v1_5
-from api.utils import decrypt, file_utils
-
-
-def crypt(line):
-    file_path = os.path.join(
-        file_utils.get_project_base_directory(),
-        "conf",
-        "public.pem")
-    rsa_key = RSA.importKey(open(file_path).read(),"Welcome")
-    cipher = Cipher_pkcs1_v1_5.new(rsa_key)
-    password_base64 = base64.b64encode(line.encode('utf-8')).decode("utf-8")
-    encrypted_password = cipher.encrypt(password_base64.encode())
-    return base64.b64encode(encrypted_password).decode('utf-8')
-
-
-if __name__ == "__main__":
-    passwd = crypt(sys.argv[1])
-    print(passwd)
-    print(decrypt(passwd))

chat_demo/index.html

@@ -0,0 +1,19 @@
+<iframe src="http://localhost:9222/next-chats/widget?shared_id=9dcfc68696c611f0bb789b9b8b765d12&from=chat&auth=U4MDU3NzkwOTZjNzExZjBiYjc4OWI5Yj&mode=master&streaming=false"
+  style="position:fixed;bottom:0;right:0;width:100px;height:100px;border:none;background:transparent;z-index:9999"
+  frameborder="0" allow="microphone;camera"></iframe>
+<script>
+window.addEventListener('message',e=>{
+  if(e.origin!=='http://localhost:9222')return;
+  if(e.data.type==='CREATE_CHAT_WINDOW'){
+    if(document.getElementById('chat-win'))return;
+    const i=document.createElement('iframe');
+    i.id='chat-win';i.src=e.data.src;
+    i.style.cssText='position:fixed;bottom:104px;right:24px;width:380px;height:500px;border:none;background:transparent;z-index:9998;display:none';
+    i.frameBorder='0';i.allow='microphone;camera';
+    document.body.appendChild(i);
+  }else if(e.data.type==='TOGGLE_CHAT'){
+    const w=document.getElementById('chat-win');
+    if(w)w.style.display=e.data.isOpen?'block':'none';
+  }else if(e.data.type==='SCROLL_PASSTHROUGH')window.scrollBy(0,e.data.deltaY);
+});
+</script>

chat_demo/widget_demo.html

@@ -0,0 +1,154 @@
+<!DOCTYPE html>
+<html lang="en">
+<head>
+    <meta charset="UTF-8">
+    <meta name="viewport" content="width=device-width, initial-scale=1.0">
+    <title>Floating Chat Widget Demo</title>
+    <style>
+        body {
+            font-family: Arial, sans-serif;
+            margin: 0;
+            padding: 40px;
+            background: linear-gradient(135deg, #667eea 0%, #764ba2 100%);
+            min-height: 100vh;
+            color: white;
+        }
+        
+        .demo-content {
+            max-width: 800px;
+            margin: 0 auto;
+        }
+        
+        .demo-content h1 {
+            text-align: center;
+            font-size: 2.5rem;
+            margin-bottom: 2rem;
+        }
+        
+        .demo-content p {
+            font-size: 1.2rem;
+            line-height: 1.6;
+            margin-bottom: 1.5rem;
+        }
+        
+        .feature-list {
+            background: rgba(255, 255, 255, 0.1);
+            border-radius: 10px;
+            padding: 2rem;
+            margin: 2rem 0;
+        }
+        
+        .feature-list h3 {
+            margin-top: 0;
+            font-size: 1.5rem;
+        }
+        
+        .feature-list ul {
+            list-style-type: none;
+            padding: 0;
+        }
+        
+        .feature-list li {
+            padding: 0.5rem 0;
+            padding-left: 1.5rem;
+            position: relative;
+        }
+        
+        .feature-list li:before {
+            content: "✓";
+            position: absolute;
+            left: 0;
+            color: #4ade80;
+            font-weight: bold;
+        }
+    </style>
+</head>
+<body>
+    <div class="demo-content">
+        <h1>🚀 Floating Chat Widget Demo</h1>
+        
+        <p>
+            Welcome to our demo page! This page simulates a real website with content. 
+            Look for the floating chat button in the bottom-right corner - just like Intercom!
+        </p>
+        
+        <div class="feature-list">
+            <h3>🎯 Widget Features</h3>
+            <ul>
+                <li>Floating button that stays visible while scrolling</li>
+                <li>Click to open/close the chat window</li>
+                <li>Minimize button to collapse the chat</li>
+                <li>Professional Intercom-style design</li>
+                <li>Unread message indicator (red badge)</li>
+                <li>Transparent background integration</li>
+                <li>Responsive design for all screen sizes</li>
+            </ul>
+        </div>
+        
+        <p>
+            The chat widget is completely separate from your website's content and won't 
+            interfere with your existing layout or functionality. It's designed to be 
+            lightweight and performant.
+        </p>
+        
+        <p>
+            Try scrolling this page - notice how the chat button stays in position. 
+            Click it to start a conversation with our AI assistant!
+        </p>
+        
+        <div class="feature-list">
+            <h3>🔧 Implementation</h3>
+            <ul>
+                <li>Simple iframe embed - just copy and paste</li>
+                <li>No JavaScript dependencies required</li>
+                <li>Works on any website or platform</li>
+                <li>Customizable appearance and behavior</li>
+                <li>Secure and privacy-focused</li>
+            </ul>
+        </div>
+        
+        <p>
+            This is just placeholder content to demonstrate how the widget integrates 
+            seamlessly with your existing website content. The widget floats above 
+            everything else without disrupting your user experience.
+        </p>
+        
+        <p style="margin-top: 4rem; text-align: center; font-style: italic;">
+            🎉 Ready to add this to your website? Get your embed code from the admin panel!
+        </p>
+    </div>
+
+    <iframe id="main-widget" src="http://localhost:9222/next-chats/widget?shared_id=9dcfc68696c611f0bb789b9b8b765d12&from=chat&auth=U4MDU3NzkwOTZjNzExZjBiYjc4OWI5Yj&visible_avatar=1&locale=zh&mode=master&streaming=false"
+    style="position:fixed;bottom:0;right:0;width:100px;height:100px;border:none;background:transparent;z-index:9999;opacity:0;transition:opacity 0.2s ease"
+    frameborder="0" allow="microphone;camera"></iframe>
+  <script>
+  window.addEventListener('message',e=>{
+    if(e.origin!=='http://localhost:9222')return;
+    if(e.data.type==='WIDGET_READY'){
+      // Show the main widget when React is ready
+      const mainWidget = document.getElementById('main-widget');
+      if(mainWidget) mainWidget.style.opacity = '1';
+    }else if(e.data.type==='CREATE_CHAT_WINDOW'){
+      if(document.getElementById('chat-win'))return;
+      const i=document.createElement('iframe');
+      i.id='chat-win';i.src=e.data.src;
+      i.style.cssText='position:fixed;bottom:104px;right:24px;width:380px;height:500px;border:none;background:transparent;z-index:9998;display:none;opacity:0;transition:opacity 0.2s ease';
+      i.frameBorder='0';i.allow='microphone;camera';
+      document.body.appendChild(i);
+    }else if(e.data.type==='TOGGLE_CHAT'){
+      const w=document.getElementById('chat-win');
+      if(w){
+        if(e.data.isOpen){
+          w.style.display='block';
+          // Wait for the iframe content to be ready before showing
+          setTimeout(() => w.style.opacity='1', 100);
+        }else{
+          w.style.opacity='0';
+          setTimeout(() => w.style.display='none', 200);
+        }
+      }
+    }else if(e.data.type==='SCROLL_PASSTHROUGH')window.scrollBy(0,e.data.deltaY);
+  });
+  </script>
+</body>
+</html>

conf/llm_factories.json

@@ -219,6 +219,70 @@
                 }
             ]
         },
+        {
+            "name": "TokenPony",
+            "logo": "",
+            "tags": "LLM",
+            "status": "1",
+            "llm": [
+                {
+                    "llm_name": "qwen3-8b",
+                    "tags": "LLM,CHAT,131k",
+                    "max_tokens": 131000,
+                    "model_type": "chat",
+                    "is_tools": true
+                },
+                {
+                    "llm_name": "deepseek-v3-0324",
+                    "tags": "LLM,CHAT,128k",
+                    "max_tokens": 128000,
+                    "model_type": "chat",
+                    "is_tools": true
+                },
+                {
+                    "llm_name": "qwen3-32b",
+                    "tags": "LLM,CHAT,131k",
+                    "max_tokens": 131000,
+                    "model_type": "chat",
+                    "is_tools": true
+                },
+                {
+                    "llm_name": "kimi-k2-instruct",
+                    "tags": "LLM,CHAT,128K",
+                    "max_tokens": 128000,
+                    "model_type": "chat",
+                    "is_tools": true
+                },
+                {
+                    "llm_name": "deepseek-r1-0528",
+                    "tags": "LLM,CHAT,164k",
+                    "max_tokens": 164000,
+                    "model_type": "chat",
+                    "is_tools": true
+                },
+                {
+                    "llm_name": "qwen3-coder-480b",
+                    "tags": "LLM,CHAT,1024k",
+                    "max_tokens": 1024000,
+                    "model_type": "chat",
+                    "is_tools": true
+                },
+                {
+                    "llm_name": "glm-4.5",
+                    "tags": "LLM,CHAT,131K",
+                    "max_tokens": 131000,
+                    "model_type": "chat",
+                    "is_tools": true
+                },
+                {
+                    "llm_name": "deepseek-v3.1",
+                    "tags": "LLM,CHAT,128k",
+                    "max_tokens": 128000,
+                    "model_type": "chat",
+                    "is_tools": true
+                }
+            ]
+        },
         {
             "name": "Tongyi-Qianwen",
             "logo": "",
@@ -338,7 +402,7 @@
                     "is_tools": true
                 },
                 {
-                    "llm_name": "qwen3-max-preview",
+                    "llm_name": "qwen3-max",
                     "tags": "LLM,CHAT,256k",
                     "max_tokens": 256000,
                     "model_type": "chat",
@@ -372,6 +436,27 @@
                     "model_type": "chat",
                     "is_tools": true
                 },
+                {
+                    "llm_name": "qwen3-vl-plus",
+                    "tags": "LLM,CHAT,IMAGE2TEXT,256k",
+                    "max_tokens": 256000,
+                    "model_type": "image2text",
+                    "is_tools": true
+                },
+                {
+                    "llm_name": "qwen3-vl-235b-a22b-instruct",
+                    "tags": "LLM,CHAT,IMAGE2TEXT,128k",
+                    "max_tokens": 128000,
+                    "model_type": "image2text",
+                    "is_tools": true
+                },
+                {
+                    "llm_name": "qwen3-vl-235b-a22b-thinking",
+                    "tags": "LLM,CHAT,IMAGE2TEXT,128k",
+                    "max_tokens": 128000,
+                    "model_type": "image2text",
+                    "is_tools": true
+                },
                 {
                     "llm_name": "qwen3-235b-a22b-instruct-2507",
                     "tags": "LLM,CHAT,128k",
@@ -393,6 +478,20 @@
                     "model_type": "chat",
                     "is_tools": true
                 },
+                {
+                    "llm_name": "qwen3-next-80b-a3b-instruct",
+                    "tags": "LLM,CHAT,128k",
+                    "max_tokens": 128000,
+                    "model_type": "chat",
+                    "is_tools": true
+                },
+                {
+                    "llm_name": "qwen3-next-80b-a3b-thinking",
+                    "tags": "LLM,CHAT,128k",
+                    "max_tokens": 128000,
+                    "model_type": "chat",
+                    "is_tools": true
+                },
                 {
                     "llm_name": "qwen3-0.6b",
                     "tags": "LLM,CHAT,32k",
@@ -558,6 +657,13 @@
                     "tags": "SPEECH2TEXT,8k",
                     "max_tokens": 8000,
                     "model_type": "speech2text"
+                },
+                {
+                    "llm_name": "qianwen-deepresearch-30b-a3b-131k",
+                    "tags": "LLM,CHAT,1M,AGENT,DEEPRESEARCH",
+                    "max_tokens": 1000000,
+                    "model_type": "chat",
+                    "is_tools": true
                 }
             ]
         },
@@ -625,7 +731,7 @@
                 },
                 {
                     "llm_name": "glm-4",
-                    "tags":"LLM,CHAT,128K",
+                    "tags": "LLM,CHAT,128K",
                     "max_tokens": 128000,
                     "model_type": "chat",
                     "is_tools": true
@@ -4477,6 +4583,273 @@
                 }
             ]
         },
+        {
+            "name": "CometAPI",
+            "logo": "",
+            "tags": "LLM,TEXT EMBEDDING,IMAGE2TEXT",
+            "status": "1",
+            "llm": [
+                {
+                    "llm_name": "gpt-5-chat-latest",
+                    "tags": "LLM,CHAT,400k",
+                    "max_tokens": 400000,
+                    "model_type": "chat",
+                    "is_tools": true
+                },
+                {
+                    "llm_name": "chatgpt-4o-latest",
+                    "tags": "LLM,CHAT,128k",
+                    "max_tokens": 128000,
+                    "model_type": "chat",
+                    "is_tools": true
+                },
+                {
+                    "llm_name": "gpt-5-mini",
+                    "tags": "LLM,CHAT,400k",
+                    "max_tokens": 400000,
+                    "model_type": "chat",
+                    "is_tools": true
+                },
+                {
+                    "llm_name": "gpt-5-nano",
+                    "tags": "LLM,CHAT,400k",
+                    "max_tokens": 400000,
+                    "model_type": "chat",
+                    "is_tools": true
+                },
+                {
+                    "llm_name": "gpt-5",
+                    "tags": "LLM,CHAT,400k",
+                    "max_tokens": 400000,
+                    "model_type": "chat",
+                    "is_tools": true
+                },
+                {
+                    "llm_name": "gpt-4.1-mini",
+                    "tags": "LLM,CHAT,1M",
+                    "max_tokens": 1047576,
+                    "model_type": "chat",
+                    "is_tools": true
+                },
+                {
+                    "llm_name": "gpt-4.1-nano",
+                    "tags": "LLM,CHAT,1M",
+                    "max_tokens": 1047576,
+                    "model_type": "chat",
+                    "is_tools": true
+                },
+                {
+                    "llm_name": "gpt-4.1",
+                    "tags": "LLM,CHAT,1M",
+                    "max_tokens": 1047576,
+                    "model_type": "chat",
+                    "is_tools": true
+                },
+                {
+                    "llm_name": "gpt-4o-mini",
+                    "tags": "LLM,CHAT,128k",
+                    "max_tokens": 128000,
+                    "model_type": "chat",
+                    "is_tools": true
+                },
+                {
+                    "llm_name": "o4-mini-2025-04-16",
+                    "tags": "LLM,CHAT,200k",
+                    "max_tokens": 200000,
+                    "model_type": "chat",
+                    "is_tools": true
+                },
+                {
+                    "llm_name": "o3-pro-2025-06-10",
+                    "tags": "LLM,CHAT,200k",
+                    "max_tokens": 200000,
+                    "model_type": "chat",
+                    "is_tools": true
+                },
+                {
+                    "llm_name": "claude-opus-4-1-20250805",
+                    "tags": "LLM,CHAT,200k,IMAGE2TEXT",
+                    "max_tokens": 200000,
+                    "model_type": "image2text",
+                    "is_tools": true
+                },
+                {
+                    "llm_name": "claude-opus-4-1-20250805-thinking",
+                    "tags": "LLM,CHAT,200k,IMAGE2TEXT",
+                    "max_tokens": 200000,
+                    "model_type": "image2text",
+                    "is_tools": true
+                },
+                {
+                    "llm_name": "claude-sonnet-4-20250514",
+                    "tags": "LLM,CHAT,200k,IMAGE2TEXT",
+                    "max_tokens": 200000,
+                    "model_type": "image2text",
+                    "is_tools": true
+                },
+                {
+                    "llm_name": "claude-sonnet-4-20250514-thinking",
+                    "tags": "LLM,CHAT,200k,IMAGE2TEXT",
+                    "max_tokens": 200000,
+                    "model_type": "image2text",
+                    "is_tools": true
+                },
+                {
+                    "llm_name": "claude-3-7-sonnet-latest",
+                    "tags": "LLM,CHAT,200k",
+                    "max_tokens": 200000,
+                    "model_type": "chat",
+                    "is_tools": true
+                },
+                {
+                    "llm_name": "claude-3-5-haiku-latest",
+                    "tags": "LLM,CHAT,200k",
+                    "max_tokens": 200000,
+                    "model_type": "chat",
+                    "is_tools": true
+                },
+                {
+                    "llm_name": "gemini-2.5-pro",
+                    "tags": "LLM,CHAT,1M,IMAGE2TEXT",
+                    "max_tokens": 1000000,
+                    "model_type": "image2text",
+                    "is_tools": true
+                },
+                {
+                    "llm_name": "gemini-2.5-flash",
+                    "tags": "LLM,CHAT,1M,IMAGE2TEXT",
+                    "max_tokens": 1000000,
+                    "model_type": "image2text",
+                    "is_tools": true
+                },
+                {
+                    "llm_name": "gemini-2.5-flash-lite",
+                    "tags": "LLM,CHAT,1M,IMAGE2TEXT",
+                    "max_tokens": 1000000,
+                    "model_type": "image2text",
+                    "is_tools": true
+                },
+                {
+                    "llm_name": "gemini-2.0-flash",
+                    "tags": "LLM,CHAT,1M,IMAGE2TEXT",
+                    "max_tokens": 1000000,
+                    "model_type": "image2text",
+                    "is_tools": true
+                },
+                {
+                    "llm_name": "grok-4-0709",
+                    "tags": "LLM,CHAT,131k",
+                    "max_tokens": 131072,
+                    "model_type": "chat",
+                    "is_tools": true
+                },
+                {
+                    "llm_name": "grok-3",
+                    "tags": "LLM,CHAT,131k",
+                    "max_tokens": 131072,
+                    "model_type": "chat",
+                    "is_tools": true
+                },
+                {
+                    "llm_name": "grok-3-mini",
+                    "tags": "LLM,CHAT,131k",
+                    "max_tokens": 131072,
+                    "model_type": "chat",
+                    "is_tools": true
+                },
+                {
+                    "llm_name": "grok-2-image-1212",
+                    "tags": "LLM,CHAT,32k,IMAGE2TEXT",
+                    "max_tokens": 32768,
+                    "model_type": "image2text",
+                    "is_tools": true
+                },
+                {
+                    "llm_name": "deepseek-v3.1",
+                    "tags": "LLM,CHAT,64k",
+                    "max_tokens": 64000,
+                    "model_type": "chat",
+                    "is_tools": true
+                },
+                {
+                    "llm_name": "deepseek-v3",
+                    "tags": "LLM,CHAT,64k",
+                    "max_tokens": 64000,
+                    "model_type": "chat",
+                    "is_tools": true
+                },
+                {
+                    "llm_name": "deepseek-r1-0528",
+                    "tags": "LLM,CHAT,164k",
+                    "max_tokens": 164000,
+                    "model_type": "chat",
+                    "is_tools": true
+                },
+                {
+                    "llm_name": "deepseek-chat",
+                    "tags": "LLM,CHAT,32k",
+                    "max_tokens": 32000,
+                    "model_type": "chat",
+                    "is_tools": true
+                },
+                {
+                    "llm_name": "deepseek-reasoner",
+                    "tags": "LLM,CHAT,64k",
+                    "max_tokens": 64000,
+                    "model_type": "chat",
+                    "is_tools": true
+                },
+                {
+                    "llm_name": "qwen3-30b-a3b",
+                    "tags": "LLM,CHAT,128k",
+                    "max_tokens": 128000,
+                    "model_type": "chat",
+                    "is_tools": true
+                },
+                {
+                    "llm_name": "qwen3-coder-plus-2025-07-22",
+                    "tags": "LLM,CHAT,128k",
+                    "max_tokens": 128000,
+                    "model_type": "chat",
+                    "is_tools": true
+                },
+                {
+                    "llm_name": "text-embedding-ada-002",
+                    "tags": "TEXT EMBEDDING,8K",
+                    "max_tokens": 8191,
+                    "model_type": "embedding",
+                    "is_tools": false
+                },
+                {
+                    "llm_name": "text-embedding-3-small",
+                    "tags": "TEXT EMBEDDING,8K",
+                    "max_tokens": 8191,
+                    "model_type": "embedding",
+                    "is_tools": false
+                },
+                {
+                    "llm_name": "text-embedding-3-large",
+                    "tags": "TEXT EMBEDDING,8K",
+                    "max_tokens": 8191,
+                    "model_type": "embedding",
+                    "is_tools": false
+                },
+                {
+                    "llm_name": "whisper-1",
+                    "tags": "SPEECH2TEXT",
+                    "max_tokens": 26214400,
+                    "model_type": "speech2text",
+                    "is_tools": false
+                },
+                {
+                    "llm_name": "tts-1",
+                    "tags": "TTS",
+                    "max_tokens": 2048,
+                    "model_type": "tts",
+                    "is_tools": false
+                }
+            ]
+        },
         {
             "name": "Meituan",
             "logo": "",
@@ -4493,4 +4866,4 @@
             ]
         }
     ]
-}
+}

conf/service_conf.yaml

@@ -1,6 +1,9 @@
 ragflow:
   host: 0.0.0.0
   http_port: 9380
+admin:
+  host: 0.0.0.0
+  http_port: 9381
 mysql:
   name: 'rag_flow'
   user: 'root'

deepdoc/parser/excel_parser.py

@@ -22,10 +22,10 @@ from openpyxl import Workbook, load_workbook
 from rag.nlp import find_codec
 
 # copied from `/openpyxl/cell/cell.py`
-ILLEGAL_CHARACTERS_RE = re.compile(r'[\000-\010]|[\013-\014]|[\016-\037]')
+ILLEGAL_CHARACTERS_RE = re.compile(r"[\000-\010]|[\013-\014]|[\016-\037]")
+
 
 class RAGFlowExcelParser:
-
     @staticmethod
     def _load_excel_to_workbook(file_like_object):
         if isinstance(file_like_object, bytes):
@@ -36,7 +36,7 @@ class RAGFlowExcelParser:
         file_head = file_like_object.read(4)
         file_like_object.seek(0)
 
-        if not (file_head.startswith(b'PK\x03\x04') or file_head.startswith(b'\xD0\xCF\x11\xE0')):
+        if not (file_head.startswith(b"PK\x03\x04") or file_head.startswith(b"\xd0\xcf\x11\xe0")):
             logging.info("Not an Excel file, converting CSV to Excel Workbook")
 
             try:
@@ -48,7 +48,7 @@ class RAGFlowExcelParser:
                 raise Exception(f"Failed to parse CSV and convert to Excel Workbook: {e_csv}")
 
         try:
-            return load_workbook(file_like_object,data_only= True)
+            return load_workbook(file_like_object, data_only=True)
         except Exception as e:
             logging.info(f"openpyxl load error: {e}, try pandas instead")
             try:
@@ -59,7 +59,7 @@ class RAGFlowExcelParser:
                 except Exception as ex:
                     logging.info(f"pandas with default engine load error: {ex}, try calamine instead")
                     file_like_object.seek(0)
-                    df = pd.read_excel(file_like_object, engine='calamine')
+                    df = pd.read_excel(file_like_object, engine="calamine")
                     return RAGFlowExcelParser._dataframe_to_workbook(df)
             except Exception as e_pandas:
                 raise Exception(f"pandas.read_excel error: {e_pandas}, original openpyxl error: {e}")
@@ -116,9 +116,7 @@ class RAGFlowExcelParser:
                 tb = ""
                 tb += f"<table><caption>{sheetname}</caption>"
                 tb += tb_rows_0
-                for r in list(
-                    rows[1 + chunk_i * chunk_rows: min(1 + (chunk_i + 1) * chunk_rows, len(rows))]
-                ):
+                for r in list(rows[1 + chunk_i * chunk_rows : min(1 + (chunk_i + 1) * chunk_rows, len(rows))]):
                     tb += "<tr>"
                     for i, c in enumerate(r):
                         if c.value is None:
@@ -133,8 +131,16 @@ class RAGFlowExcelParser:
 
     def markdown(self, fnm):
         import pandas as pd
+
         file_like_object = BytesIO(fnm) if not isinstance(fnm, str) else fnm
-        df = pd.read_excel(file_like_object)
+        try:
+            file_like_object.seek(0)
+            df = pd.read_excel(file_like_object)
+        except Exception as e:
+            logging.warning(f"Parse spreadsheet error: {e}, trying to interpret as CSV file")
+            file_like_object.seek(0)
+            df = pd.read_csv(file_like_object)
+        df = df.replace(r"^\s*$", "", regex=True)
         return df.to_markdown(index=False)
 
     def __call__(self, fnm):

deepdoc/parser/figure_parser.py

@@ -19,7 +19,7 @@ from PIL import Image
 
 from api.utils.api_utils import timeout
 from rag.app.picture import vision_llm_chunk as picture_vision_llm_chunk
-from rag.prompts import vision_llm_figure_describe_prompt
+from rag.prompts.generator import vision_llm_figure_describe_prompt
 
 
 def vision_figure_parser_figure_data_wrapper(figures_data_without_positions):

deepdoc/parser/html_parser.py

@@ -37,7 +37,7 @@ TITLE_TAGS = {"h1": "#", "h2": "##", "h3": "###", "h4": "#####", "h5": "#####",
 
 
 class RAGFlowHtmlParser:
-    def __call__(self, fnm, binary=None, chunk_token_num=None):
+    def __call__(self, fnm, binary=None, chunk_token_num=512):
         if binary:
             encoding = find_codec(binary)
             txt = binary.decode(encoding, errors="ignore")

deepdoc/parser/pdf_parser.py

@@ -34,10 +34,10 @@ from pypdf import PdfReader as pdf2_read
 
 from api import settings
 from api.utils.file_utils import get_project_base_directory
-from deepdoc.vision import OCR, LayoutRecognizer, Recognizer, TableStructureRecognizer
+from deepdoc.vision import OCR, AscendLayoutRecognizer, LayoutRecognizer, Recognizer, TableStructureRecognizer
 from rag.app.picture import vision_llm_chunk as picture_vision_llm_chunk
 from rag.nlp import rag_tokenizer
-from rag.prompts import vision_llm_describe_prompt
+from rag.prompts.generator import vision_llm_describe_prompt
 from rag.settings import PARALLEL_DEVICES
 
 LOCK_KEY_pdfplumber = "global_shared_lock_pdfplumber"
@@ -64,33 +64,38 @@ class RAGFlowPdfParser:
         if PARALLEL_DEVICES > 1:
             self.parallel_limiter = [trio.CapacityLimiter(1) for _ in range(PARALLEL_DEVICES)]
 
+        layout_recognizer_type = os.getenv("LAYOUT_RECOGNIZER_TYPE", "onnx").lower()
+        if layout_recognizer_type not in ["onnx", "ascend"]:
+            raise RuntimeError("Unsupported layout recognizer type.")
+
         if hasattr(self, "model_speciess"):
-            self.layouter = LayoutRecognizer("layout." + self.model_speciess)
+            recognizer_domain = "layout." + self.model_speciess
         else:
-            self.layouter = LayoutRecognizer("layout")
+            recognizer_domain = "layout"
+
+        if layout_recognizer_type == "ascend":
+            logging.debug("Using Ascend LayoutRecognizer")
+            self.layouter = AscendLayoutRecognizer(recognizer_domain)
+        else:  # onnx
+            logging.debug("Using Onnx LayoutRecognizer")
+            self.layouter = LayoutRecognizer(recognizer_domain)
         self.tbl_det = TableStructureRecognizer()
 
         self.updown_cnt_mdl = xgb.Booster()
         if not settings.LIGHTEN:
             try:
                 import torch.cuda
+
                 if torch.cuda.is_available():
                     self.updown_cnt_mdl.set_param({"device": "cuda"})
             except Exception:
                 logging.exception("RAGFlowPdfParser __init__")
         try:
-            model_dir = os.path.join(
-                get_project_base_directory(),
-                "rag/res/deepdoc")
-            self.updown_cnt_mdl.load_model(os.path.join(
-                model_dir, "updown_concat_xgb.model"))
+            model_dir = os.path.join(get_project_base_directory(), "rag/res/deepdoc")
+            self.updown_cnt_mdl.load_model(os.path.join(model_dir, "updown_concat_xgb.model"))
         except Exception:
-            model_dir = snapshot_download(
-                repo_id="InfiniFlow/text_concat_xgb_v1.0",
-                local_dir=os.path.join(get_project_base_directory(), "rag/res/deepdoc"),
-                local_dir_use_symlinks=False)
-            self.updown_cnt_mdl.load_model(os.path.join(
-                model_dir, "updown_concat_xgb.model"))
+            model_dir = snapshot_download(repo_id="InfiniFlow/text_concat_xgb_v1.0", local_dir=os.path.join(get_project_base_directory(), "rag/res/deepdoc"), local_dir_use_symlinks=False)
+            self.updown_cnt_mdl.load_model(os.path.join(model_dir, "updown_concat_xgb.model"))
 
         self.page_from = 0
         self.column_num = 1
@@ -102,13 +107,10 @@ class RAGFlowPdfParser:
         return c["bottom"] - c["top"]
 
     def _x_dis(self, a, b):
-        return min(abs(a["x1"] - b["x0"]), abs(a["x0"] - b["x1"]),
-                   abs(a["x0"] + a["x1"] - b["x0"] - b["x1"]) / 2)
+        return min(abs(a["x1"] - b["x0"]), abs(a["x0"] - b["x1"]), abs(a["x0"] + a["x1"] - b["x0"] - b["x1"]) / 2)
 
-    def _y_dis(
-            self, a, b):
-        return (
-            b["top"] + b["bottom"] - a["top"] - a["bottom"]) / 2
+    def _y_dis(self, a, b):
+        return (b["top"] + b["bottom"] - a["top"] - a["bottom"]) / 2
 
     def _match_proj(self, b):
         proj_patt = [
@@ -130,10 +132,7 @@ class RAGFlowPdfParser:
         LEN = 6
         tks_down = rag_tokenizer.tokenize(down["text"][:LEN]).split()
         tks_up = rag_tokenizer.tokenize(up["text"][-LEN:]).split()
-        tks_all = up["text"][-LEN:].strip() \
-            + (" " if re.match(r"[a-zA-Z0-9]+",
-                               up["text"][-1] + down["text"][0]) else "") \
-            + down["text"][:LEN].strip()
+        tks_all = up["text"][-LEN:].strip() + (" " if re.match(r"[a-zA-Z0-9]+", up["text"][-1] + down["text"][0]) else "") + down["text"][:LEN].strip()
         tks_all = rag_tokenizer.tokenize(tks_all).split()
         fea = [
             up.get("R", -1) == down.get("R", -1),
@@ -144,39 +143,30 @@ class RAGFlowPdfParser:
             down["layout_type"] == "text",
             up["layout_type"] == "table",
             down["layout_type"] == "table",
-            True if re.search(
-                r"([。？！；!?;+)）]|[a-z]\.)$",
-                up["text"]) else False,
+            True if re.search(r"([。？！；!?;+)）]|[a-z]\.)$", up["text"]) else False,
             True if re.search(r"[，：‘“、0-9（+-]$", up["text"]) else False,
-            True if re.search(
-                r"(^.?[/,?;:\]，。；：’”？！》】）-])",
-                down["text"]) else False,
+            True if re.search(r"(^.?[/,?;:\]，。；：’”？！》】）-])", down["text"]) else False,
             True if re.match(r"[\(（][^\(\)（）]+[）\)]$", up["text"]) else False,
             True if re.search(r"[，,][^。.]+$", up["text"]) else False,
             True if re.search(r"[，,][^。.]+$", up["text"]) else False,
-            True if re.search(r"[\(（][^\)）]+$", up["text"])
-            and re.search(r"[\)）]", down["text"]) else False,
+            True if re.search(r"[\(（][^\)）]+$", up["text"]) and re.search(r"[\)）]", down["text"]) else False,
             self._match_proj(down),
             True if re.match(r"[A-Z]", down["text"]) else False,
             True if re.match(r"[A-Z]", up["text"][-1]) else False,
             True if re.match(r"[a-z0-9]", up["text"][-1]) else False,
             True if re.match(r"[0-9.%,-]+$", down["text"]) else False,
-            up["text"].strip()[-2:] == down["text"].strip()[-2:] if len(up["text"].strip()
-                                                                        ) > 1 and len(
-                down["text"].strip()) > 1 else False,
+            up["text"].strip()[-2:] == down["text"].strip()[-2:] if len(up["text"].strip()) > 1 and len(down["text"].strip()) > 1 else False,
             up["x0"] > down["x1"],
-            abs(self.__height(up) - self.__height(down)) / min(self.__height(up),
-                                                               self.__height(down)),
+            abs(self.__height(up) - self.__height(down)) / min(self.__height(up), self.__height(down)),
             self._x_dis(up, down) / max(w, 0.000001),
-            (len(up["text"]) - len(down["text"])) /
-            max(len(up["text"]), len(down["text"])),
+            (len(up["text"]) - len(down["text"])) / max(len(up["text"]), len(down["text"])),
             len(tks_all) - len(tks_up) - len(tks_down),
             len(tks_down) - len(tks_up),
             tks_down[-1] == tks_up[-1] if tks_down and tks_up else False,
             max(down["in_row"], up["in_row"]),
             abs(down["in_row"] - up["in_row"]),
             len(tks_down) == 1 and rag_tokenizer.tag(tks_down[0]).find("n") >= 0,
-            len(tks_up) == 1 and rag_tokenizer.tag(tks_up[0]).find("n") >= 0
+            len(tks_up) == 1 and rag_tokenizer.tag(tks_up[0]).find("n") >= 0,
         ]
         return fea
 
@@ -187,9 +177,7 @@ class RAGFlowPdfParser:
         for i in range(len(arr) - 1):
             for j in range(i, -1, -1):
                 # restore the order using th
-                if abs(arr[j + 1]["x0"] - arr[j]["x0"]) < threshold \
-                        and arr[j + 1]["top"] < arr[j]["top"] \
-                        and arr[j + 1]["page_number"] == arr[j]["page_number"]:
+                if abs(arr[j + 1]["x0"] - arr[j]["x0"]) < threshold and arr[j + 1]["top"] < arr[j]["top"] and arr[j + 1]["page_number"] == arr[j]["page_number"]:
                     tmp = arr[j]
                     arr[j] = arr[j + 1]
                     arr[j + 1] = tmp
@@ -197,8 +185,7 @@ class RAGFlowPdfParser:
 
     def _has_color(self, o):
         if o.get("ncs", "") == "DeviceGray":
-            if o["stroking_color"] and o["stroking_color"][0] == 1 and o["non_stroking_color"] and \
-                    o["non_stroking_color"][0] == 1:
+            if o["stroking_color"] and o["stroking_color"][0] == 1 and o["non_stroking_color"] and o["non_stroking_color"][0] == 1:
                 if re.match(r"[a-zT_\[\]\(\)-]+", o.get("text", "")):
                     return False
         return True
@@ -216,8 +203,7 @@ class RAGFlowPdfParser:
             if not tbls:
                 continue
             for tb in tbls:  # for table
-                left, top, right, bott = tb["x0"] - MARGIN, tb["top"] - MARGIN, \
-                    tb["x1"] + MARGIN, tb["bottom"] + MARGIN
+                left, top, right, bott = tb["x0"] - MARGIN, tb["top"] - MARGIN, tb["x1"] + MARGIN, tb["bottom"] + MARGIN
                 left *= ZM
                 top *= ZM
                 right *= ZM
@@ -232,14 +218,13 @@ class RAGFlowPdfParser:
         tbcnt = np.cumsum(tbcnt)
         for i in range(len(tbcnt) - 1):  # for page
             pg = []
-            for j, tb_items in enumerate(
-                    recos[tbcnt[i]: tbcnt[i + 1]]):  # for table
-                poss = pos[tbcnt[i]: tbcnt[i + 1]]
+            for j, tb_items in enumerate(recos[tbcnt[i] : tbcnt[i + 1]]):  # for table
+                poss = pos[tbcnt[i] : tbcnt[i + 1]]
                 for it in tb_items:  # for table components
-                    it["x0"] = (it["x0"] + poss[j][0])
-                    it["x1"] = (it["x1"] + poss[j][0])
-                    it["top"] = (it["top"] + poss[j][1])
-                    it["bottom"] = (it["bottom"] + poss[j][1])
+                    it["x0"] = it["x0"] + poss[j][0]
+                    it["x1"] = it["x1"] + poss[j][0]
+                    it["top"] = it["top"] + poss[j][1]
+                    it["bottom"] = it["bottom"] + poss[j][1]
                     for n in ["x0", "x1", "top", "bottom"]:
                         it[n] /= ZM
                     it["top"] += self.page_cum_height[i]
@@ -250,8 +235,7 @@ class RAGFlowPdfParser:
             self.tb_cpns.extend(pg)
 
         def gather(kwd, fzy=10, ption=0.6):
-            eles = Recognizer.sort_Y_firstly(
-                [r for r in self.tb_cpns if re.match(kwd, r["label"])], fzy)
+            eles = Recognizer.sort_Y_firstly([r for r in self.tb_cpns if re.match(kwd, r["label"])], fzy)
             eles = Recognizer.layouts_cleanup(self.boxes, eles, 5, ption)
             return Recognizer.sort_Y_firstly(eles, 0)
 
@@ -259,8 +243,7 @@ class RAGFlowPdfParser:
         headers = gather(r".*header$")
         rows = gather(r".* (row|header)")
         spans = gather(r".*spanning")
-        clmns = sorted([r for r in self.tb_cpns if re.match(
-            r"table column$", r["label"])], key=lambda x: (x["pn"], x["layoutno"], x["x0"]))
+        clmns = sorted([r for r in self.tb_cpns if re.match(r"table column$", r["label"])], key=lambda x: (x["pn"], x["layoutno"], x["x0"]))
         clmns = Recognizer.layouts_cleanup(self.boxes, clmns, 5, 0.5)
         for b in self.boxes:
             if b.get("layout_type", "") != "table":
@@ -271,8 +254,7 @@ class RAGFlowPdfParser:
                 b["R_top"] = rows[ii]["top"]
                 b["R_bott"] = rows[ii]["bottom"]
 
-            ii = Recognizer.find_overlapped_with_threshold(
-                b, headers, thr=0.3)
+            ii = Recognizer.find_overlapped_with_threshold(b, headers, thr=0.3)
             if ii is not None:
                 b["H_top"] = headers[ii]["top"]
                 b["H_bott"] = headers[ii]["bottom"]
@@ -305,12 +287,12 @@ class RAGFlowPdfParser:
             return
         bxs = [(line[0], line[1][0]) for line in bxs]
         bxs = Recognizer.sort_Y_firstly(
-            [{"x0": b[0][0] / ZM, "x1": b[1][0] / ZM,
-              "top": b[0][1] / ZM, "text": "", "txt": t,
-              "bottom": b[-1][1] / ZM,
-              "chars": [],
-              "page_number": pagenum} for b, t in bxs if b[0][0] <= b[1][0] and b[0][1] <= b[-1][1]],
-            self.mean_height[pagenum-1] / 3
+            [
+                {"x0": b[0][0] / ZM, "x1": b[1][0] / ZM, "top": b[0][1] / ZM, "text": "", "txt": t, "bottom": b[-1][1] / ZM, "chars": [], "page_number": pagenum}
+                for b, t in bxs
+                if b[0][0] <= b[1][0] and b[0][1] <= b[-1][1]
+            ],
+            self.mean_height[pagenum - 1] / 3,
         )
 
         # merge chars in the same rect
@@ -321,7 +303,7 @@ class RAGFlowPdfParser:
                 continue
             ch = c["bottom"] - c["top"]
             bh = bxs[ii]["bottom"] - bxs[ii]["top"]
-            if abs(ch - bh) / max(ch, bh) >= 0.7 and c["text"] != ' ':
+            if abs(ch - bh) / max(ch, bh) >= 0.7 and c["text"] != " ":
                 self.lefted_chars.append(c)
                 continue
             bxs[ii]["chars"].append(c)
@@ -345,8 +327,7 @@ class RAGFlowPdfParser:
         img_np = np.array(img)
         for b in bxs:
             if not b["text"]:
-                left, right, top, bott = b["x0"] * ZM, b["x1"] * \
-                                         ZM, b["top"] * ZM, b["bottom"] * ZM
+                left, right, top, bott = b["x0"] * ZM, b["x1"] * ZM, b["top"] * ZM, b["bottom"] * ZM
                 b["box_image"] = self.ocr.get_rotate_crop_image(img_np, np.array([[left, top], [right, top], [right, bott], [left, bott]], dtype=np.float32))
                 boxes_to_reg.append(b)
             del b["txt"]
@@ -356,21 +337,17 @@ class RAGFlowPdfParser:
             del boxes_to_reg[i]["box_image"]
         logging.info(f"__ocr recognize {len(bxs)} boxes cost {timer() - start}s")
         bxs = [b for b in bxs if b["text"]]
-        if self.mean_height[pagenum-1] == 0:
-            self.mean_height[pagenum-1] = np.median([b["bottom"] - b["top"]
-                                              for b in bxs])
+        if self.mean_height[pagenum - 1] == 0:
+            self.mean_height[pagenum - 1] = np.median([b["bottom"] - b["top"] for b in bxs])
         self.boxes.append(bxs)
 
     def _layouts_rec(self, ZM, drop=True):
         assert len(self.page_images) == len(self.boxes)
-        self.boxes, self.page_layout = self.layouter(
-            self.page_images, self.boxes, ZM, drop=drop)
+        self.boxes, self.page_layout = self.layouter(self.page_images, self.boxes, ZM, drop=drop)
         # cumlative Y
         for i in range(len(self.boxes)):
-            self.boxes[i]["top"] += \
-                self.page_cum_height[self.boxes[i]["page_number"] - 1]
-            self.boxes[i]["bottom"] += \
-                self.page_cum_height[self.boxes[i]["page_number"] - 1]
+            self.boxes[i]["top"] += self.page_cum_height[self.boxes[i]["page_number"] - 1]
+            self.boxes[i]["bottom"] += self.page_cum_height[self.boxes[i]["page_number"] - 1]
 
     def _text_merge(self):
         # merge adjusted boxes
@@ -390,12 +367,10 @@ class RAGFlowPdfParser:
         while i < len(bxs) - 1:
             b = bxs[i]
             b_ = bxs[i + 1]
-            if b.get("layoutno", "0") != b_.get("layoutno", "1") or b.get("layout_type", "") in ["table", "figure",
-                                                                                                 "equation"]:
+            if b.get("layoutno", "0") != b_.get("layoutno", "1") or b.get("layout_type", "") in ["table", "figure", "equation"]:
                 i += 1
                 continue
-            if abs(self._y_dis(b, b_)
-                   ) < self.mean_height[bxs[i]["page_number"] - 1] / 3:
+            if abs(self._y_dis(b, b_)) < self.mean_height[bxs[i]["page_number"] - 1] / 3:
                 # merge
                 bxs[i]["x1"] = b_["x1"]
                 bxs[i]["top"] = (b["top"] + b_["top"]) / 2
@@ -408,16 +383,14 @@ class RAGFlowPdfParser:
 
             dis_thr = 1
             dis = b["x1"] - b_["x0"]
-            if b.get("layout_type", "") != "text" or b_.get(
-                    "layout_type", "") != "text":
+            if b.get("layout_type", "") != "text" or b_.get("layout_type", "") != "text":
                 if end_with(b, "，") or start_with(b_, "（，"):
                     dis_thr = -8
                 else:
                     i += 1
                     continue
 
-            if abs(self._y_dis(b, b_)) < self.mean_height[bxs[i]["page_number"] - 1] / 5 \
-                    and dis >= dis_thr and b["x1"] < b_["x1"]:
+            if abs(self._y_dis(b, b_)) < self.mean_height[bxs[i]["page_number"] - 1] / 5 and dis >= dis_thr and b["x1"] < b_["x1"]:
                 # merge
                 bxs[i]["x1"] = b_["x1"]
                 bxs[i]["top"] = (b["top"] + b_["top"]) / 2
@@ -429,23 +402,22 @@ class RAGFlowPdfParser:
         self.boxes = bxs
 
     def _naive_vertical_merge(self, zoomin=3):
-        bxs = Recognizer.sort_Y_firstly(
-            self.boxes, np.median(
-                self.mean_height) / 3)
+        import math
+        bxs = Recognizer.sort_Y_firstly(self.boxes, np.median(self.mean_height) / 3)
 
         column_width = np.median([b["x1"] - b["x0"] for b in self.boxes])
+        if not column_width or math.isnan(column_width):
+            column_width = self.mean_width[0]
         self.column_num = int(self.page_images[0].size[0] / zoomin / column_width)
         if column_width < self.page_images[0].size[0] / zoomin / self.column_num:
-            logging.info("Multi-column................... {} {}".format(column_width,
-                  self.page_images[0].size[0] / zoomin / self.column_num))
+            logging.info("Multi-column................... {} {}".format(column_width, self.page_images[0].size[0] / zoomin / self.column_num))
             self.boxes = self.sort_X_by_page(self.boxes, column_width / self.column_num)
 
         i = 0
         while i + 1 < len(bxs):
             b = bxs[i]
             b_ = bxs[i + 1]
-            if b["page_number"] < b_["page_number"] and re.match(
-                    r"[0-9  •一—-]+$", b["text"]):
+            if b["page_number"] < b_["page_number"] and re.match(r"[0-9  •一—-]+$", b["text"]):
                 bxs.pop(i)
                 continue
             if not b["text"].strip():
@@ -453,8 +425,7 @@ class RAGFlowPdfParser:
                 continue
             concatting_feats = [
                 b["text"].strip()[-1] in ",;:'\"，、‘“；：-",
-                len(b["text"].strip()) > 1 and b["text"].strip(
-                )[-2] in ",;:'\"，‘“、；：",
+                len(b["text"].strip()) > 1 and b["text"].strip()[-2] in ",;:'\"，‘“、；：",
                 b_["text"].strip() and b_["text"].strip()[0] in "。；？！?”）),，、：",
             ]
             # features for not concating
@@ -462,21 +433,20 @@ class RAGFlowPdfParser:
                 b.get("layoutno", 0) != b_.get("layoutno", 0),
                 b["text"].strip()[-1] in "。？！?",
                 self.is_english and b["text"].strip()[-1] in ".!?",
-                b["page_number"] == b_["page_number"] and b_["top"] -
-                b["bottom"] > self.mean_height[b["page_number"] - 1] * 1.5,
-                b["page_number"] < b_["page_number"] and abs(
-                    b["x0"] - b_["x0"]) > self.mean_width[b["page_number"] - 1] * 4,
+                b["page_number"] == b_["page_number"] and b_["top"] - b["bottom"] > self.mean_height[b["page_number"] - 1] * 1.5,
+                b["page_number"] < b_["page_number"] and abs(b["x0"] - b_["x0"]) > self.mean_width[b["page_number"] - 1] * 4,
             ]
             # split features
-            detach_feats = [b["x1"] < b_["x0"],
-                            b["x0"] > b_["x1"]]
+            detach_feats = [b["x1"] < b_["x0"], b["x0"] > b_["x1"]]
             if (any(feats) and not any(concatting_feats)) or any(detach_feats):
-                logging.debug("{} {} {} {}".format(
-                    b["text"],
-                    b_["text"],
-                    any(feats),
-                    any(concatting_feats),
-                ))
+                logging.debug(
+                    "{} {} {} {}".format(
+                        b["text"],
+                        b_["text"],
+                        any(feats),
+                        any(concatting_feats),
+                    )
+                )
                 i += 1
                 continue
             # merge up and down
@@ -529,14 +499,11 @@ class RAGFlowPdfParser:
                     if not concat_between_pages and down["page_number"] > up["page_number"]:
                         break
 
-                    if up.get("R", "") != down.get(
-                            "R", "") and up["text"][-1] != "，":
+                    if up.get("R", "") != down.get("R", "") and up["text"][-1] != "，":
                         i += 1
                         continue
 
-                    if re.match(r"[0-9]{2,3}/[0-9]{3}$", up["text"]) \
-                            or re.match(r"[0-9]{2,3}/[0-9]{3}$", down["text"]) \
-                            or not down["text"].strip():
+                    if re.match(r"[0-9]{2,3}/[0-9]{3}$", up["text"]) or re.match(r"[0-9]{2,3}/[0-9]{3}$", down["text"]) or not down["text"].strip():
                         i += 1
                         continue
 
@@ -544,14 +511,12 @@ class RAGFlowPdfParser:
                         i += 1
                         continue
 
-                    if up["x1"] < down["x0"] - 10 * \
-                            mw or up["x0"] > down["x1"] + 10 * mw:
+                    if up["x1"] < down["x0"] - 10 * mw or up["x0"] > down["x1"] + 10 * mw:
                         i += 1
                         continue
 
                     if i - dp < 5 and up.get("layout_type") == "text":
-                        if up.get("layoutno", "1") == down.get(
-                                "layoutno", "2"):
+                        if up.get("layoutno", "1") == down.get("layoutno", "2"):
                             dfs(down, i + 1)
                             boxes.pop(i)
                             return
@@ -559,8 +524,7 @@ class RAGFlowPdfParser:
                         continue
 
                     fea = self._updown_concat_features(up, down)
-                    if self.updown_cnt_mdl.predict(
-                            xgb.DMatrix([fea]))[0] <= 0.5:
+                    if self.updown_cnt_mdl.predict(xgb.DMatrix([fea]))[0] <= 0.5:
                         i += 1
                         continue
                     dfs(down, i + 1)
@@ -584,16 +548,14 @@ class RAGFlowPdfParser:
                 c["text"] = c["text"].strip()
                 if not c["text"]:
                     continue
-                if t["text"] and re.match(
-                        r"[0-9\.a-zA-Z]+$", t["text"][-1] + c["text"][-1]):
+                if t["text"] and re.match(r"[0-9\.a-zA-Z]+$", t["text"][-1] + c["text"][-1]):
                     t["text"] += " "
                 t["text"] += c["text"]
                 t["x0"] = min(t["x0"], c["x0"])
                 t["x1"] = max(t["x1"], c["x1"])
                 t["page_number"] = min(t["page_number"], c["page_number"])
                 t["bottom"] = c["bottom"]
-                if not t["layout_type"] \
-                        and c["layout_type"]:
+                if not t["layout_type"] and c["layout_type"]:
                     t["layout_type"] = c["layout_type"]
             boxes.append(t)
 
@@ -605,25 +567,20 @@ class RAGFlowPdfParser:
         findit = False
         i = 0
         while i < len(self.boxes):
-            if not re.match(r"(contents|目录|目次|table of contents|致谢|acknowledge)$",
-                            re.sub(r"( | |\u3000)+", "", self.boxes[i]["text"].lower())):
+            if not re.match(r"(contents|目录|目次|table of contents|致谢|acknowledge)$", re.sub(r"( | |\u3000)+", "", self.boxes[i]["text"].lower())):
                 i += 1
                 continue
             findit = True
-            eng = re.match(
-                r"[0-9a-zA-Z :'.-]{5,}",
-                self.boxes[i]["text"].strip())
+            eng = re.match(r"[0-9a-zA-Z :'.-]{5,}", self.boxes[i]["text"].strip())
             self.boxes.pop(i)
             if i >= len(self.boxes):
                 break
-            prefix = self.boxes[i]["text"].strip()[:3] if not eng else " ".join(
-                self.boxes[i]["text"].strip().split()[:2])
+            prefix = self.boxes[i]["text"].strip()[:3] if not eng else " ".join(self.boxes[i]["text"].strip().split()[:2])
             while not prefix:
                 self.boxes.pop(i)
                 if i >= len(self.boxes):
                     break
-                prefix = self.boxes[i]["text"].strip()[:3] if not eng else " ".join(
-                    self.boxes[i]["text"].strip().split()[:2])
+                prefix = self.boxes[i]["text"].strip()[:3] if not eng else " ".join(self.boxes[i]["text"].strip().split()[:2])
             self.boxes.pop(i)
             if i >= len(self.boxes) or not prefix:
                 break
@@ -662,10 +619,12 @@ class RAGFlowPdfParser:
                 self.boxes.pop(i + 1)
                 continue
 
-            if b["text"].strip()[0] != b_["text"].strip()[0] \
-                    or b["text"].strip()[0].lower() in set("qwertyuopasdfghjklzxcvbnm") \
-                    or rag_tokenizer.is_chinese(b["text"].strip()[0]) \
-                    or b["top"] > b_["bottom"]:
+            if (
+                b["text"].strip()[0] != b_["text"].strip()[0]
+                or b["text"].strip()[0].lower() in set("qwertyuopasdfghjklzxcvbnm")
+                or rag_tokenizer.is_chinese(b["text"].strip()[0])
+                or b["top"] > b_["bottom"]
+            ):
                 i += 1
                 continue
             b_["text"] = b["text"] + "\n" + b_["text"]
@@ -685,12 +644,8 @@ class RAGFlowPdfParser:
             if "layoutno" not in self.boxes[i]:
                 i += 1
                 continue
-            lout_no = str(self.boxes[i]["page_number"]) + \
-                "-" + str(self.boxes[i]["layoutno"])
-            if TableStructureRecognizer.is_caption(self.boxes[i]) or self.boxes[i]["layout_type"] in ["table caption",
-                                                                                                      "title",
-                                                                                                      "figure caption",
-                                                                                                      "reference"]:
+            lout_no = str(self.boxes[i]["page_number"]) + "-" + str(self.boxes[i]["layoutno"])
+            if TableStructureRecognizer.is_caption(self.boxes[i]) or self.boxes[i]["layout_type"] in ["table caption", "title", "figure caption", "reference"]:
                 nomerge_lout_no.append(lst_lout_no)
             if self.boxes[i]["layout_type"] == "table":
                 if re.match(r"(数据|资料|图表)*来源[:： ]", self.boxes[i]["text"]):
@@ -716,8 +671,7 @@ class RAGFlowPdfParser:
 
         # merge table on different pages
         nomerge_lout_no = set(nomerge_lout_no)
-        tbls = sorted([(k, bxs) for k, bxs in tables.items()],
-                      key=lambda x: (x[1][0]["top"], x[1][0]["x0"]))
+        tbls = sorted([(k, bxs) for k, bxs in tables.items()], key=lambda x: (x[1][0]["top"], x[1][0]["x0"]))
 
         i = len(tbls) - 1
         while i - 1 >= 0:
@@ -758,9 +712,7 @@ class RAGFlowPdfParser:
                         if b.get("layout_type", "").find("caption") >= 0:
                             continue
                         y_dis = self._y_dis(c, b)
-                        x_dis = self._x_dis(
-                            c, b) if not x_overlapped(
-                            c, b) else 0
+                        x_dis = self._x_dis(c, b) if not x_overlapped(c, b) else 0
                         dis = y_dis * y_dis + x_dis * x_dis
                         if dis < minv:
                             mink = k
@@ -774,18 +726,10 @@ class RAGFlowPdfParser:
             #    continue
             if tv < fv and tk:
                 tables[tk].insert(0, c)
-                logging.debug(
-                    "TABLE:" +
-                    self.boxes[i]["text"] +
-                    "; Cap: " +
-                    tk)
+                logging.debug("TABLE:" + self.boxes[i]["text"] + "; Cap: " + tk)
             elif fk:
                 figures[fk].insert(0, c)
-                logging.debug(
-                    "FIGURE:" +
-                    self.boxes[i]["text"] +
-                    "; Cap: " +
-                    tk)
+                logging.debug("FIGURE:" + self.boxes[i]["text"] + "; Cap: " + tk)
             self.boxes.pop(i)
 
         def cropout(bxs, ltype, poss):
@@ -794,29 +738,19 @@ class RAGFlowPdfParser:
             if len(pn) < 2:
                 pn = list(pn)[0]
                 ht = self.page_cum_height[pn]
-                b = {
-                    "x0": np.min([b["x0"] for b in bxs]),
-                    "top": np.min([b["top"] for b in bxs]) - ht,
-                    "x1": np.max([b["x1"] for b in bxs]),
-                    "bottom": np.max([b["bottom"] for b in bxs]) - ht
-                }
+                b = {"x0": np.min([b["x0"] for b in bxs]), "top": np.min([b["top"] for b in bxs]) - ht, "x1": np.max([b["x1"] for b in bxs]), "bottom": np.max([b["bottom"] for b in bxs]) - ht}
                 louts = [layout for layout in self.page_layout[pn] if layout["type"] == ltype]
                 ii = Recognizer.find_overlapped(b, louts, naive=True)
                 if ii is not None:
                     b = louts[ii]
                 else:
-                    logging.warning(
-                        f"Missing layout match: {pn + 1},%s" %
-                        (bxs[0].get(
-                            "layoutno", "")))
+                    logging.warning(f"Missing layout match: {pn + 1},%s" % (bxs[0].get("layoutno", "")))
 
                 left, top, right, bott = b["x0"], b["top"], b["x1"], b["bottom"]
                 if right < left:
                     right = left + 1
                 poss.append((pn + self.page_from, left, right, top, bott))
-                return self.page_images[pn] \
-                    .crop((left * ZM, top * ZM,
-                           right * ZM, bott * ZM))
+                return self.page_images[pn].crop((left * ZM, top * ZM, right * ZM, bott * ZM))
             pn = {}
             for b in bxs:
                 p = b["page_number"] - 1
@@ -825,10 +759,7 @@ class RAGFlowPdfParser:
                 pn[p].append(b)
             pn = sorted(pn.items(), key=lambda x: x[0])
             imgs = [cropout(arr, ltype, poss) for p, arr in pn]
-            pic = Image.new("RGB",
-                            (int(np.max([i.size[0] for i in imgs])),
-                             int(np.sum([m.size[1] for m in imgs]))),
-                            (245, 245, 245))
+            pic = Image.new("RGB", (int(np.max([i.size[0] for i in imgs])), int(np.sum([m.size[1] for m in imgs]))), (245, 245, 245))
             height = 0
             for img in imgs:
                 pic.paste(img, (0, int(height)))
@@ -848,30 +779,20 @@ class RAGFlowPdfParser:
             poss = []
 
             if separate_tables_figures:
-                figure_results.append(
-                    (cropout(
-                        bxs,
-                        "figure", poss),
-                     [txt]))
+                figure_results.append((cropout(bxs, "figure", poss), [txt]))
                 figure_positions.append(poss)
             else:
-                res.append(
-                    (cropout(
-                        bxs,
-                        "figure", poss),
-                     [txt]))
+                res.append((cropout(bxs, "figure", poss), [txt]))
                 positions.append(poss)
 
         for k, bxs in tables.items():
             if not bxs:
                 continue
-            bxs = Recognizer.sort_Y_firstly(bxs, np.mean(
-                [(b["bottom"] - b["top"]) / 2 for b in bxs]))
+            bxs = Recognizer.sort_Y_firstly(bxs, np.mean([(b["bottom"] - b["top"]) / 2 for b in bxs]))
 
             poss = []
 
-            res.append((cropout(bxs, "table", poss),
-                        self.tbl_det.construct_table(bxs, html=return_html, is_english=self.is_english)))
+            res.append((cropout(bxs, "table", poss), self.tbl_det.construct_table(bxs, html=return_html, is_english=self.is_english)))
             positions.append(poss)
 
         if separate_tables_figures:
@@ -905,7 +826,7 @@ class RAGFlowPdfParser:
             (r"[0-9]+）", 10),
             (r"[\(（][0-9]+[）\)]", 11),
             (r"[零一二三四五六七八九十百]+是", 12),
-            (r"[⚫•➢✓]", 12)
+            (r"[⚫•➢✓]", 12),
         ]:
             if re.match(p, line):
                 return j
@@ -924,12 +845,9 @@ class RAGFlowPdfParser:
             if pn[-1] - 1 >= page_images_cnt:
                 return ""
 
-        return "@@{}\t{:.1f}\t{:.1f}\t{:.1f}\t{:.1f}##" \
-            .format("-".join([str(p) for p in pn]),
-                    bx["x0"], bx["x1"], top, bott)
+        return "@@{}\t{:.1f}\t{:.1f}\t{:.1f}\t{:.1f}##".format("-".join([str(p) for p in pn]), bx["x0"], bx["x1"], top, bott)
 
     def __filterout_scraps(self, boxes, ZM):
-
         def width(b):
             return b["x1"] - b["x0"]
 
@@ -939,8 +857,7 @@ class RAGFlowPdfParser:
         def usefull(b):
             if b.get("layout_type"):
                 return True
-            if width(
-                    b) > self.page_images[b["page_number"] - 1].size[0] / ZM / 3:
+            if width(b) > self.page_images[b["page_number"] - 1].size[0] / ZM / 3:
                 return True
             if b["bottom"] - b["top"] > self.mean_height[b["page_number"] - 1]:
                 return True
@@ -952,31 +869,23 @@ class RAGFlowPdfParser:
             widths = []
             pw = self.page_images[boxes[0]["page_number"] - 1].size[0] / ZM
             mh = self.mean_height[boxes[0]["page_number"] - 1]
-            mj = self.proj_match(
-                boxes[0]["text"]) or boxes[0].get(
-                "layout_type",
-                "") == "title"
+            mj = self.proj_match(boxes[0]["text"]) or boxes[0].get("layout_type", "") == "title"
 
             def dfs(line, st):
                 nonlocal mh, pw, lines, widths
                 lines.append(line)
                 widths.append(width(line))
-                mmj = self.proj_match(
-                    line["text"]) or line.get(
-                    "layout_type",
-                    "") == "title"
+                mmj = self.proj_match(line["text"]) or line.get("layout_type", "") == "title"
                 for i in range(st + 1, min(st + 20, len(boxes))):
                     if (boxes[i]["page_number"] - line["page_number"]) > 0:
                         break
-                    if not mmj and self._y_dis(
-                            line, boxes[i]) >= 3 * mh and height(line) < 1.5 * mh:
+                    if not mmj and self._y_dis(line, boxes[i]) >= 3 * mh and height(line) < 1.5 * mh:
                         break
 
                     if not usefull(boxes[i]):
                         continue
-                    if mmj or \
-                            (self._x_dis(boxes[i], line) < pw / 10): \
-                            # and abs(width(boxes[i])-width_mean)/max(width(boxes[i]),width_mean)<0.5):
+                    if mmj or (self._x_dis(boxes[i], line) < pw / 10):
+                        # and abs(width(boxes[i])-width_mean)/max(width(boxes[i]),width_mean)<0.5):
                         # concat following
                         dfs(boxes[i], i)
                         boxes.pop(i)
@@ -992,11 +901,9 @@ class RAGFlowPdfParser:
             boxes.pop(0)
             mw = np.mean(widths)
             if mj or mw / pw >= 0.35 or mw > 200:
-                res.append(
-                    "\n".join([c["text"] + self._line_tag(c, ZM) for c in lines]))
+                res.append("\n".join([c["text"] + self._line_tag(c, ZM) for c in lines]))
             else:
-                logging.debug("REMOVED: " +
-                              "<<".join([c["text"] for c in lines]))
+                logging.debug("REMOVED: " + "<<".join([c["text"] for c in lines]))
 
         return "\n\n".join(res)
 
@@ -1004,16 +911,14 @@ class RAGFlowPdfParser:
     def total_page_number(fnm, binary=None):
         try:
             with sys.modules[LOCK_KEY_pdfplumber]:
-                pdf = pdfplumber.open(
-                    fnm) if not binary else pdfplumber.open(BytesIO(binary))
+                pdf = pdfplumber.open(fnm) if not binary else pdfplumber.open(BytesIO(binary))
             total_page = len(pdf.pages)
             pdf.close()
             return total_page
         except Exception:
             logging.exception("total_page_number")
 
-    def __images__(self, fnm, zoomin=3, page_from=0,
-                   page_to=299, callback=None):
+    def __images__(self, fnm, zoomin=3, page_from=0, page_to=299, callback=None):
         self.lefted_chars = []
         self.mean_height = []
         self.mean_width = []
@@ -1025,10 +930,9 @@ class RAGFlowPdfParser:
         start = timer()
         try:
             with sys.modules[LOCK_KEY_pdfplumber]:
-                with (pdfplumber.open(fnm) if isinstance(fnm, str) else pdfplumber.open(BytesIO(fnm))) as pdf:
+                with pdfplumber.open(fnm) if isinstance(fnm, str) else pdfplumber.open(BytesIO(fnm)) as pdf:
                     self.pdf = pdf
-                    self.page_images = [p.to_image(resolution=72 * zoomin, antialias=True).annotated for i, p in
-                                        enumerate(self.pdf.pages[page_from:page_to])]
+                    self.page_images = [p.to_image(resolution=72 * zoomin, antialias=True).annotated for i, p in enumerate(self.pdf.pages[page_from:page_to])]
 
                     try:
                         self.page_chars = [[c for c in page.dedupe_chars().chars if self._has_color(c)] for page in self.pdf.pages[page_from:page_to]]
@@ -1044,11 +948,11 @@ class RAGFlowPdfParser:
 
         self.outlines = []
         try:
-            with (pdf2_read(fnm if isinstance(fnm, str)
-                            else BytesIO(fnm))) as pdf:
+            with pdf2_read(fnm if isinstance(fnm, str) else BytesIO(fnm)) as pdf:
                 self.pdf = pdf
 
                 outlines = self.pdf.outline
+
                 def dfs(arr, depth):
                     for a in arr:
                         if isinstance(a, dict):
@@ -1065,11 +969,11 @@ class RAGFlowPdfParser:
             logging.warning("Miss outlines")
 
         logging.debug("Images converted.")
-        self.is_english = [re.search(r"[a-zA-Z0-9,/¸;:'\[\]\(\)!@#$%^&*\"?<>._-]{30,}", "".join(
-            random.choices([c["text"] for c in self.page_chars[i]], k=min(100, len(self.page_chars[i]))))) for i in
-            range(len(self.page_chars))]
-        if sum([1 if e else 0 for e in self.is_english]) > len(
-                self.page_images) / 2:
+        self.is_english = [
+            re.search(r"[a-zA-Z0-9,/¸;:'\[\]\(\)!@#$%^&*\"?<>._-]{30,}", "".join(random.choices([c["text"] for c in self.page_chars[i]], k=min(100, len(self.page_chars[i])))))
+            for i in range(len(self.page_chars))
+        ]
+        if sum([1 if e else 0 for e in self.is_english]) > len(self.page_images) / 2:
             self.is_english = True
         else:
             self.is_english = False
@@ -1077,10 +981,12 @@ class RAGFlowPdfParser:
         async def __img_ocr(i, id, img, chars, limiter):
             j = 0
             while j + 1 < len(chars):
-                if chars[j]["text"] and chars[j + 1]["text"] \
-                        and re.match(r"[0-9a-zA-Z,.:;!%]+", chars[j]["text"] + chars[j + 1]["text"]) \
-                        and chars[j + 1]["x0"] - chars[j]["x1"] >= min(chars[j + 1]["width"],
-                                                                       chars[j]["width"]) / 2:
+                if (
+                    chars[j]["text"]
+                    and chars[j + 1]["text"]
+                    and re.match(r"[0-9a-zA-Z,.:;!%]+", chars[j]["text"] + chars[j + 1]["text"])
+                    and chars[j + 1]["x0"] - chars[j]["x1"] >= min(chars[j + 1]["width"], chars[j]["width"]) / 2
+                ):
                     chars[j]["text"] += " "
                 j += 1
 
@@ -1096,12 +1002,8 @@ class RAGFlowPdfParser:
         async def __img_ocr_launcher():
             def __ocr_preprocess():
                 chars = self.page_chars[i] if not self.is_english else []
-                self.mean_height.append(
-                    np.median(sorted([c["height"] for c in chars])) if chars else 0
-                )
-                self.mean_width.append(
-                    np.median(sorted([c["width"] for c in chars])) if chars else 8
-                )
+                self.mean_height.append(np.median(sorted([c["height"] for c in chars])) if chars else 0)
+                self.mean_width.append(np.median(sorted([c["width"] for c in chars])) if chars else 8)
                 self.page_cum_height.append(img.size[1] / zoomin)
                 return chars
 
@@ -1110,8 +1012,7 @@ class RAGFlowPdfParser:
                     for i, img in enumerate(self.page_images):
                         chars = __ocr_preprocess()
 
-                        nursery.start_soon(__img_ocr, i, i % PARALLEL_DEVICES, img, chars,
-                                           self.parallel_limiter[i % PARALLEL_DEVICES])
+                        nursery.start_soon(__img_ocr, i, i % PARALLEL_DEVICES, img, chars, self.parallel_limiter[i % PARALLEL_DEVICES])
                         await trio.sleep(0.1)
             else:
                 for i, img in enumerate(self.page_images):
@@ -1124,11 +1025,9 @@ class RAGFlowPdfParser:
 
         logging.info(f"__images__ {len(self.page_images)} pages cost {timer() - start}s")
 
-        if not self.is_english and not any(
-                [c for c in self.page_chars]) and self.boxes:
+        if not self.is_english and not any([c for c in self.page_chars]) and self.boxes:
             bxes = [b for bxs in self.boxes for b in bxs]
-            self.is_english = re.search(r"[\na-zA-Z0-9,/¸;:'\[\]\(\)!@#$%^&*\"?<>._-]{30,}",
-                                        "".join([b["text"] for b in random.choices(bxes, k=min(30, len(bxes)))]))
+            self.is_english = re.search(r"[\na-zA-Z0-9,/¸;:'\[\]\(\)!@#$%^&*\"?<>._-]{30,}", "".join([b["text"] for b in random.choices(bxes, k=min(30, len(bxes)))]))
 
         logging.debug("Is it English:", self.is_english)
 
@@ -1144,8 +1043,7 @@ class RAGFlowPdfParser:
         self._text_merge()
         self._concat_downward()
         self._filter_forpages()
-        tbls = self._extract_table_figure(
-            need_image, zoomin, return_html, False)
+        tbls = self._extract_table_figure(need_image, zoomin, return_html, False)
         return self.__filterout_scraps(deepcopy(self.boxes), zoomin), tbls
 
     def parse_into_bboxes(self, fnm, callback=None, zoomin=3):
@@ -1177,11 +1075,11 @@ class RAGFlowPdfParser:
         def insert_table_figures(tbls_or_figs, layout_type):
             def min_rectangle_distance(rect1, rect2):
                 import math
+
                 pn1, left1, right1, top1, bottom1 = rect1
                 pn2, left2, right2, top2, bottom2 = rect2
-                if (right1 >= left2 and right2 >= left1 and
-                        bottom1 >= top2 and bottom2 >= top1):
-                    return 0 + (pn1-pn2)*10000
+                if right1 >= left2 and right2 >= left1 and bottom1 >= top2 and bottom2 >= top1:
+                    return 0 + (pn1 - pn2) * 10000
                 if right1 < left2:
                     dx = left2 - right1
                 elif right2 < left1:
@@ -1194,18 +1092,16 @@ class RAGFlowPdfParser:
                     dy = top1 - bottom2
                 else:
                     dy = 0
-                return math.sqrt(dx*dx + dy*dy) + (pn1-pn2)*10000
+                return math.sqrt(dx * dx + dy * dy) + (pn1 - pn2) * 10000
 
             for (img, txt), poss in tbls_or_figs:
                 bboxes = [(i, (b["page_number"], b["x0"], b["x1"], b["top"], b["bottom"])) for i, b in enumerate(self.boxes)]
-                dists = [(min_rectangle_distance((pn, left, right, top, bott), rect),i) for i, rect in bboxes for pn, left, right, top, bott in poss]
+                dists = [(min_rectangle_distance((pn, left, right, top, bott), rect), i) for i, rect in bboxes for pn, left, right, top, bott in poss]
                 min_i = np.argmin(dists, axis=0)[0]
                 min_i, rect = bboxes[dists[min_i][-1]]
                 if isinstance(txt, list):
                     txt = "\n".join(txt)
-                self.boxes.insert(min_i, {
-                    "page_number": rect[0], "x0": rect[1], "x1": rect[2], "top": rect[3], "bottom": rect[4], "layout_type": layout_type, "text": txt, "image": img
-                })
+                self.boxes.insert(min_i, {"page_number": rect[0], "x0": rect[1], "x1": rect[2], "top": rect[3], "bottom": rect[4], "layout_type": layout_type, "text": txt, "image": img})
 
         for b in self.boxes:
             b["position_tag"] = self._line_tag(b, zoomin)
@@ -1225,12 +1121,9 @@ class RAGFlowPdfParser:
     def extract_positions(txt):
         poss = []
         for tag in re.findall(r"@@[0-9-]+\t[0-9.\t]+##", txt):
-            pn, left, right, top, bottom = tag.strip(
-                "#").strip("@").split("\t")
-            left, right, top, bottom = float(left), float(
-                right), float(top), float(bottom)
-            poss.append(([int(p) - 1 for p in pn.split("-")],
-                         left, right, top, bottom))
+            pn, left, right, top, bottom = tag.strip("#").strip("@").split("\t")
+            left, right, top, bottom = float(left), float(right), float(top), float(bottom)
+            poss.append(([int(p) - 1 for p in pn.split("-")], left, right, top, bottom))
         return poss
 
     def crop(self, text, ZM=3, need_position=False):
@@ -1241,15 +1134,12 @@ class RAGFlowPdfParser:
                 return None, None
             return
 
-        max_width = max(
-            np.max([right - left for (_, left, right, _, _) in poss]), 6)
+        max_width = max(np.max([right - left for (_, left, right, _, _) in poss]), 6)
         GAP = 6
         pos = poss[0]
-        poss.insert(0, ([pos[0][0]], pos[1], pos[2], max(
-            0, pos[3] - 120), max(pos[3] - GAP, 0)))
+        poss.insert(0, ([pos[0][0]], pos[1], pos[2], max(0, pos[3] - 120), max(pos[3] - GAP, 0)))
         pos = poss[-1]
-        poss.append(([pos[0][-1]], pos[1], pos[2], min(self.page_images[pos[0][-1]].size[1] / ZM, pos[4] + GAP),
-                     min(self.page_images[pos[0][-1]].size[1] / ZM, pos[4] + 120)))
+        poss.append(([pos[0][-1]], pos[1], pos[2], min(self.page_images[pos[0][-1]].size[1] / ZM, pos[4] + GAP), min(self.page_images[pos[0][-1]].size[1] / ZM, pos[4] + 120)))
 
         positions = []
         for ii, (pns, left, right, top, bottom) in enumerate(poss):
@@ -1257,28 +1147,14 @@ class RAGFlowPdfParser:
             bottom *= ZM
             for pn in pns[1:]:
                 bottom += self.page_images[pn - 1].size[1]
-            imgs.append(
-                self.page_images[pns[0]].crop((left * ZM, top * ZM,
-                                               right *
-                                               ZM, min(
-                                                   bottom, self.page_images[pns[0]].size[1])
-                                               ))
-            )
+            imgs.append(self.page_images[pns[0]].crop((left * ZM, top * ZM, right * ZM, min(bottom, self.page_images[pns[0]].size[1]))))
             if 0 < ii < len(poss) - 1:
-                positions.append((pns[0] + self.page_from, left, right, top, min(
-                    bottom, self.page_images[pns[0]].size[1]) / ZM))
+                positions.append((pns[0] + self.page_from, left, right, top, min(bottom, self.page_images[pns[0]].size[1]) / ZM))
             bottom -= self.page_images[pns[0]].size[1]
             for pn in pns[1:]:
-                imgs.append(
-                    self.page_images[pn].crop((left * ZM, 0,
-                                               right * ZM,
-                                               min(bottom,
-                                                   self.page_images[pn].size[1])
-                                               ))
-                )
+                imgs.append(self.page_images[pn].crop((left * ZM, 0, right * ZM, min(bottom, self.page_images[pn].size[1]))))
                 if 0 < ii < len(poss) - 1:
-                    positions.append((pn + self.page_from, left, right, 0, min(
-                        bottom, self.page_images[pn].size[1]) / ZM))
+                    positions.append((pn + self.page_from, left, right, 0, min(bottom, self.page_images[pn].size[1]) / ZM))
                 bottom -= self.page_images[pn].size[1]
 
         if not imgs:
@@ -1290,14 +1166,12 @@ class RAGFlowPdfParser:
             height += img.size[1] + GAP
         height = int(height)
         width = int(np.max([i.size[0] for i in imgs]))
-        pic = Image.new("RGB",
-                        (width, height),
-                        (245, 245, 245))
+        pic = Image.new("RGB", (width, height), (245, 245, 245))
         height = 0
         for ii, img in enumerate(imgs):
             if ii == 0 or ii + 1 == len(imgs):
-                img = img.convert('RGBA')
-                overlay = Image.new('RGBA', img.size, (0, 0, 0, 0))
+                img = img.convert("RGBA")
+                overlay = Image.new("RGBA", img.size, (0, 0, 0, 0))
                 overlay.putalpha(128)
                 img = Image.alpha_composite(img, overlay).convert("RGB")
             pic.paste(img, (0, int(height)))
@@ -1312,14 +1186,12 @@ class RAGFlowPdfParser:
         pn = bx["page_number"]
         top = bx["top"] - self.page_cum_height[pn - 1]
         bott = bx["bottom"] - self.page_cum_height[pn - 1]
-        poss.append((pn, bx["x0"], bx["x1"], top, min(
-            bott, self.page_images[pn - 1].size[1] / ZM)))
+        poss.append((pn, bx["x0"], bx["x1"], top, min(bott, self.page_images[pn - 1].size[1] / ZM)))
         while bott * ZM > self.page_images[pn - 1].size[1]:
             bott -= self.page_images[pn - 1].size[1] / ZM
             top = 0
             pn += 1
-            poss.append((pn, bx["x0"], bx["x1"], top, min(
-                bott, self.page_images[pn - 1].size[1] / ZM)))
+            poss.append((pn, bx["x0"], bx["x1"], top, min(bott, self.page_images[pn - 1].size[1] / ZM)))
         return poss
 
 
@@ -1328,9 +1200,7 @@ class PlainParser:
         self.outlines = []
         lines = []
         try:
-            self.pdf = pdf2_read(
-                filename if isinstance(
-                    filename, str) else BytesIO(filename))
+            self.pdf = pdf2_read(filename if isinstance(filename, str) else BytesIO(filename))
             for page in self.pdf.pages[from_page:to_page]:
                 lines.extend([t for t in page.extract_text().split("\n")])
 
@@ -1367,10 +1237,8 @@ class VisionParser(RAGFlowPdfParser):
     def __images__(self, fnm, zoomin=3, page_from=0, page_to=299, callback=None):
         try:
             with sys.modules[LOCK_KEY_pdfplumber]:
-                self.pdf = pdfplumber.open(fnm) if isinstance(
-                    fnm, str) else pdfplumber.open(BytesIO(fnm))
-                self.page_images = [p.to_image(resolution=72 * zoomin).annotated for i, p in
-                                    enumerate(self.pdf.pages[page_from:page_to])]
+                self.pdf = pdfplumber.open(fnm) if isinstance(fnm, str) else pdfplumber.open(BytesIO(fnm))
+                self.page_images = [p.to_image(resolution=72 * zoomin).annotated for i, p in enumerate(self.pdf.pages[page_from:page_to])]
                 self.total_page = len(self.pdf.pages)
         except Exception:
             self.page_images = None
@@ -1397,15 +1265,15 @@ class VisionParser(RAGFlowPdfParser):
             text = picture_vision_llm_chunk(
                 binary=img_binary,
                 vision_model=self.vision_model,
-                prompt=vision_llm_describe_prompt(page=pdf_page_num+1),
+                prompt=vision_llm_describe_prompt(page=pdf_page_num + 1),
                 callback=callback,
             )
             if kwargs.get("callback"):
-                kwargs["callback"](idx*1./len(self.page_images), f"Processed: {idx+1}/{len(self.page_images)}")
+                kwargs["callback"](idx * 1.0 / len(self.page_images), f"Processed: {idx + 1}/{len(self.page_images)}")
 
             if text:
                 width, height = self.page_images[idx].size
-                all_docs.append((text, f"{pdf_page_num+1} 0 {width/zoomin} 0 {height/zoomin}"))
+                all_docs.append((text, f"{pdf_page_num + 1} 0 {width / zoomin} 0 {height / zoomin}"))
         return all_docs, []
 
 

deepdoc/vision/__init__.py

@@ -16,24 +16,28 @@
 import io
 import sys
 import threading
+
 import pdfplumber
 
 from .ocr import OCR
 from .recognizer import Recognizer
+from .layout_recognizer import AscendLayoutRecognizer
 from .layout_recognizer import LayoutRecognizer4YOLOv10 as LayoutRecognizer
 from .table_structure_recognizer import TableStructureRecognizer
 
-
 LOCK_KEY_pdfplumber = "global_shared_lock_pdfplumber"
 if LOCK_KEY_pdfplumber not in sys.modules:
     sys.modules[LOCK_KEY_pdfplumber] = threading.Lock()
 
 
 def init_in_out(args):
-    from PIL import Image
     import os
     import traceback
+
+    from PIL import Image
+
     from api.utils.file_utils import traversal_files
+
     images = []
     outputs = []
 
@@ -44,8 +48,7 @@ def init_in_out(args):
         nonlocal outputs, images
         with sys.modules[LOCK_KEY_pdfplumber]:
             pdf = pdfplumber.open(fnm)
-            images = [p.to_image(resolution=72 * zoomin).annotated for i, p in
-                                enumerate(pdf.pages)]
+            images = [p.to_image(resolution=72 * zoomin).annotated for i, p in enumerate(pdf.pages)]
 
         for i, page in enumerate(images):
             outputs.append(os.path.split(fnm)[-1] + f"_{i}.jpg")
@@ -57,10 +60,10 @@ def init_in_out(args):
             pdf_pages(fnm)
             return
         try:
-            fp = open(fnm, 'rb')
+            fp = open(fnm, "rb")
             binary = fp.read()
             fp.close()
-            images.append(Image.open(io.BytesIO(binary)).convert('RGB'))
+            images.append(Image.open(io.BytesIO(binary)).convert("RGB"))
             outputs.append(os.path.split(fnm)[-1])
         except Exception:
             traceback.print_exc()
@@ -81,6 +84,7 @@ __all__ = [
     "OCR",
     "Recognizer",
     "LayoutRecognizer",
+    "AscendLayoutRecognizer",
     "TableStructureRecognizer",
     "init_in_out",
 ]

deepdoc/vision/layout_recognizer.py

@@ -14,6 +14,8 @@
 #  limitations under the License.
 #
 
+import logging
+import math
 import os
 import re
 from collections import Counter
@@ -45,28 +47,22 @@ class LayoutRecognizer(Recognizer):
 
     def __init__(self, domain):
         try:
-            model_dir = os.path.join(
-                get_project_base_directory(),
-                "rag/res/deepdoc")
+            model_dir = os.path.join(get_project_base_directory(), "rag/res/deepdoc")
             super().__init__(self.labels, domain, model_dir)
         except Exception:
-            model_dir = snapshot_download(repo_id="InfiniFlow/deepdoc",
-                                          local_dir=os.path.join(get_project_base_directory(), "rag/res/deepdoc"),
-                                          local_dir_use_symlinks=False)
+            model_dir = snapshot_download(repo_id="InfiniFlow/deepdoc", local_dir=os.path.join(get_project_base_directory(), "rag/res/deepdoc"), local_dir_use_symlinks=False)
             super().__init__(self.labels, domain, model_dir)
 
         self.garbage_layouts = ["footer", "header", "reference"]
         self.client = None
         if os.environ.get("TENSORRT_DLA_SVR"):
             from deepdoc.vision.dla_cli import DLAClient
+
             self.client = DLAClient(os.environ["TENSORRT_DLA_SVR"])
 
     def __call__(self, image_list, ocr_res, scale_factor=3, thr=0.2, batch_size=16, drop=True):
         def __is_garbage(b):
-            patt = [r"^•+$", "^[0-9]{1,2} / ?[0-9]{1,2}$",
-                    r"^[0-9]{1,2} of [0-9]{1,2}$", "^http://[^ ]{12,}",
-                    "\\(cid *: *[0-9]+ *\\)"
-                    ]
+            patt = [r"^•+$", "^[0-9]{1,2} / ?[0-9]{1,2}$", r"^[0-9]{1,2} of [0-9]{1,2}$", "^http://[^ ]{12,}", "\\(cid *: *[0-9]+ *\\)"]
             return any([re.search(p, b["text"]) for p in patt])
 
         if self.client:
@@ -82,18 +78,23 @@ class LayoutRecognizer(Recognizer):
         page_layout = []
         for pn, lts in enumerate(layouts):
             bxs = ocr_res[pn]
-            lts = [{"type": b["type"],
+            lts = [
+                {
+                    "type": b["type"],
                     "score": float(b["score"]),
-                    "x0": b["bbox"][0] / scale_factor, "x1": b["bbox"][2] / scale_factor,
-                    "top": b["bbox"][1] / scale_factor, "bottom": b["bbox"][-1] / scale_factor,
+                    "x0": b["bbox"][0] / scale_factor,
+                    "x1": b["bbox"][2] / scale_factor,
+                    "top": b["bbox"][1] / scale_factor,
+                    "bottom": b["bbox"][-1] / scale_factor,
                     "page_number": pn,
-                    } for b in lts if float(b["score"]) >= 0.4 or b["type"] not in self.garbage_layouts]
-            lts = self.sort_Y_firstly(lts, np.mean(
-                [lt["bottom"] - lt["top"] for lt in lts]) / 2)
+                }
+                for b in lts
+                if float(b["score"]) >= 0.4 or b["type"] not in self.garbage_layouts
+            ]
+            lts = self.sort_Y_firstly(lts, np.mean([lt["bottom"] - lt["top"] for lt in lts]) / 2)
             lts = self.layouts_cleanup(bxs, lts)
             page_layout.append(lts)
 
-            # Tag layout type, layouts are ready
             def findLayout(ty):
                 nonlocal bxs, lts, self
                 lts_ = [lt for lt in lts if lt["type"] == ty]
@@ -106,21 +107,17 @@ class LayoutRecognizer(Recognizer):
                         bxs.pop(i)
                         continue
 
-                    ii = self.find_overlapped_with_threshold(bxs[i], lts_,
-                                                              thr=0.4)
-                    if ii is None:  # belong to nothing
+                    ii = self.find_overlapped_with_threshold(bxs[i], lts_, thr=0.4)
+                    if ii is None:
                         bxs[i]["layout_type"] = ""
                         i += 1
                         continue
                     lts_[ii]["visited"] = True
                     keep_feats = [
-                        lts_[
-                            ii]["type"] == "footer" and bxs[i]["bottom"] < image_list[pn].size[1] * 0.9 / scale_factor,
-                        lts_[
-                            ii]["type"] == "header" and bxs[i]["top"] > image_list[pn].size[1] * 0.1 / scale_factor,
+                        lts_[ii]["type"] == "footer" and bxs[i]["bottom"] < image_list[pn].size[1] * 0.9 / scale_factor,
+                        lts_[ii]["type"] == "header" and bxs[i]["top"] > image_list[pn].size[1] * 0.1 / scale_factor,
                     ]
-                    if drop and lts_[
-                            ii]["type"] in self.garbage_layouts and not any(keep_feats):
+                    if drop and lts_[ii]["type"] in self.garbage_layouts and not any(keep_feats):
                         if lts_[ii]["type"] not in garbages:
                             garbages[lts_[ii]["type"]] = []
                         garbages[lts_[ii]["type"]].append(bxs[i]["text"])
@@ -128,17 +125,14 @@ class LayoutRecognizer(Recognizer):
                         continue
 
                     bxs[i]["layoutno"] = f"{ty}-{ii}"
-                    bxs[i]["layout_type"] = lts_[ii]["type"] if lts_[
-                        ii]["type"] != "equation" else "figure"
+                    bxs[i]["layout_type"] = lts_[ii]["type"] if lts_[ii]["type"] != "equation" else "figure"
                     i += 1
 
-            for lt in ["footer", "header", "reference", "figure caption",
-                       "table caption", "title", "table", "text", "figure", "equation"]:
+            for lt in ["footer", "header", "reference", "figure caption", "table caption", "title", "table", "text", "figure", "equation"]:
                 findLayout(lt)
 
             # add box to figure layouts which has not text box
-            for i, lt in enumerate(
-                    [lt for lt in lts if lt["type"] in ["figure", "equation"]]):
+            for i, lt in enumerate([lt for lt in lts if lt["type"] in ["figure", "equation"]]):
                 if lt.get("visited"):
                     continue
                 lt = deepcopy(lt)
@@ -206,13 +200,11 @@ class LayoutRecognizer4YOLOv10(LayoutRecognizer):
             img = cv2.resize(img, new_unpad, interpolation=cv2.INTER_LINEAR)
             top, bottom = int(round(dh - 0.1)) if self.center else 0, int(round(dh + 0.1))
             left, right = int(round(dw - 0.1)) if self.center else 0, int(round(dw + 0.1))
-            img = cv2.copyMakeBorder(
-                img, top, bottom, left, right, cv2.BORDER_CONSTANT, value=(114, 114, 114)
-            )  # add border
+            img = cv2.copyMakeBorder(img, top, bottom, left, right, cv2.BORDER_CONSTANT, value=(114, 114, 114))  # add border
             img /= 255.0
             img = img.transpose(2, 0, 1)
             img = img[np.newaxis, :, :, :].astype(np.float32)
-            inputs.append({self.input_names[0]: img, "scale_factor": [shape[1]/ww, shape[0]/hh, dw, dh]})
+            inputs.append({self.input_names[0]: img, "scale_factor": [shape[1] / ww, shape[0] / hh, dw, dh]})
 
         return inputs
 
@@ -230,8 +222,7 @@ class LayoutRecognizer4YOLOv10(LayoutRecognizer):
         boxes[:, 2] -= inputs["scale_factor"][2]
         boxes[:, 1] -= inputs["scale_factor"][3]
         boxes[:, 3] -= inputs["scale_factor"][3]
-        input_shape = np.array([inputs["scale_factor"][0], inputs["scale_factor"][1], inputs["scale_factor"][0],
-                                inputs["scale_factor"][1]])
+        input_shape = np.array([inputs["scale_factor"][0], inputs["scale_factor"][1], inputs["scale_factor"][0], inputs["scale_factor"][1]])
         boxes = np.multiply(boxes, input_shape, dtype=np.float32)
 
         unique_class_ids = np.unique(class_ids)
@@ -243,8 +234,223 @@ class LayoutRecognizer4YOLOv10(LayoutRecognizer):
             class_keep_boxes = nms(class_boxes, class_scores, 0.45)
             indices.extend(class_indices[class_keep_boxes])
 
-        return [{
-            "type": self.label_list[class_ids[i]].lower(),
-            "bbox": [float(t) for t in boxes[i].tolist()],
-            "score": float(scores[i])
-        } for i in indices]
+        return [{"type": self.label_list[class_ids[i]].lower(), "bbox": [float(t) for t in boxes[i].tolist()], "score": float(scores[i])} for i in indices]
+
+
+class AscendLayoutRecognizer(Recognizer):
+    labels = [
+        "title",
+        "Text",
+        "Reference",
+        "Figure",
+        "Figure caption",
+        "Table",
+        "Table caption",
+        "Table caption",
+        "Equation",
+        "Figure caption",
+    ]
+
+    def __init__(self, domain):
+        from ais_bench.infer.interface import InferSession
+
+        model_dir = os.path.join(get_project_base_directory(), "rag/res/deepdoc")
+        model_file_path = os.path.join(model_dir, domain + ".om")
+
+        if not os.path.exists(model_file_path):
+            raise ValueError(f"Model file not found: {model_file_path}")
+
+        device_id = int(os.getenv("ASCEND_LAYOUT_RECOGNIZER_DEVICE_ID", 0))
+        self.session = InferSession(device_id=device_id, model_path=model_file_path)
+        self.input_shape = self.session.get_inputs()[0].shape[2:4]  # H,W
+        self.garbage_layouts = ["footer", "header", "reference"]
+
+    def preprocess(self, image_list):
+        inputs = []
+        H, W = self.input_shape
+        for img in image_list:
+            h, w = img.shape[:2]
+            img = cv2.cvtColor(img, cv2.COLOR_BGR2RGB).astype(np.float32)
+
+            r = min(H / h, W / w)
+            new_unpad = (int(round(w * r)), int(round(h * r)))
+            dw, dh = (W - new_unpad[0]) / 2.0, (H - new_unpad[1]) / 2.0
+
+            img = cv2.resize(img, new_unpad, interpolation=cv2.INTER_LINEAR)
+            top, bottom = int(round(dh - 0.1)), int(round(dh + 0.1))
+            left, right = int(round(dw - 0.1)), int(round(dw + 0.1))
+            img = cv2.copyMakeBorder(img, top, bottom, left, right, cv2.BORDER_CONSTANT, value=(114, 114, 114))
+
+            img /= 255.0
+            img = img.transpose(2, 0, 1)[np.newaxis, :, :, :].astype(np.float32)
+
+            inputs.append(
+                {
+                    "image": img,
+                    "scale_factor": [w / new_unpad[0], h / new_unpad[1]],
+                    "pad": [dw, dh],
+                    "orig_shape": [h, w],
+                }
+            )
+        return inputs
+
+    def postprocess(self, boxes, inputs, thr=0.25):
+        arr = np.squeeze(boxes)
+        if arr.ndim == 1:
+            arr = arr.reshape(1, -1)
+
+        results = []
+        if arr.shape[1] == 6:
+            # [x1,y1,x2,y2,score,cls]
+            m = arr[:, 4] >= thr
+            arr = arr[m]
+            if arr.size == 0:
+                return []
+            xyxy = arr[:, :4].astype(np.float32)
+            scores = arr[:, 4].astype(np.float32)
+            cls_ids = arr[:, 5].astype(np.int32)
+
+            if "pad" in inputs:
+                dw, dh = inputs["pad"]
+                sx, sy = inputs["scale_factor"]
+                xyxy[:, [0, 2]] -= dw
+                xyxy[:, [1, 3]] -= dh
+                xyxy *= np.array([sx, sy, sx, sy], dtype=np.float32)
+            else:
+                # backup
+                sx, sy = inputs["scale_factor"]
+                xyxy *= np.array([sx, sy, sx, sy], dtype=np.float32)
+
+            keep_indices = []
+            for c in np.unique(cls_ids):
+                idx = np.where(cls_ids == c)[0]
+                k = nms(xyxy[idx], scores[idx], 0.45)
+                keep_indices.extend(idx[k])
+
+            for i in keep_indices:
+                cid = int(cls_ids[i])
+                if 0 <= cid < len(self.labels):
+                    results.append({"type": self.labels[cid].lower(), "bbox": [float(t) for t in xyxy[i].tolist()], "score": float(scores[i])})
+            return results
+
+        raise ValueError(f"Unexpected output shape: {arr.shape}")
+
+    def __call__(self, image_list, ocr_res, scale_factor=3, thr=0.2, batch_size=16, drop=True):
+        import re
+        from collections import Counter
+
+        assert len(image_list) == len(ocr_res)
+
+        images = [np.array(im) if not isinstance(im, np.ndarray) else im for im in image_list]
+        layouts_all_pages = []  # list of list[{"type","score","bbox":[x1,y1,x2,y2]}]
+
+        conf_thr = max(thr, 0.08)
+
+        batch_loop_cnt = math.ceil(float(len(images)) / batch_size)
+        for bi in range(batch_loop_cnt):
+            s = bi * batch_size
+            e = min((bi + 1) * batch_size, len(images))
+            batch_images = images[s:e]
+
+            inputs_list = self.preprocess(batch_images)
+            logging.debug("preprocess done")
+
+            for ins in inputs_list:
+                feeds = [ins["image"]]
+                out_list = self.session.infer(feeds=feeds, mode="static")
+
+                for out in out_list:
+                    lts = self.postprocess(out, ins, conf_thr)
+
+                    page_lts = []
+                    for b in lts:
+                        if float(b["score"]) >= 0.4 or b["type"] not in self.garbage_layouts:
+                            x0, y0, x1, y1 = b["bbox"]
+                            page_lts.append(
+                                {
+                                    "type": b["type"],
+                                    "score": float(b["score"]),
+                                    "x0": float(x0) / scale_factor,
+                                    "x1": float(x1) / scale_factor,
+                                    "top": float(y0) / scale_factor,
+                                    "bottom": float(y1) / scale_factor,
+                                    "page_number": len(layouts_all_pages),
+                                }
+                            )
+                    layouts_all_pages.append(page_lts)
+
+        def _is_garbage_text(box):
+            patt = [r"^•+$", r"^[0-9]{1,2} / ?[0-9]{1,2}$", r"^[0-9]{1,2} of [0-9]{1,2}$", r"^http://[^ ]{12,}", r"\(cid *: *[0-9]+ *\)"]
+            return any(re.search(p, box.get("text", "")) for p in patt)
+
+        boxes_out = []
+        page_layout = []
+        garbages = {}
+
+        for pn, lts in enumerate(layouts_all_pages):
+            if lts:
+                avg_h = np.mean([lt["bottom"] - lt["top"] for lt in lts])
+                lts = self.sort_Y_firstly(lts, avg_h / 2 if avg_h > 0 else 0)
+
+            bxs = ocr_res[pn]
+            lts = self.layouts_cleanup(bxs, lts)
+            page_layout.append(lts)
+
+            def _tag_layout(ty):
+                nonlocal bxs, lts
+                lts_of_ty = [lt for lt in lts if lt["type"] == ty]
+                i = 0
+                while i < len(bxs):
+                    if bxs[i].get("layout_type"):
+                        i += 1
+                        continue
+                    if _is_garbage_text(bxs[i]):
+                        bxs.pop(i)
+                        continue
+
+                    ii = self.find_overlapped_with_threshold(bxs[i], lts_of_ty, thr=0.4)
+                    if ii is None:
+                        bxs[i]["layout_type"] = ""
+                        i += 1
+                        continue
+
+                    lts_of_ty[ii]["visited"] = True
+
+                    keep_feats = [
+                        lts_of_ty[ii]["type"] == "footer" and bxs[i]["bottom"] < image_list[pn].shape[0] * 0.9 / scale_factor,
+                        lts_of_ty[ii]["type"] == "header" and bxs[i]["top"] > image_list[pn].shape[0] * 0.1 / scale_factor,
+                    ]
+                    if drop and lts_of_ty[ii]["type"] in self.garbage_layouts and not any(keep_feats):
+                        garbages.setdefault(lts_of_ty[ii]["type"], []).append(bxs[i].get("text", ""))
+                        bxs.pop(i)
+                        continue
+
+                    bxs[i]["layoutno"] = f"{ty}-{ii}"
+                    bxs[i]["layout_type"] = lts_of_ty[ii]["type"] if lts_of_ty[ii]["type"] != "equation" else "figure"
+                    i += 1
+
+            for ty in ["footer", "header", "reference", "figure caption", "table caption", "title", "table", "text", "figure", "equation"]:
+                _tag_layout(ty)
+
+            figs = [lt for lt in lts if lt["type"] in ["figure", "equation"]]
+            for i, lt in enumerate(figs):
+                if lt.get("visited"):
+                    continue
+                lt = deepcopy(lt)
+                lt.pop("type", None)
+                lt["text"] = ""
+                lt["layout_type"] = "figure"
+                lt["layoutno"] = f"figure-{i}"
+                bxs.append(lt)
+
+            boxes_out.extend(bxs)
+
+        garbag_set = set()
+        for k, lst in garbages.items():
+            cnt = Counter(lst)
+            for g, c in cnt.items():
+                if c > 1:
+                    garbag_set.add(g)
+
+        ocr_res_new = [b for b in boxes_out if b["text"].strip() not in garbag_set]
+        return ocr_res_new, page_layout

deepdoc/vision/ocr.py

@@ -13,7 +13,7 @@
 #  See the License for the specific language governing permissions and
 #  limitations under the License.
 #
-
+import gc
 import logging
 import copy
 import time
@@ -348,6 +348,13 @@ class TextRecognizer:
 
         return img
 
+    def close(self):
+        # close session and release manually
+        logging.info('Close TextRecognizer.')
+        if hasattr(self, "predictor"):
+            del self.predictor
+        gc.collect()
+
     def __call__(self, img_list):
         img_num = len(img_list)
         # Calculate the aspect ratio of all text bars
@@ -395,6 +402,9 @@ class TextRecognizer:
 
         return rec_res, time.time() - st
 
+    def __del__(self):
+        self.close()
+
 
 class TextDetector:
     def __init__(self, model_dir, device_id: int | None = None):
@@ -479,6 +489,12 @@ class TextDetector:
         dt_boxes = np.array(dt_boxes_new)
         return dt_boxes
 
+    def close(self):
+        logging.info("Close TextDetector.")
+        if hasattr(self, "predictor"):
+            del self.predictor
+        gc.collect()
+
     def __call__(self, img):
         ori_im = img.copy()
         data = {'image': img}
@@ -508,6 +524,9 @@ class TextDetector:
 
         return dt_boxes, time.time() - st
 
+    def __del__(self):
+        self.close()
+
 
 class OCR:
     def __init__(self, model_dir=None):

deepdoc/vision/recognizer.py

@@ -13,7 +13,7 @@
 #  See the License for the specific language governing permissions and
 #  limitations under the License.
 #
-
+import gc
 import logging
 import os
 import math
@@ -406,6 +406,12 @@ class Recognizer:
             "score": float(scores[i])
         } for i in indices]
 
+    def close(self):
+        logging.info("Close recognizer.")
+        if hasattr(self, "ort_sess"):
+            del self.ort_sess
+        gc.collect()
+
     def __call__(self, image_list, thr=0.7, batch_size=16):
         res = []
         images = []
@@ -430,5 +436,7 @@ class Recognizer:
 
         return res
 
+    def __del__(self):
+        self.close()
 
 

deepdoc/vision/table_structure_recognizer.py

@@ -23,6 +23,7 @@ from huggingface_hub import snapshot_download
 
 from api.utils.file_utils import get_project_base_directory
 from rag.nlp import rag_tokenizer
+
 from .recognizer import Recognizer
 
 
@@ -38,31 +39,49 @@ class TableStructureRecognizer(Recognizer):
 
     def __init__(self):
         try:
-            super().__init__(self.labels, "tsr", os.path.join(
-                    get_project_base_directory(),
-                    "rag/res/deepdoc"))
+            super().__init__(self.labels, "tsr", os.path.join(get_project_base_directory(), "rag/res/deepdoc"))
         except Exception:
-            super().__init__(self.labels, "tsr", snapshot_download(repo_id="InfiniFlow/deepdoc",
-                                              local_dir=os.path.join(get_project_base_directory(), "rag/res/deepdoc"),
-                                              local_dir_use_symlinks=False))
+            super().__init__(
+                self.labels,
+                "tsr",
+                snapshot_download(
+                    repo_id="InfiniFlow/deepdoc",
+                    local_dir=os.path.join(get_project_base_directory(), "rag/res/deepdoc"),
+                    local_dir_use_symlinks=False,
+                ),
+            )
 
     def __call__(self, images, thr=0.2):
-        tbls = super().__call__(images, thr)
+        table_structure_recognizer_type = os.getenv("TABLE_STRUCTURE_RECOGNIZER_TYPE", "onnx").lower()
+        if table_structure_recognizer_type not in ["onnx", "ascend"]:
+            raise RuntimeError("Unsupported table structure recognizer type.")
+
+        if table_structure_recognizer_type == "onnx":
+            logging.debug("Using Onnx table structure recognizer", flush=True)
+            tbls = super().__call__(images, thr)
+        else:  # ascend
+            logging.debug("Using Ascend table structure recognizer", flush=True)
+            tbls = self._run_ascend_tsr(images, thr)
+
         res = []
         # align left&right for rows, align top&bottom for columns
         for tbl in tbls:
-            lts = [{"label": b["type"],
+            lts = [
+                {
+                    "label": b["type"],
                     "score": b["score"],
-                    "x0": b["bbox"][0], "x1": b["bbox"][2],
-                    "top": b["bbox"][1], "bottom": b["bbox"][-1]
-                    } for b in tbl]
+                    "x0": b["bbox"][0],
+                    "x1": b["bbox"][2],
+                    "top": b["bbox"][1],
+                    "bottom": b["bbox"][-1],
+                }
+                for b in tbl
+            ]
             if not lts:
                 continue
 
-            left = [b["x0"] for b in lts if b["label"].find(
-                "row") > 0 or b["label"].find("header") > 0]
-            right = [b["x1"] for b in lts if b["label"].find(
-                "row") > 0 or b["label"].find("header") > 0]
+            left = [b["x0"] for b in lts if b["label"].find("row") > 0 or b["label"].find("header") > 0]
+            right = [b["x1"] for b in lts if b["label"].find("row") > 0 or b["label"].find("header") > 0]
             if not left:
                 continue
             left = np.mean(left) if len(left) > 4 else np.min(left)
@@ -93,11 +112,8 @@ class TableStructureRecognizer(Recognizer):
 
     @staticmethod
     def is_caption(bx):
-        patt = [
-            r"[图表]+[ 0-9:：]{2,}"
-        ]
-        if any([re.match(p, bx["text"].strip()) for p in patt]) \
-                or bx.get("layout_type", "").find("caption") >= 0:
+        patt = [r"[图表]+[ 0-9:：]{2,}"]
+        if any([re.match(p, bx["text"].strip()) for p in patt]) or bx.get("layout_type", "").find("caption") >= 0:
             return True
         return False
 
@@ -115,7 +131,7 @@ class TableStructureRecognizer(Recognizer):
             (r"^[0-9A-Z/\._~-]+$", "Ca"),
             (r"^[A-Z]*[a-z' -]+$", "En"),
             (r"^[0-9.,+-]+[0-9A-Za-z/$￥%<>（）()' -]+$", "NE"),
-            (r"^.{1}$", "Sg")
+            (r"^.{1}$", "Sg"),
         ]
         for p, n in patt:
             if re.search(p, b["text"].strip()):
@@ -156,21 +172,19 @@ class TableStructureRecognizer(Recognizer):
         rowh = [b["R_bott"] - b["R_top"] for b in boxes if "R" in b]
         rowh = np.min(rowh) if rowh else 0
         boxes = Recognizer.sort_R_firstly(boxes, rowh / 2)
-        #for b in boxes:print(b)
+        # for b in boxes:print(b)
         boxes[0]["rn"] = 0
         rows = [[boxes[0]]]
         btm = boxes[0]["bottom"]
         for b in boxes[1:]:
             b["rn"] = len(rows) - 1
             lst_r = rows[-1]
-            if lst_r[-1].get("R", "") != b.get("R", "") \
-                    or (b["top"] >= btm - 3 and lst_r[-1].get("R", "-1") != b.get("R", "-2")
-                        ):  # new row
+            if lst_r[-1].get("R", "") != b.get("R", "") or (b["top"] >= btm - 3 and lst_r[-1].get("R", "-1") != b.get("R", "-2")):  # new row
                 btm = b["bottom"]
                 b["rn"] += 1
                 rows.append([b])
                 continue
-            btm = (btm + b["bottom"]) / 2.
+            btm = (btm + b["bottom"]) / 2.0
             rows[-1].append(b)
 
         colwm = [b["C_right"] - b["C_left"] for b in boxes if "C" in b]
@@ -186,14 +200,14 @@ class TableStructureRecognizer(Recognizer):
         for b in boxes[1:]:
             b["cn"] = len(cols) - 1
             lst_c = cols[-1]
-            if (int(b.get("C", "1")) - int(lst_c[-1].get("C", "1")) == 1 and b["page_number"] == lst_c[-1][
-                "page_number"]) \
-                    or (b["x0"] >= right and lst_c[-1].get("C", "-1") != b.get("C", "-2")):  # new col
+            if (int(b.get("C", "1")) - int(lst_c[-1].get("C", "1")) == 1 and b["page_number"] == lst_c[-1]["page_number"]) or (
+                b["x0"] >= right and lst_c[-1].get("C", "-1") != b.get("C", "-2")
+            ):  # new col
                 right = b["x1"]
                 b["cn"] += 1
                 cols.append([b])
                 continue
-            right = (right + b["x1"]) / 2.
+            right = (right + b["x1"]) / 2.0
             cols[-1].append(b)
 
         tbl = [[[] for _ in range(len(cols))] for _ in range(len(rows))]
@@ -214,10 +228,8 @@ class TableStructureRecognizer(Recognizer):
                 if e > 1:
                     j += 1
                     continue
-                f = (j > 0 and tbl[ii][j - 1] and tbl[ii]
-                     [j - 1][0].get("text")) or j == 0
-                ff = (j + 1 < len(tbl[ii]) and tbl[ii][j + 1] and tbl[ii]
-                      [j + 1][0].get("text")) or j + 1 >= len(tbl[ii])
+                f = (j > 0 and tbl[ii][j - 1] and tbl[ii][j - 1][0].get("text")) or j == 0
+                ff = (j + 1 < len(tbl[ii]) and tbl[ii][j + 1] and tbl[ii][j + 1][0].get("text")) or j + 1 >= len(tbl[ii])
                 if f and ff:
                     j += 1
                     continue
@@ -228,13 +240,11 @@ class TableStructureRecognizer(Recognizer):
                 if j > 0 and not f:
                     for i in range(len(tbl)):
                         if tbl[i][j - 1]:
-                            left = min(left, np.min(
-                                [bx["x0"] - a["x1"] for a in tbl[i][j - 1]]))
+                            left = min(left, np.min([bx["x0"] - a["x1"] for a in tbl[i][j - 1]]))
                 if j + 1 < len(tbl[0]) and not ff:
                     for i in range(len(tbl)):
                         if tbl[i][j + 1]:
-                            right = min(right, np.min(
-                                [a["x0"] - bx["x1"] for a in tbl[i][j + 1]]))
+                            right = min(right, np.min([a["x0"] - bx["x1"] for a in tbl[i][j + 1]]))
                 assert left < 100000 or right < 100000
                 if left < right:
                     for jj in range(j, len(tbl[0])):
@@ -260,8 +270,7 @@ class TableStructureRecognizer(Recognizer):
                     for i in range(len(tbl)):
                         tbl[i].pop(j)
                 cols.pop(j)
-        assert len(cols) == len(tbl[0]), "Column NO. miss matched: %d vs %d" % (
-            len(cols), len(tbl[0]))
+        assert len(cols) == len(tbl[0]), "Column NO. miss matched: %d vs %d" % (len(cols), len(tbl[0]))
 
         if len(cols) >= 4:
             # remove single in row
@@ -277,10 +286,8 @@ class TableStructureRecognizer(Recognizer):
                 if e > 1:
                     i += 1
                     continue
-                f = (i > 0 and tbl[i - 1][jj] and tbl[i - 1]
-                     [jj][0].get("text")) or i == 0
-                ff = (i + 1 < len(tbl) and tbl[i + 1][jj] and tbl[i + 1]
-                      [jj][0].get("text")) or i + 1 >= len(tbl)
+                f = (i > 0 and tbl[i - 1][jj] and tbl[i - 1][jj][0].get("text")) or i == 0
+                ff = (i + 1 < len(tbl) and tbl[i + 1][jj] and tbl[i + 1][jj][0].get("text")) or i + 1 >= len(tbl)
                 if f and ff:
                     i += 1
                     continue
@@ -292,13 +299,11 @@ class TableStructureRecognizer(Recognizer):
                 if i > 0 and not f:
                     for j in range(len(tbl[i - 1])):
                         if tbl[i - 1][j]:
-                            up = min(up, np.min(
-                                [bx["top"] - a["bottom"] for a in tbl[i - 1][j]]))
+                            up = min(up, np.min([bx["top"] - a["bottom"] for a in tbl[i - 1][j]]))
                 if i + 1 < len(tbl) and not ff:
                     for j in range(len(tbl[i + 1])):
                         if tbl[i + 1][j]:
-                            down = min(down, np.min(
-                                [a["top"] - bx["bottom"] for a in tbl[i + 1][j]]))
+                            down = min(down, np.min([a["top"] - bx["bottom"] for a in tbl[i + 1][j]]))
                 assert up < 100000 or down < 100000
                 if up < down:
                     for ii in range(i, len(tbl)):
@@ -333,22 +338,15 @@ class TableStructureRecognizer(Recognizer):
                 cnt += 1
                 if max_type == "Nu" and arr[0]["btype"] == "Nu":
                     continue
-                if any([a.get("H") for a in arr]) \
-                        or (max_type == "Nu" and arr[0]["btype"] != "Nu"):
+                if any([a.get("H") for a in arr]) or (max_type == "Nu" and arr[0]["btype"] != "Nu"):
                     h += 1
             if h / cnt > 0.5:
                 hdset.add(i)
 
         if html:
-            return TableStructureRecognizer.__html_table(cap, hdset,
-                                                         TableStructureRecognizer.__cal_spans(boxes, rows,
-                                                                                              cols, tbl, True)
-                                                         )
+            return TableStructureRecognizer.__html_table(cap, hdset, TableStructureRecognizer.__cal_spans(boxes, rows, cols, tbl, True))
 
-        return TableStructureRecognizer.__desc_table(cap, hdset,
-                                                     TableStructureRecognizer.__cal_spans(boxes, rows, cols, tbl,
-                                                                                          False),
-                                                     is_english)
+        return TableStructureRecognizer.__desc_table(cap, hdset, TableStructureRecognizer.__cal_spans(boxes, rows, cols, tbl, False), is_english)
 
     @staticmethod
     def __html_table(cap, hdset, tbl):
@@ -367,10 +365,8 @@ class TableStructureRecognizer(Recognizer):
                     continue
                 txt = ""
                 if arr:
-                    h = min(np.min([c["bottom"] - c["top"]
-                            for c in arr]) / 2, 10)
-                    txt = " ".join([c["text"]
-                                   for c in Recognizer.sort_Y_firstly(arr, h)])
+                    h = min(np.min([c["bottom"] - c["top"] for c in arr]) / 2, 10)
+                    txt = " ".join([c["text"] for c in Recognizer.sort_Y_firstly(arr, h)])
                 txts.append(txt)
                 sp = ""
                 if arr[0].get("colspan"):
@@ -436,15 +432,11 @@ class TableStructureRecognizer(Recognizer):
                     if headers[j][k].find(headers[j - 1][k]) >= 0:
                         continue
                     if len(headers[j][k]) > len(headers[j - 1][k]):
-                        headers[j][k] += (de if headers[j][k]
-                                          else "") + headers[j - 1][k]
+                        headers[j][k] += (de if headers[j][k] else "") + headers[j - 1][k]
                     else:
-                        headers[j][k] = headers[j - 1][k] \
-                            + (de if headers[j - 1][k] else "") \
-                            + headers[j][k]
+                        headers[j][k] = headers[j - 1][k] + (de if headers[j - 1][k] else "") + headers[j][k]
 
-        logging.debug(
-            f">>>>>>>>>>>>>>>>>{cap}：SIZE:{rowno}X{clmno} Header: {hdr_rowno}")
+        logging.debug(f">>>>>>>>>>>>>>>>>{cap}：SIZE:{rowno}X{clmno} Header: {hdr_rowno}")
         row_txt = []
         for i in range(rowno):
             if i in hdr_rowno:
@@ -503,14 +495,10 @@ class TableStructureRecognizer(Recognizer):
     @staticmethod
     def __cal_spans(boxes, rows, cols, tbl, html=True):
         # caculate span
-        clft = [np.mean([c.get("C_left", c["x0"]) for c in cln])
-                for cln in cols]
-        crgt = [np.mean([c.get("C_right", c["x1"]) for c in cln])
-                for cln in cols]
-        rtop = [np.mean([c.get("R_top", c["top"]) for c in row])
-                for row in rows]
-        rbtm = [np.mean([c.get("R_btm", c["bottom"])
-                         for c in row]) for row in rows]
+        clft = [np.mean([c.get("C_left", c["x0"]) for c in cln]) for cln in cols]
+        crgt = [np.mean([c.get("C_right", c["x1"]) for c in cln]) for cln in cols]
+        rtop = [np.mean([c.get("R_top", c["top"]) for c in row]) for row in rows]
+        rbtm = [np.mean([c.get("R_btm", c["bottom"]) for c in row]) for row in rows]
         for b in boxes:
             if "SP" not in b:
                 continue
@@ -585,3 +573,40 @@ class TableStructureRecognizer(Recognizer):
                 tbl[rowspan[0]][colspan[0]] = arr
 
         return tbl
+
+    def _run_ascend_tsr(self, image_list, thr=0.2, batch_size=16):
+        import math
+
+        from ais_bench.infer.interface import InferSession
+
+        model_dir = os.path.join(get_project_base_directory(), "rag/res/deepdoc")
+        model_file_path = os.path.join(model_dir, "tsr.om")
+
+        if not os.path.exists(model_file_path):
+            raise ValueError(f"Model file not found: {model_file_path}")
+
+        device_id = int(os.getenv("ASCEND_LAYOUT_RECOGNIZER_DEVICE_ID", 0))
+        session = InferSession(device_id=device_id, model_path=model_file_path)
+
+        images = [np.array(im) if not isinstance(im, np.ndarray) else im for im in image_list]
+        results = []
+
+        conf_thr = max(thr, 0.08)
+
+        batch_loop_cnt = math.ceil(float(len(images)) / batch_size)
+        for bi in range(batch_loop_cnt):
+            s = bi * batch_size
+            e = min((bi + 1) * batch_size, len(images))
+            batch_images = images[s:e]
+
+            inputs_list = self.preprocess(batch_images)
+            for ins in inputs_list:
+                feeds = []
+                if "image" in ins:
+                    feeds.append(ins["image"])
+                else:
+                    feeds.append(ins[self.input_names[0]])
+                output_list = session.infer(feeds=feeds, mode="static")
+                bb = self.postprocess(output_list, ins, conf_thr)
+                results.append(bb)
+        return results

docker/service_conf.yaml.template

@@ -1,6 +1,9 @@
 ragflow:
   host: ${RAGFLOW_HOST:-0.0.0.0}
   http_port: 9380
+admin:
+  host: ${RAGFLOW_HOST:-0.0.0.0}
+  http_port: 9381
 mysql:
   name: '${MYSQL_DBNAME:-rag_flow}'
   user: '${MYSQL_USER:-root}'
@@ -29,7 +32,6 @@ redis:
   db: 1
   password: '${REDIS_PASSWORD:-infini_rag_flow}'
   host: '${REDIS_HOST:-redis}:6379'
-
 # postgres:
 #   name: '${POSTGRES_DBNAME:-rag_flow}'
 #   user: '${POSTGRES_USER:-rag_flow}'
@@ -65,15 +67,26 @@ redis:
 #   secret: 'secret'
 #   tenant_id: 'tenant_id'
 #   container_name: 'container_name'
+# The OSS object storage uses the MySQL configuration above by default. If you need to switch to another object storage service, please uncomment and configure the following parameters.
+# opendal:
+#   scheme: 'mysql'  # Storage type, such as s3, oss, azure, etc.
+#   config:
+#     oss_table: 'opendal_storage'
 # user_default_llm:
-#   factory: 'Tongyi-Qianwen'
-#   api_key: 'sk-xxxxxxxxxxxxx'
-#   base_url: ''
+#   factory: 'BAAI'
+#   api_key: 'backup'
+#   base_url: 'backup_base_url'
 #   default_models:
-#     chat_model: 'qwen-plus'
-#     embedding_model: 'BAAI/bge-large-zh-v1.5@BAAI'
-#     rerank_model: ''
-#     asr_model: ''
+#     chat_model:
+#       name: 'qwen2.5-7b-instruct'
+#       factory: 'xxxx'
+#       api_key: 'xxxx'
+#       base_url: 'https://api.xx.com'
+#     embedding_model:
+#       name: 'bge-m3'
+#     rerank_model: 'bge-reranker-v2'
+#     asr_model:
+#       model: 'whisper-large-v3' # alias of name
 #     image2text_model: ''
 # oauth:
 #   oauth2:
@@ -109,3 +122,14 @@ redis:
 #   switch: false
 #   component: false
 #   dataset: false
+# smtp:
+#   mail_server: ""
+#   mail_port: 465
+#   mail_use_ssl: true
+#   mail_use_tls: false
+#   mail_username: ""
+#   mail_password: ""
+#   mail_default_sender:
+#     - "RAGFlow" # display name
+#     - "" # sender email address
+#   mail_frontend_url: "https://your-frontend.example.com"

docs/develop/mcp/_category_.json

@@ -3,6 +3,6 @@
   "position": 40,
   "link": {
     "type": "generated-index",
-    "description": "Guides and references on accessing RAGFlow's knowledge bases via MCP."
+    "description": "Guides and references on accessing RAGFlow's datasets via MCP."
   }
 }

docs/develop/mcp/launch_mcp_server.md

@@ -14,9 +14,9 @@ A RAGFlow Model Context Protocol (MCP) server is designed as an independent comp
 An MCP server can start up in either self-host mode (default) or host mode: 
 
 - **Self-host mode**:  
-  When launching an MCP server in self-host mode, you must provide an API key to authenticate the MCP server with the RAGFlow server. In this mode, the MCP server can access *only* the datasets (knowledge bases) of a specified tenant on the RAGFlow server.
+  When launching an MCP server in self-host mode, you must provide an API key to authenticate the MCP server with the RAGFlow server. In this mode, the MCP server can access *only* the datasets of a specified tenant on the RAGFlow server.
 - **Host mode**:  
-  In host mode, each MCP client can access their own knowledge bases on the RAGFlow server. However, each client request must include a valid API key to authenticate the client with the RAGFlow server.
+  In host mode, each MCP client can access their own datasets on the RAGFlow server. However, each client request must include a valid API key to authenticate the client with the RAGFlow server.
 
 Once a connection is established, an MCP server communicates with its client in MCP HTTP+SSE (Server-Sent Events) mode, unidirectionally pushing responses from the RAGFlow server to its client in real time.
 

docs/faq.mdx

@@ -498,7 +498,7 @@ To switch your document engine from Elasticsearch to [Infinity](https://github.c
 
 ### Where are my uploaded files stored in RAGFlow's image?
 
-All uploaded files are stored in Minio, RAGFlow's object storage solution. For instance, if you upload your file directly to a knowledge base, it is located at `<knowledgebase_id>/filename`.
+All uploaded files are stored in Minio, RAGFlow's object storage solution. For instance, if you upload your file directly to a dataset, it is located at `<knowledgebase_id>/filename`.
 
 ---
 
@@ -507,3 +507,16 @@ All uploaded files are stored in Minio, RAGFlow's object storage solution. For i
 You can control the batch size for document parsing and embedding by setting the environment variables `DOC_BULK_SIZE` and `EMBEDDING_BATCH_SIZE`. Increasing these values may improve throughput for large-scale data processing, but will also increase memory usage. Adjust them according to your hardware resources.
 
 ---
+
+### How to accelerate the question-answering speed of my chat assistant?
+
+See [here](./guides/chat/best_practices/accelerate_question_answering.mdx).
+
+---
+
+### How to accelerate the question-answering speed of my Agent?
+
+See [here](./guides/agent/best_practices/accelerate_agent_question_answering.md).
+
+---
+

docs/guides/agent/agent_component_reference/agent.mdx

@@ -26,6 +26,84 @@ An **Agent** component is essential when you need the LLM to assist with summari
 
 2. If your Agent involves dataset retrieval, ensure you [have properly configured your target knowledge base(s)](../../dataset/configure_knowledge_base.md).
 
+## Quickstart
+
+### 1. Click on an **Agent** component to show its configuration panel  
+
+The corresponding configuration panel appears to the right of the canvas. Use this panel to define and fine-tune the **Agent** component's behavior.
+
+### 2. Select your model
+
+Click **Model**, and select a chat model from the dropdown menu. 
+
+:::tip NOTE
+If no model appears, check if your have added a chat model on the **Model providers** page.
+:::
+
+### 3. Update system prompt (Optional)
+
+The system prompt typically defines your model's role. You can either keep the system prompt as is or customize it to override the default.
+
+
+### 4. Update user prompt
+
+The user prompt typically defines your model's task. You will find the `sys.query` variable auto-populated. Type `/` or click **(x)** to view or add variables.
+
+In this quickstart, we assume your **Agent** component is used standalone (without tools or sub-Agents below), then you may also need to specify retrieved chunks using the `formalized_content` variable:
+
+![](https://raw.githubusercontent.com/infiniflow/ragflow-docs/main/images/standalone_user_prompt_variable.jpg)
+
+### 5. Skip Tools and Agent
+
+The **+ Add tools** and **+ Add agent** sections are used *only* when you need to configure your **Agent** component as a planner (with tools or sub-Agents beneath). In this quickstart, we assume your **Agent** component is used standalone (without tools or sub-Agents beneath). 
+
+### 6. Choose the next component
+
+When necessary, click the **+** button on the **Agent** component to choose the next component in the worflow from the dropdown list.
+
+## Connect to an MCP server as a client
+
+:::danger IMPORTANT
+In this section, we assume your **Agent** will be configured as a planner, with a Tavily tool beneath it.
+:::
+
+### 1. Navigate to the MCP configuration page
+
+![](https://raw.githubusercontent.com/infiniflow/ragflow-docs/main/images/mcp_page.jpg)
+
+### 2. Configure your Tavily MCP server 
+
+Update your MCP server's name, URL (including the API key), server type, and other necessary settings. When configured correctly, the available tools will be displayed.
+
+![](https://raw.githubusercontent.com/infiniflow/ragflow-docs/main/images/edit_mcp_server.jpg)
+
+### 3. Navigate to your Agent's editing page
+
+### 4. Connect to your MCP server
+
+1. Click **+ Add tools**:
+
+![](https://raw.githubusercontent.com/infiniflow/ragflow-docs/main/images/add_tools.jpg)
+
+2. Click **MCP** to show the available MCP servers.
+
+3. Select your MCP server:
+
+   *The target MCP server appears below your Agent component, and your Agent will autonomously decide when to invoke the available tools it offers.*
+
+![](https://raw.githubusercontent.com/infiniflow/ragflow-docs/main/images/choose_tavily_mcp_server.jpg)
+
+### 5. Update system prompt to specify trigger conditions (Optional)
+
+To ensure reliable tool calls, you may specify within the system prompt which tasks should trigger each tool call.
+
+### 6. View the availabe tools of your MCP server
+
+On the canvas, click the newly-populated Tavily server to view and select its available tools:
+
+![](https://raw.githubusercontent.com/infiniflow/ragflow-docs/main/images/tavily_mcp_server.jpg)
+
+
 ## Configurations
 
 ### Model
@@ -69,7 +147,7 @@ An **Agent** component relies on keys (variables) to specify its data inputs. It
 
 #### Advanced usage
 
-From v0.20.5 onwards, four framework-level prompt blocks are available in the **System prompt** field. Type `/` or click **(x)** to view them; they appear under the **Framework** entry in the dropdown menu.
+From v0.20.5 onwards, four framework-level prompt blocks are available in the **System prompt** field, enabling you to customize and *override* prompts at the framework level. Type `/` or click **(x)** to view them; they appear under the **Framework** entry in the dropdown menu.
 
 - `task_analysis` prompt block
   - This block is responsible for analyzing tasks — either a user task or a task assigned by the lead Agent when the **Agent** component is acting as a Sub-Agent.
@@ -100,6 +178,12 @@ From v0.20.5 onwards, four framework-level prompt blocks are available in the **
 - `citation_guidelines` prompt block
   - Reference design: [citation_prompt.md](https://github.com/infiniflow/ragflow/blob/main/rag/prompts/citation_prompt.md)
 
+*The screenshots below show the framework prompt blocks available to an **Agent** component, both as a standalone and as a planner (with a Tavily tool below):*
+
+![standalone](https://raw.githubusercontent.com/infiniflow/ragflow-docs/main/images/standalone_agent_framework_block.jpg)
+
+![planner](https://raw.githubusercontent.com/infiniflow/ragflow-docs/main/images/planner_agent_framework_blocks.jpg)
+
 ### User prompt
 
 The user-defined prompt. Defaults to `sys.query`, the user query. As a general rule, when using the **Agent** component as a standalone module (not as a planner), you usually need to specify the corresponding **Retrieval** component’s output variable (`formalized_content`) here as part of the input to the LLM.
@@ -129,7 +213,7 @@ Defines the maximum number of attempts the agent will make to retry a failed tas
 
 The waiting period in seconds that the agent observes before retrying a failed task, helping to prevent immediate repeated attempts and allowing system conditions to improve. Defaults to 1 second.
 
-### Max rounds
+### Max reflection rounds
 
 Defines the maximum number reflection rounds of the selected chat model. Defaults to 1 round.
 
@@ -145,18 +229,4 @@ The global variable name for the output of the **Agent** component, which can be
 
 ### Why does it take so long for my Agent to respond?
 
-An Agent’s response time generally depends on two key factors: the LLM’s capabilities and the prompt, the latter reflecting task complexity. When using an Agent, you should always balance task demands with the LLM’s ability. See [How to balance task complexity with an Agent's performance and speed?](#how-to-balance-task-complexity-with-an-agents-performance-and-speed) for details.
-
-## Best practices
-
-### How to balance task complexity with an Agent’s performance and speed?
-
-- For simple tasks, such as retrieval, rewriting, formatting, or structured data extraction, use concise prompts, remove planning or reasoning instructions, enforce output length limits, and select smaller or Turbo-class models. This significantly reduces latency and cost with minimal impact on quality.
-
-- For complex tasks, like multi-step reasoning, cross-document synthesis, or tool-based workflows, maintain or enhance prompts that include planning, reflection, and verification steps.
-
-- In multi-Agent orchestration systems, delegate simple subtasks to sub-Agents using smaller, faster models, and reserve more powerful models for the lead Agent to handle complexity and uncertainty.
-
-:::tip KEY INSIGHT
-Focus on minimizing output tokens — through summarization, bullet points, or explicit length limits — as this has far greater impact on reducing latency than optimizing input size.
-:::
+See [here](../best_practices/accelerate_agent_question_answering.md) for details.

docs/guides/agent/agent_component_reference/begin.mdx

@@ -67,14 +67,14 @@ You can tune document parsing and embedding efficiency by setting the environmen
 
 ## Frequently asked questions
 
-### Is the uploaded file in a knowledge base?
+### Is the uploaded file in a dataset?
 
-No. Files uploaded to an agent as input are not stored in a knowledge base and hence will not be processed using RAGFlow's built-in OCR, DLR or TSR models, or chunked using RAGFlow's built-in chunking methods.
+No. Files uploaded to an agent as input are not stored in a dataset and hence will not be processed using RAGFlow's built-in OCR, DLR or TSR models, or chunked using RAGFlow's built-in chunking methods.
 
 ### File size limit for an uploaded file
 
 There is no _specific_ file size limit for a file uploaded to an agent. However, note that model providers typically have a default or explicit maximum token setting, which can range from 8196 to 128k: The plain text part of the uploaded file will be passed in as the key value, but if the file's token count exceeds this limit, the string will be truncated and incomplete.
 
 :::tip NOTE
-The variables `MAX_CONTENT_LENGTH` in `/docker/.env` and `client_max_body_size` in `/docker/nginx/nginx.conf` set the file size limit for each upload to a knowledge base or **File Management**. These settings DO NOT apply in this scenario.
+The variables `MAX_CONTENT_LENGTH` in `/docker/.env` and `client_max_body_size` in `/docker/nginx/nginx.conf` set the file size limit for each upload to a dataset or **File Management**. These settings DO NOT apply in this scenario.
 :::

docs/guides/agent/agent_component_reference/code.mdx

@@ -49,6 +49,10 @@ You can specify multiple input sources for the **Code** component. Click **+ Add
 
 This field allows you to enter and edit your source code.
 
+:::danger IMPORTANT
+If your code implementation includes defined variables, whether input or output variables, ensure they are also specified in the corresponding **Input** or **Output** sections.
+:::
+
 #### A Python code example
 
 ```Python 
@@ -77,6 +81,15 @@ This field allows you to enter and edit your source code.
 
 You define the output variable(s) of the **Code** component here.
 
+:::danger IMPORTANT
+If you define output variables here, ensure they are also defined in your code implementation; otherwise, their values will be `null`. The following are two examples:
+
+
+![](https://raw.githubusercontent.com/infiniflow/ragflow-docs/main/images/set_object_output.jpg)
+
+![](https://raw.githubusercontent.com/infiniflow/ragflow-docs/main/images/set_nested_object_output.png)
+:::
+
 ### Output
 
 The defined output variable(s) will be auto-populated here.

docs/guides/agent/agent_component_reference/execute_sql.md

@@ -0,0 +1,79 @@
+---
+sidebar_position: 25
+slug: /execute_sql
+---
+
+# Execute SQL tool
+
+A tool that execute SQL queries on a specified relational database.
+
+---
+
+The **Execute SQL** tool enables you to connect to a relational database and run SQL queries, whether entered directly or generated by the system’s Text2SQL capability via an **Agent** component. 
+
+## Prerequisites
+
+- A database instance properly configured and running.
+- The database must be one of the following types:
+  - MySQL
+  - PostgreSQL
+  - MariaDB
+  - Microsoft SQL Server
+
+## Examples
+
+You can pair an **Agent** component with the **Execute SQL** tool, with the **Agent** generating SQL statements and the **Execute SQL** tool handling database connection and query execution. An example of this setup can be found in the **SQL Assistant** Agent template shown below:
+
+![](https://raw.githubusercontent.com/infiniflow/ragflow-docs/main/images/exeSQL.jpg)
+
+## Configurations
+
+### SQL statement
+
+This text input field allows you to write static SQL queries, such as `SELECT * FROM my_table`, and dynamic SQL queries using variables.
+
+:::tip NOTE
+Click **(x)** or type `/` to insert variables.
+:::
+
+For dynamic SQL queries, you can include variables in your SQL queries, such as `SELECT * FROM /sys.query`; if an **Agent** component is paired with the **Execute SQL** tool to generate SQL tasks (see the [Examples](#examples) section), you can directly insert that **Agent**'s output, `content`, into this field.
+
+### Database type
+
+The supported database type. Currently the following database types are available:
+
+- MySQL
+- PostreSQL
+- MariaDB
+- Microsoft SQL Server (Myssql)
+
+### Database
+
+Appears only when you select **Split** as method.
+
+### Username
+
+The username with access privileges to the database.
+
+### Host
+
+The IP address of the database server.
+
+### Port
+
+The port number on which the database server is listening.
+
+### Password
+
+The password for the database user.
+
+### Max records
+
+The maximum number of records returned by the SQL query to control response size and improve efficiency. Defaults to `1024`.
+
+### Output
+
+The **Execute SQL** tool provides two output variables:
+
+- `formalized_content`: A string. If you reference this variable in a **Message** component, the returned records are displayed as a table.
+- `json`: An object array. If you reference this variable in a **Message** component, the returned records will be presented as key-value pairs.

docs/guides/agent/agent_component_reference/retrieval.mdx

@@ -9,7 +9,7 @@ A component that retrieves information from specified datasets.
 
 ## Scenarios
 
-A **Retrieval** component is essential in most RAG scenarios, where information is extracted from designated knowledge bases before being sent to the LLM for content generation. A **Retrieval** component can operate either as a standalone workflow module or as a tool for an **Agent** component. In the latter role, the **Agent** component has autonomous control over when to invoke it for query and retrieval.
+A **Retrieval** component is essential in most RAG scenarios, where information is extracted from designated datasets before being sent to the LLM for content generation. A **Retrieval** component can operate either as a standalone workflow module or as a tool for an **Agent** component. In the latter role, the **Agent** component has autonomous control over when to invoke it for query and retrieval.
 
 The following screenshot shows a reference design using the **Retrieval** component, where the component serves as a tool for an **Agent** component. You can find it from the **Report Agent Using Knowledge Base** Agent template.
 
@@ -17,7 +17,7 @@ The following screenshot shows a reference design using the **Retrieval** compon
 
 ## Prerequisites
 
-Ensure you [have properly configured your target knowledge base(s)](../../dataset/configure_knowledge_base.md).
+Ensure you [have properly configured your target dataset(s)](../../dataset/configure_knowledge_base.md).
 
 ## Quickstart
 
@@ -36,9 +36,9 @@ The **Retrieval** component depends on query variables to specify its queries.
 
 By default, you can use `sys.query`, which is the user query and the default output of the **Begin** component. All global variables defined before the **Retrieval** component can also be used as query statements. Use the `(x)` button or type `/` to show all the available query variables.
 
-### 3. Select knowledge base(s) to query
+### 3. Select dataset(s) to query
 
-You can specify one or multiple knowledge bases to retrieve data from. If selecting mutiple, ensure they use the same embedding model.
+You can specify one or multiple datasets to retrieve data from. If selecting mutiple, ensure they use the same embedding model.
 
 ### 4. Expand **Advanced Settings** to configure the retrieval method
 
@@ -52,7 +52,7 @@ Using a rerank model will *significantly* increase the system's response time. I
 
 ### 5. Enable cross-language search
 
-If your user query is different from the languages of the knowledge bases, you can select the target languages in the **Cross-language search** dropdown menu. The model will then translates queries to ensure accurate matching of semantic meaning across languages.
+If your user query is different from the languages of the datasets, you can select the target languages in the **Cross-language search** dropdown menu. The model will then translates queries to ensure accurate matching of semantic meaning across languages.
 
 
 ### 6. Test retrieval results
@@ -76,10 +76,10 @@ The **Retrieval** component relies on query variables to specify its queries. Al
 
 ### Knowledge bases 
 
-Select the knowledge base(s) to retrieve data from.
+Select the dataset(s) to retrieve data from.
 
-- If no knowledge base is selected, meaning conversations with the agent will not be based on any knowledge base, ensure that the **Empty response** field is left blank to avoid an error.
-- If you select multiple knowledge bases, you must ensure that the knowledge bases (datasets) you select use the same embedding model; otherwise, an error message would occur.
+- If no dataset is selected, meaning conversations with the agent will not be based on any dataset, ensure that the **Empty response** field is left blank to avoid an error.
+- If you select multiple datasets, you must ensure that the datasets you select use the same embedding model; otherwise, an error message would occur.
 
 ### Similarity threshold
 
@@ -110,11 +110,11 @@ Using a rerank model will *significantly* increase the system's response time.
 
 ### Empty response
 
-- Set this as a response if no results are retrieved from the knowledge base(s) for your query, or 
+- Set this as a response if no results are retrieved from the dataset(s) for your query, or 
 - Leave this field blank to allow the chat model to improvise when nothing is found.
 
 :::caution WARNING
-If you do not specify a knowledge base, you must leave this field blank; otherwise, an error would occur.
+If you do not specify a dataset, you must leave this field blank; otherwise, an error would occur.
 :::
 
 ### Cross-language search
@@ -124,10 +124,10 @@ Select one or more languages for cross‑language search. If no language is sele
 ### Use knowledge graph
 
 :::caution IMPORTANT
-Before enabling this feature, ensure you have properly [constructed a knowledge graph from each target knowledge base](../../dataset/construct_knowledge_graph.md).
+Before enabling this feature, ensure you have properly [constructed a knowledge graph from each target dataset](../../dataset/construct_knowledge_graph.md).
 :::
 
-Whether to use knowledge graph(s) in the specified knowledge base(s) during retrieval for multi-hop question answering. When enabled, this would involve iterative searches across entity, relationship, and community report chunks, greatly increasing retrieval time.
+Whether to use knowledge graph(s) in the specified dataset(s) during retrieval for multi-hop question answering. When enabled, this would involve iterative searches across entity, relationship, and community report chunks, greatly increasing retrieval time.
 
 ### Output
 

docs/guides/agent/agent_introduction.md

@@ -27,7 +27,7 @@ Agents and RAG are complementary techniques, each enhancing the other’s capabi
 Before proceeding, ensure that:  
 
 1. You have properly set the LLM to use. See the guides on [Configure your API key](../models/llm_api_key_setup.md) or [Deploy a local LLM](../models/deploy_local_llm.mdx) for more information.
-2. You have a knowledge base configured and the corresponding files properly parsed. See the guide on [Configure a knowledge base](../dataset/configure_knowledge_base.md) for more information.
+2. You have a dataset configured and the corresponding files properly parsed. See the guide on [Configure a dataset](../dataset/configure_knowledge_base.md) for more information.
 
 :::
 

docs/guides/agent/best_practices/_category_.json

@@ -0,0 +1,8 @@
+{
+  "label": "Best practices",
+  "position": 30,
+  "link": {
+    "type": "generated-index",
+    "description": "Best practices on Agent configuration."
+  }
+}

docs/guides/agent/best_practices/accelerate_agent_question_answering.md

@@ -0,0 +1,58 @@
+---
+sidebar_position: 1
+slug: /accelerate_agent_question_answering
+---
+
+# Accelerate answering
+
+A checklist to speed up question answering.
+
+---
+
+Please note that some of your settings may consume a significant amount of time. If you often find that your question answering is time-consuming, here is a checklist to consider:
+
+## Balance task complexity with an Agent’s performance and speed?
+
+An Agent’s response time generally depends on many factors, e.g., the LLM’s capabilities and the prompt, the latter reflecting task complexity. When using an Agent, you should always balance task demands with the LLM’s ability. 
+
+- For simple tasks, such as retrieval, rewriting, formatting, or structured data extraction, use concise prompts, remove planning or reasoning instructions, enforce output length limits, and select smaller or Turbo-class models. This significantly reduces latency and cost with minimal impact on quality.
+
+- For complex tasks, like multi-step reasoning, cross-document synthesis, or tool-based workflows, maintain or enhance prompts that include planning, reflection, and verification steps.
+
+- In multi-Agent orchestration systems, delegate simple subtasks to sub-Agents using smaller, faster models, and reserve more powerful models for the lead Agent to handle complexity and uncertainty.
+
+:::tip KEY INSIGHT
+Focus on minimizing output tokens — through summarization, bullet points, or explicit length limits — as this has far greater impact on reducing latency than optimizing input size.
+:::
+
+## Disable Reasoning
+
+Disabling the **Reasoning** toggle will reduce the LLM's thinking time. For a model like Qwen3, you also need to add `/no_think` to the system prompt to disable reasoning.
+
+## Disable Rerank model
+
+- Leaving the **Rerank model** field empty (in the corresponding **Retrieval** component) will significantly decrease retrieval time.
+- When using a rerank model, ensure you have a GPU for acceleration; otherwise, the reranking process will be *prohibitively* slow.
+
+:::tip NOTE 
+Please note that rerank models are essential in certain scenarios. There is always a trade-off between speed and performance; you must weigh the pros against cons for your specific case.
+:::
+
+## Check the time taken for each task
+
+Click the light bulb icon above the *current* dialogue and scroll down the popup window to view the time taken for each task:  
+
+
+
+| Item name         | Description                                                                                   |
+| ----------------- | --------------------------------------------------------------------------------------------- |
+| Total             | Total time spent on this conversation round, including chunk retrieval and answer generation. |
+| Check LLM         | Time to validate the specified LLM.                                                           |
+| Create retriever  | Time to create a chunk retriever.                                                             |
+| Bind embedding    | Time to initialize an embedding model instance.                                               |
+| Bind LLM          | Time to initialize an LLM instance.                                                           |
+| Tune question     | Time to optimize the user query using the context of the mult-turn conversation.              |
+| Bind reranker     | Time to initialize an reranker model instance for chunk retrieval.                            |
+| Generate keywords | Time to extract keywords from the user query.                                                 |
+| Retrieval         | Time to retrieve the chunks.                                                                  |
+| Generate answer   | Time to generate the answer.                                                                  |

docs/guides/ai_search.md

@@ -22,7 +22,7 @@ When debugging your chat assistant, you can use AI search as a reference to veri
 ## Prerequisites
 
 - Ensure that you have configured the system's default models on the **Model providers** page.
-- Ensure that the intended knowledge bases are properly configured and the intended documents have finished file parsing.
+- Ensure that the intended datasets are properly configured and the intended documents have finished file parsing.
 
 ## Frequently asked questions
 

docs/guides/chat/best_practices/accelerate_question_answering.mdx

@@ -6,21 +6,22 @@ slug: /accelerate_question_answering
 # Accelerate answering
 import APITable from '@site/src/components/APITable';
 
-A checklist to speed up question answering.
+A checklist to speed up question answering for your chat assistant.
 
 ---
 
 Please note that some of your settings may consume a significant amount of time. If you often find that your question answering is time-consuming, here is a checklist to consider:
 
-- In the **Prompt engine** tab of your **Chat Configuration** dialogue, disabling **Multi-turn optimization** will reduce the time required to get an answer from the LLM.
-- In the **Prompt engine** tab of your **Chat Configuration** dialogue, leaving the **Rerank model** field empty will significantly decrease retrieval time.
+- Disabling **Multi-turn optimization** will reduce the time required to get an answer from the LLM.
+- Leaving the **Rerank model** field empty will significantly decrease retrieval time.
+- Disabling the **Reasoning** toggle will reduce the LLM's thinking time. For a model like Qwen3, you also need to add `/no_think` to the system prompt to disable reasoning.
 - When using a rerank model, ensure you have a GPU for acceleration; otherwise, the reranking process will be *prohibitively* slow.
 
 :::tip NOTE 
 Please note that rerank models are essential in certain scenarios. There is always a trade-off between speed and performance; you must weigh the pros against cons for your specific case.
 :::
 
-- In the **Assistant settings** tab of your **Chat Configuration** dialogue, disabling **Keyword analysis** will reduce the time to receive an answer from the LLM.
+- Disabling **Keyword analysis** will reduce the time to receive an answer from the LLM.
 - When chatting with your chat assistant, click the light bulb icon above the *current* dialogue and scroll down the popup window to view the time taken for each task:  
    ![enlighten](https://github.com/user-attachments/assets/fedfa2ee-21a7-451b-be66-20125619923c)  
 

docs/guides/chat/set_chat_variables.md

@@ -25,13 +25,13 @@ In the **Variable** section, you add, remove, or update variables.
 
 ### `{knowledge}` - a reserved variable
 
-`{knowledge}` is the system's reserved variable, representing the chunks retrieved from the knowledge base(s) specified by **Knowledge bases** under the **Assistant settings** tab. If your chat assistant is associated with certain knowledge bases, you can keep it as is.
+`{knowledge}` is the system's reserved variable, representing the chunks retrieved from the dataset(s) specified by **Knowledge bases** under the **Assistant settings** tab. If your chat assistant is associated with certain datasets, you can keep it as is.
 
 :::info NOTE
 It currently makes no difference whether  `{knowledge}` is set as optional or mandatory, but please note this design will be updated in due course.
 :::
 
-From v0.17.0 onward, you can start an AI chat without specifying knowledge bases. In this case, we recommend removing the `{knowledge}` variable to prevent unnecessary reference and keeping the **Empty response** field empty to avoid errors.
+From v0.17.0 onward, you can start an AI chat without specifying datasets. In this case, we recommend removing the `{knowledge}` variable to prevent unnecessary reference and keeping the **Empty response** field empty to avoid errors.
 
 ### Custom variables
 
@@ -45,15 +45,15 @@ Besides `{knowledge}`, you can also define your own variables to pair with the s
 After you add or remove variables in the **Variable** section, ensure your changes are reflected in the system prompt to avoid inconsistencies or errors. Here's an example:
 
 ```
-You are an intelligent assistant. Please answer the question by summarizing chunks from the specified knowledge base(s)...
+You are an intelligent assistant. Please answer the question by summarizing chunks from the specified dataset(s)...
 
 Your answers should follow a professional and {style} style.
 
 ...
 
-Here is the knowledge base:
+Here is the dataset:
 {knowledge}
-The above is the knowledge base.
+The above is the dataset.
 ```
 
 :::tip NOTE

docs/guides/chat/start_chat.md

@@ -9,7 +9,7 @@ Initiate an AI-powered chat with a configured chat assistant.
 
 ---
 
-Knowledge base, hallucination-free chat, and file management are the three pillars of RAGFlow. Chats in RAGFlow are based on a particular knowledge base or multiple knowledge bases. Once you have created your knowledge base, finished file parsing, and [run a retrieval test](../dataset/run_retrieval_test.md), you can go ahead and start an AI conversation.
+Knowledge base, hallucination-free chat, and file management are the three pillars of RAGFlow. Chats in RAGFlow are based on a particular dataset or multiple datasets. Once you have created your dataset, finished file parsing, and [run a retrieval test](../dataset/run_retrieval_test.md), you can go ahead and start an AI conversation.
 
 ## Start an AI chat
 
@@ -21,12 +21,12 @@ You start an AI conversation by creating an assistant.
 
 2. Update **Assistant settings**:
 
-   - **Assistant name** is the name of your chat assistant. Each assistant corresponds to a dialogue with a unique combination of knowledge bases, prompts, hybrid search configurations, and large model settings.
+   - **Assistant name** is the name of your chat assistant. Each assistant corresponds to a dialogue with a unique combination of datasets, prompts, hybrid search configurations, and large model settings.
    - **Empty response**:
-     - If you wish to *confine* RAGFlow's answers to your knowledge bases, leave a response here. Then, when it doesn't retrieve an answer, it *uniformly* responds with what you set here.
-     - If you wish RAGFlow to *improvise* when it doesn't retrieve an answer from your knowledge bases, leave it blank, which may give rise to hallucinations.
+     - If you wish to *confine* RAGFlow's answers to your datasets, leave a response here. Then, when it doesn't retrieve an answer, it *uniformly* responds with what you set here.
+     - If you wish RAGFlow to *improvise* when it doesn't retrieve an answer from your datasets, leave it blank, which may give rise to hallucinations.
    - **Show quote**: This is a key feature of RAGFlow and enabled by default. RAGFlow does not work like a black box. Instead, it clearly shows the sources of information that its responses are based on.
-   - Select the corresponding knowledge bases. You can select one or multiple knowledge bases, but ensure that they use the same embedding model, otherwise an error would occur.
+   - Select the corresponding datasets. You can select one or multiple datasets, but ensure that they use the same embedding model, otherwise an error would occur.
 
 3. Update **Prompt engine**:
 
@@ -37,14 +37,14 @@ You start an AI conversation by creating an assistant.
      - If **Rerank model** is selected, the hybrid score system uses keyword similarity and reranker score, and the default weight assigned to the reranker score is 1-0.7=0.3.
    - **Top N** determines the *maximum* number of chunks to feed to the LLM. In other words, even if more chunks are retrieved, only the top N chunks are provided as input.
    - **Multi-turn optimization** enhances user queries using existing context in a multi-round conversation. It is enabled by default. When enabled, it will consume additional LLM tokens and significantly increase the time to generate answers.
-   - **Use knowledge graph** indicates whether to use knowledge graph(s) in the specified knowledge base(s) during retrieval for multi-hop question answering. When enabled, this would involve iterative searches across entity, relationship, and community report chunks, greatly increasing retrieval time.
+   - **Use knowledge graph** indicates whether to use knowledge graph(s) in the specified dataset(s) during retrieval for multi-hop question answering. When enabled, this would involve iterative searches across entity, relationship, and community report chunks, greatly increasing retrieval time.
    - **Reasoning** indicates whether to generate answers through reasoning processes like Deepseek-R1/OpenAI o1. Once enabled, the chat model autonomously integrates Deep Research during question answering when encountering an unknown topic. This involves the chat model dynamically searching external knowledge and generating final answers through reasoning.
    - **Rerank model** sets the reranker model to use. It is left empty by default.
      - If **Rerank model** is left empty, the hybrid score system uses keyword similarity and vector similarity, and the default weight assigned to the vector similarity component is 1-0.7=0.3.
      - If **Rerank model** is selected, the hybrid score system uses keyword similarity and reranker score, and the default weight assigned to the reranker score is 1-0.7=0.3.
    - [Cross-language search](../../references/glossary.mdx#cross-language-search): Optional  
      Select one or more target languages from the dropdown menu. The system’s default chat model will then translate your query into the selected target language(s). This translation ensures accurate semantic matching across languages, allowing you to retrieve relevant results regardless of language differences.  
-     - When selecting target languages, please ensure that these languages are present in the knowledge base to guarantee an effective search.
+     - When selecting target languages, please ensure that these languages are present in the dataset to guarantee an effective search.
      - If no target language is selected, the system will search only in the language of your query, which may cause relevant information in other languages to be missed.
    - **Variable** refers to the variables (keys) to be used in the system prompt. `{knowledge}` is a reserved variable. Click **Add** to add more variables for the system prompt.
       - If you are uncertain about the logic behind **Variable**, leave it *as-is*.
@@ -106,7 +106,7 @@ RAGFlow offers HTTP and Python APIs for you to integrate RAGFlow's capabilities
 
 You can use iframe to embed the created chat assistant into a third-party webpage:
 
-1. Before proceeding, you must [acquire an API key](../models/llm_api_key_setup.md); otherwise, an error message would appear.
+1. Before proceeding, you must [acquire an API key](../../develop/acquire_ragflow_api_key.md); otherwise, an error message would appear.
 2. Hover over an intended chat assistant **>** **Edit** to show the **iframe** window:
 
    ![chat-embed](https://raw.githubusercontent.com/infiniflow/ragflow-docs/main/images/embed_chat_into_webpage.jpg)

docs/guides/dataset/_category_.json

@@ -3,6 +3,6 @@
   "position": 0,
   "link": {
     "type": "generated-index",
-    "description": "Guides on configuring a knowledge base."
+    "description": "Guides on configuring a dataset."
   }
 }

docs/guides/dataset/autokeyword_autoquestion.mdx

@@ -6,7 +6,7 @@ slug: /autokeyword_autoquestion
 # Auto-keyword Auto-question
 import APITable from '@site/src/components/APITable';
 
-Use a chat model to generate keywords or questions from each chunk in the knowledge base.
+Use a chat model to generate keywords or questions from each chunk in the dataset.
 
 ---
 
@@ -18,7 +18,7 @@ Enabling this feature increases document indexing time and uses extra tokens, as
 
 ## What is Auto-keyword?
 
-Auto-keyword refers to the auto-keyword generation feature of RAGFlow. It uses a chat model to generate a set of keywords or synonyms from each chunk to correct errors and enhance retrieval accuracy. This feature is implemented as a slider under **Page rank** on the **Configuration** page of your knowledge base.
+Auto-keyword refers to the auto-keyword generation feature of RAGFlow. It uses a chat model to generate a set of keywords or synonyms from each chunk to correct errors and enhance retrieval accuracy. This feature is implemented as a slider under **Page rank** on the **Configuration** page of your dataset.
 
 **Values**:  
 
@@ -33,7 +33,7 @@ Auto-keyword refers to the auto-keyword generation feature of RAGFlow. It uses a
 
 ## What is Auto-question?
 
-Auto-question is a feature of RAGFlow that automatically generates questions from chunks of data using a chat model. These questions (e.g. who, what, and why) also help correct errors and improve the matching of user queries. The feature usually works with FAQ retrieval scenarios involving product manuals or policy documents. And you can find this feature as a slider under **Page rank** on the **Configuration** page of your knowledge base.
+Auto-question is a feature of RAGFlow that automatically generates questions from chunks of data using a chat model. These questions (e.g. who, what, and why) also help correct errors and improve the matching of user queries. The feature usually works with FAQ retrieval scenarios involving product manuals or policy documents. And you can find this feature as a slider under **Page rank** on the **Configuration** page of your dataset.
 
 **Values**:
 
@@ -48,7 +48,7 @@ Auto-question is a feature of RAGFlow that automatically generates questions fro
 
 ## Tips from the community
 
-The Auto-keyword or Auto-question values relate closely to the chunking size in your knowledge base. However, if you are new to this feature and unsure which value(s) to start with, the following are some value settings we gathered from our community. While they may not be accurate, they provide a starting point at the very least.
+The Auto-keyword or Auto-question values relate closely to the chunking size in your dataset. However, if you are new to this feature and unsure which value(s) to start with, the following are some value settings we gathered from our community. While they may not be accurate, they provide a starting point at the very least.
 
 ```mdx-code-block
 <APITable>

docs/guides/dataset/best_practices/_category_.json

@@ -3,6 +3,6 @@
   "position": 11,
   "link": {
     "type": "generated-index",
-    "description": "Best practices on configuring a knowledge base."
+    "description": "Best practices on configuring a dataset."
   }
 }

docs/guides/dataset/best_practices/accelerate_doc_indexing.mdx

@@ -13,7 +13,7 @@ A checklist to speed up document parsing and indexing.
 Please note that some of your settings may consume a significant amount of time. If you often find that document parsing is time-consuming, here is a checklist to consider:
 
 - Use GPU to reduce embedding time.
-- On the configuration page of your knowledge base, switch off **Use RAPTOR to enhance retrieval**.
+- On the configuration page of your dataset, switch off **Use RAPTOR to enhance retrieval**.
 - Extracting knowledge graph (GraphRAG) is time-consuming.
-- Disable **Auto-keyword** and **Auto-question** on the configuration page of your knowledge base, as both depend on the LLM.
-- **v0.17.0+:** If all PDFs in your knowledge base are plain text and do not require GPU-intensive processes like OCR (Optical Character Recognition), TSR (Table Structure Recognition), or DLA (Document Layout Analysis), you can choose **Naive** over **DeepDoc** or other time-consuming large model options in the **Document parser** dropdown. This will substantially reduce document parsing time.
+- Disable **Auto-keyword** and **Auto-question** on the configuration page of your dataset, as both depend on the LLM.
+- **v0.17.0+:** If all PDFs in your dataset are plain text and do not require GPU-intensive processes like OCR (Optical Character Recognition), TSR (Table Structure Recognition), or DLA (Document Layout Analysis), you can choose **Naive** over **DeepDoc** or other time-consuming large model options in the **Document parser** dropdown. This will substantially reduce document parsing time.

docs/guides/dataset/configure_knowledge_base.md

@@ -3,28 +3,28 @@ sidebar_position: -1
 slug: /configure_knowledge_base
 ---
 
-# Configure knowledge base
+# Configure dataset
 
-Knowledge base, hallucination-free chat, and file management are the three pillars of RAGFlow. RAGFlow's AI chats are based on knowledge bases. Each of RAGFlow's knowledge bases serves as a knowledge source, *parsing* files uploaded from your local machine and file references generated in **File Management** into the real 'knowledge' for future AI chats. This guide demonstrates some basic usages of the knowledge base feature, covering the following topics:
+Most of RAGFlow's chat assistants and Agents are based on datasets. Each of RAGFlow's datasets serves as a knowledge source, *parsing* files uploaded from your local machine and file references generated in **File Management** into the real 'knowledge' for future AI chats. This guide demonstrates some basic usages of the dataset feature, covering the following topics:
 
-- Create a knowledge base
-- Configure a knowledge base
-- Search for a knowledge base
-- Delete a knowledge base
+- Create a dataset
+- Configure a dataset
+- Search for a dataset
+- Delete a dataset
 
-## Create knowledge base
+## Create dataset
 
-With multiple knowledge bases, you can build more flexible, diversified question answering. To create your first knowledge base:
+With multiple datasets, you can build more flexible, diversified question answering. To create your first dataset:
 
-![create knowledge base](https://raw.githubusercontent.com/infiniflow/ragflow-docs/main/images/create_knowledge_base.jpg)
+![create dataset](https://raw.githubusercontent.com/infiniflow/ragflow-docs/main/images/create_knowledge_base.jpg)
 
-_Each time a knowledge base is created, a folder with the same name is generated in the **root/.knowledgebase** directory._
+_Each time a dataset is created, a folder with the same name is generated in the **root/.knowledgebase** directory._
 
-## Configure knowledge base
+## Configure dataset
 
-The following screenshot shows the configuration page of a knowledge base. A proper configuration of your knowledge base is crucial for future AI chats. For example, choosing the wrong embedding model or chunking method would cause unexpected semantic loss or mismatched answers in chats. 
+The following screenshot shows the configuration page of a dataset. A proper configuration of your dataset is crucial for future AI chats. For example, choosing the wrong embedding model or chunking method would cause unexpected semantic loss or mismatched answers in chats. 
 
-![knowledge base configuration](https://raw.githubusercontent.com/infiniflow/ragflow-docs/main/images/configure_knowledge_base.jpg)
+![dataset configuration](https://raw.githubusercontent.com/infiniflow/ragflow-docs/main/images/configure_knowledge_base.jpg)
 
 This section covers the following topics:
 
@@ -52,7 +52,7 @@ RAGFlow offers multiple chunking template to facilitate chunking files of differ
 | Presentation |                                                                       | PDF, PPTX                                                                                     |
 | Picture      |                                                                       | JPEG, JPG, PNG, TIF, GIF                                                                      |
 | One          | Each document is chunked in its entirety (as one).                    | DOCX, XLSX, XLS (Excel 97-2003), PDF, TXT                                                      |
-| Tag          | The knowledge base functions as a tag set for the others.             | XLSX, CSV/TXT                                                                                 |
+| Tag          | The dataset functions as a tag set for the others.             | XLSX, CSV/TXT                                                                                 |
 
 You can also change a file's chunking method on the **Datasets** page.
 
@@ -60,7 +60,7 @@ You can also change a file's chunking method on the **Datasets** page.
 
 ### Select embedding model
 
-An embedding model converts chunks into embeddings. It cannot be changed once the knowledge base has chunks. To switch to a different embedding model, you must delete all existing chunks in the knowledge base. The obvious reason is that we *must* ensure that files in a specific knowledge base are converted to embeddings using the *same* embedding model (ensure that they are compared in the same embedding space).
+An embedding model converts chunks into embeddings. It cannot be changed once the dataset has chunks. To switch to a different embedding model, you must delete all existing chunks in the dataset. The obvious reason is that we *must* ensure that files in a specific dataset are converted to embeddings using the *same* embedding model (ensure that they are compared in the same embedding space).
 
 The following embedding models can be deployed locally:
 
@@ -73,19 +73,19 @@ These two embedding models are optimized specifically for English and Chinese, s
 
 ### Upload file
 
-- RAGFlow's **File Management** allows you to link a file to multiple knowledge bases, in which case each target knowledge base holds a reference to the file.
-- In **Knowledge Base**, you are also given the option of uploading a single file or a folder of files (bulk upload) from your local machine to a knowledge base, in which case the knowledge base holds file copies. 
+- RAGFlow's **File Management** allows you to link a file to multiple datasets, in which case each target dataset holds a reference to the file.
+- In **Knowledge Base**, you are also given the option of uploading a single file or a folder of files (bulk upload) from your local machine to a dataset, in which case the dataset holds file copies. 
 
-While uploading files directly to a knowledge base seems more convenient, we *highly* recommend uploading files to **File Management** and then linking them to the target knowledge bases. This way, you can avoid permanently deleting files uploaded to the knowledge base. 
+While uploading files directly to a dataset seems more convenient, we *highly* recommend uploading files to **File Management** and then linking them to the target datasets. This way, you can avoid permanently deleting files uploaded to the dataset. 
 
 ### Parse file
 
-File parsing is a crucial topic in knowledge base configuration. The meaning of file parsing in RAGFlow is twofold: chunking files based on file layout and building embedding and full-text (keyword) indexes on these chunks. After having selected the chunking method and embedding model, you can start parsing a file:
+File parsing is a crucial topic in dataset configuration. The meaning of file parsing in RAGFlow is twofold: chunking files based on file layout and building embedding and full-text (keyword) indexes on these chunks. After having selected the chunking method and embedding model, you can start parsing a file:
 
 ![parse file](https://raw.githubusercontent.com/infiniflow/ragflow-docs/main/images/parse_file.jpg)
 
 - As shown above, RAGFlow allows you to use a different chunking method for a particular file, offering flexibility beyond the default method. 
-- As shown above, RAGFlow allows you to enable or disable individual files, offering finer control over knowledge base-based AI chats. 
+- As shown above, RAGFlow allows you to enable or disable individual files, offering finer control over dataset-based AI chats. 
 
 ### Intervene with file parsing results
 
@@ -122,17 +122,17 @@ RAGFlow uses multiple recall of both full-text search and vector search in its c
 
 See [Run retrieval test](./run_retrieval_test.md) for details.
 
-## Search for knowledge base
+## Search for dataset
 
-As of RAGFlow v0.20.5, the search feature is still in a rudimentary form, supporting only knowledge base search by name.
+As of RAGFlow v0.20.5, the search feature is still in a rudimentary form, supporting only dataset search by name.
 
-![search knowledge base](https://raw.githubusercontent.com/infiniflow/ragflow-docs/main/images/search_datasets.jpg)
+![search dataset](https://raw.githubusercontent.com/infiniflow/ragflow-docs/main/images/search_datasets.jpg)
 
-## Delete knowledge base
+## Delete dataset
 
-You are allowed to delete a knowledge base. Hover your mouse over the three dot of the intended knowledge base card and the **Delete** option appears. Once you delete a knowledge base, the associated folder under **root/.knowledge** directory is AUTOMATICALLY REMOVED. The consequence is:
+You are allowed to delete a dataset. Hover your mouse over the three dot of the intended dataset card and the **Delete** option appears. Once you delete a dataset, the associated folder under **root/.knowledge** directory is AUTOMATICALLY REMOVED. The consequence is:
 
-- The files uploaded directly to the knowledge base are gone;  
+- The files uploaded directly to the dataset are gone;  
 - The file references, which you created from within **File Management**, are gone, but the associated files still exist in **File Management**. 
 
-![delete knowledge base](https://raw.githubusercontent.com/infiniflow/ragflow-docs/main/images/delete_datasets.jpg)
+![delete dataset](https://raw.githubusercontent.com/infiniflow/ragflow-docs/main/images/delete_datasets.jpg)

docs/guides/dataset/construct_knowledge_graph.md

@@ -5,7 +5,7 @@ slug: /construct_knowledge_graph
 
 # Construct knowledge graph
 
-Generate a knowledge graph for your knowledge base.
+Generate a knowledge graph for your dataset.
 
 ---
 
@@ -13,7 +13,7 @@ To enhance multi-hop question-answering, RAGFlow adds a knowledge graph construc
 
 ![Image](https://github.com/user-attachments/assets/1ec21d8e-f255-4d65-9918-69b72dfa142b)
 
-From v0.16.0 onward, RAGFlow supports constructing a knowledge graph on a knowledge base, allowing you to construct a *unified* graph across multiple files within your knowledge base. When a newly uploaded file starts parsing, the generated graph will automatically update.
+From v0.16.0 onward, RAGFlow supports constructing a knowledge graph on a dataset, allowing you to construct a *unified* graph across multiple files within your dataset. When a newly uploaded file starts parsing, the generated graph will automatically update.
 
 :::danger WARNING
 Constructing a knowledge graph requires significant memory, computational resources, and tokens.
@@ -37,7 +37,7 @@ The system's default chat model is used to generate knowledge graph. Before proc
 
 ### Entity types (*Required*)
 
-The types of the entities to extract from your knowledge base. The default types are: **organization**, **person**, **event**, and **category**. Add or remove types to suit your specific knowledge base.
+The types of the entities to extract from your dataset. The default types are: **organization**, **person**, **event**, and **category**. Add or remove types to suit your specific dataset.
 
 ### Method
 
@@ -62,12 +62,12 @@ In a knowledge graph, a community is a cluster of entities linked by relationshi
 
 ## Procedure
 
-1. On the **Configuration** page of your knowledge base, switch on **Extract knowledge graph** or adjust its settings as needed, and click **Save** to confirm your changes.
+1. On the **Configuration** page of your dataset, switch on **Extract knowledge graph** or adjust its settings as needed, and click **Save** to confirm your changes.
 
-   - *The default knowledge graph configurations for your knowledge base are now set and files uploaded from this point onward will automatically use these settings during parsing.*
+   - *The default knowledge graph configurations for your dataset are now set and files uploaded from this point onward will automatically use these settings during parsing.*
    - *Files parsed before this update will retain their original knowledge graph settings.*
 
-2. The knowledge graph of your knowledge base does *not* automatically update *until* a newly uploaded file is parsed.
+2. The knowledge graph of your dataset does *not* automatically update *until* a newly uploaded file is parsed.
 
    _A **Knowledge graph** entry appears under **Configuration** once a knowledge graph is created._
 
@@ -75,13 +75,13 @@ In a knowledge graph, a community is a cluster of entities linked by relationshi
 4. To use the created knowledge graph, do either of the following:
    
    - In the **Chat setting** panel of your chat app, switch on the **Use knowledge graph** toggle.
-   - If you are using an agent, click the **Retrieval** agent component to specify the knowledge base(s) and switch on the **Use knowledge graph** toggle.
+   - If you are using an agent, click the **Retrieval** agent component to specify the dataset(s) and switch on the **Use knowledge graph** toggle.
 
 ## Frequently asked questions
 
-### Can I have different knowledge graph settings for different files in my knowledge base?
+### Can I have different knowledge graph settings for different files in my dataset?
 
-Yes, you can. Just one graph is generated per knowledge base. The smaller graphs of your files will be *combined* into one big, unified graph at the end of the graph extraction process.
+Yes, you can. Just one graph is generated per dataset. The smaller graphs of your files will be *combined* into one big, unified graph at the end of the graph extraction process.
 
 ### Does the knowledge graph automatically update when I remove a related file?
 
@@ -89,7 +89,7 @@ Nope. The knowledge graph does *not* automatically update *until* a newly upload
 
 ### How to remove a generated knowledge graph?
 
-To remove the generated knowledge graph, delete all related files in your knowledge base. Although the **Knowledge graph** entry will still be visible, the graph has actually been deleted.
+To remove the generated knowledge graph, delete all related files in your dataset. Although the **Knowledge graph** entry will still be visible, the graph has actually been deleted.
 
 ### Where is the created knowledge graph stored?
 

docs/guides/dataset/enable_excel2html.md

@@ -12,7 +12,7 @@ Convert complex Excel spreadsheets into HTML tables.
 When using the **General** chunking method, you can enable the **Excel to HTML** toggle to convert spreadsheet files into HTML tables. If it is disabled, spreadsheet tables will be represented as key-value pairs. For complex tables that cannot be simply represented this way, you must enable this feature.
 
 :::caution WARNING
-The feature is disabled by default. If your knowledge base contains spreadsheets with complex tables and you do not enable this feature, RAGFlow will not throw an error but your tables are likely to be garbled.
+The feature is disabled by default. If your dataset contains spreadsheets with complex tables and you do not enable this feature, RAGFlow will not throw an error but your tables are likely to be garbled.
 :::
 
 ## Scenarios
@@ -27,12 +27,12 @@ Works with complex tables that cannot be represented as key-value pairs. Example
 
 ## Procedure
 
-1. On your knowledge base's **Configuration** page, select **General** as the chunking method.
+1. On your dataset's **Configuration** page, select **General** as the chunking method.
 
    _The **Excel to HTML** toggle appears._
 
-2. Enable **Excel to HTML** if your knowledge base contains complex spreadsheet tables that cannot be represented as key-value pairs.
-3. Leave **Excel to HTML** disabled if your knowledge base has no spreadsheet tables or if its spreadsheet tables can be represented as key-value pairs.
+2. Enable **Excel to HTML** if your dataset contains complex spreadsheet tables that cannot be represented as key-value pairs.
+3. Leave **Excel to HTML** disabled if your dataset has no spreadsheet tables or if its spreadsheet tables can be represented as key-value pairs.
 4. If question-answering regarding complex tables is unsatisfactory, check if **Excel to HTML** is enabled.
 
 ## Frequently asked questions

docs/guides/dataset/enable_raptor.md

@@ -43,7 +43,7 @@ The system's default chat model is used to summarize clustered content. Before p
 
 ## Configurations
 
-The RAPTOR feature is disabled by default. To enable it, manually switch on the **Use RAPTOR to enhance retrieval** toggle on your knowledge base's **Configuration** page.
+The RAPTOR feature is disabled by default. To enable it, manually switch on the **Use RAPTOR to enhance retrieval** toggle on your dataset's **Configuration** page.
 
 ### Prompt
 

docs/guides/dataset/run_retrieval_test.md

@@ -5,11 +5,11 @@ slug: /run_retrieval_test
 
 # Run retrieval test
 
-Conduct a retrieval test on your knowledge base to check whether the intended chunks can be retrieved.
+Conduct a retrieval test on your dataset to check whether the intended chunks can be retrieved.
 
 ---
 
-After your files are uploaded and parsed, it is recommended that you run a retrieval test before proceeding with the chat assistant configuration. Running a retrieval test is *not* an unnecessary or superfluous step at all! Just like fine-tuning a precision instrument, RAGFlow requires careful tuning to deliver optimal question answering performance. Your knowledge base settings, chat assistant configurations, and the specified large and small models can all significantly impact the final results. Running a retrieval test verifies whether the intended chunks can be recovered, allowing you to quickly identify areas for improvement or pinpoint any issue that needs addressing. For instance, when debugging your question answering system, if you know that the correct chunks can be retrieved, you can focus your efforts elsewhere. For example, in issue [#5627](https://github.com/infiniflow/ragflow/issues/5627), the problem was found to be due to the LLM's limitations.
+After your files are uploaded and parsed, it is recommended that you run a retrieval test before proceeding with the chat assistant configuration. Running a retrieval test is *not* an unnecessary or superfluous step at all! Just like fine-tuning a precision instrument, RAGFlow requires careful tuning to deliver optimal question answering performance. Your dataset settings, chat assistant configurations, and the specified large and small models can all significantly impact the final results. Running a retrieval test verifies whether the intended chunks can be recovered, allowing you to quickly identify areas for improvement or pinpoint any issue that needs addressing. For instance, when debugging your question answering system, if you know that the correct chunks can be retrieved, you can focus your efforts elsewhere. For example, in issue [#5627](https://github.com/infiniflow/ragflow/issues/5627), the problem was found to be due to the LLM's limitations.
 
 During a retrieval test, chunks created from your specified chunking method are retrieved using a hybrid search. This search combines weighted keyword similarity with either weighted vector cosine similarity or a weighted reranking score, depending on your settings:
 
@@ -65,7 +65,7 @@ Using a knowledge graph in a retrieval test will significantly increase the time
 To perform a [cross-language search](../../references/glossary.mdx#cross-language-search), select one or more target languages from the dropdown menu. The system’s default chat model will then translate your query entered in the Test text field into the selected target language(s). This translation ensures accurate semantic matching across languages, allowing you to retrieve relevant results regardless of language differences.
 
 :::tip NOTE
-- When selecting target languages, please ensure that these languages are present in the knowledge base to guarantee an effective search.
+- When selecting target languages, please ensure that these languages are present in the dataset to guarantee an effective search.
 - If no target language is selected, the system will search only in the language of your query, which may cause relevant information in other languages to be missed.
 :::
 
@@ -75,7 +75,7 @@ This field is where you put in your testing query.
 
 ## Procedure
 
-1. Navigate to the **Retrieval testing** page of your knowledge base, enter your query in **Test text**, and click **Testing** to run the test.
+1. Navigate to the **Retrieval testing** page of your dataset, enter your query in **Test text**, and click **Testing** to run the test.
 2. If the results are unsatisfactory, tune the options listed in the Configuration section and rerun the test.
 
    *The following is a screenshot of a retrieval test conducted without using knowledge graph. It demonstrates a hybrid search combining weighted keyword similarity and weighted vector cosine similarity. The overall hybrid similarity score is 28.56, calculated as 25.17 (term similarity score) x 0.7 + 36.49 (vector similarity score) x 0.3:*  

docs/guides/dataset/select_pdf_parser.md

@@ -27,7 +27,7 @@ RAGFlow isn't one-size-fits-all. It is built for flexibility and supports deeper
 
 ## Procedure
 
-1. On your knowledge base's **Configuration** page, select a chunking method, say **General**.
+1. On your dataset's **Configuration** page, select a chunking method, say **General**.
 
    _The **PDF parser** dropdown menu appears._
 

docs/guides/dataset/set_metadata.md

@@ -9,7 +9,7 @@ Add metadata to an uploaded file
 
 ---
 
-On the **Dataset** page of your knowledge base, you can add metadata to any uploaded file. This approach enables you to 'tag' additional information like URL, author, date, and more to an existing file. In an AI-powered chat, such information will be sent to the LLM with the retrieved chunks for content generation.
+On the **Dataset** page of your dataset, you can add metadata to any uploaded file. This approach enables you to 'tag' additional information like URL, author, date, and more to an existing file. In an AI-powered chat, such information will be sent to the LLM with the retrieved chunks for content generation.
 
 For example, if you have a dataset of HTML files and want the LLM to cite the source URL when responding to your query, add a `"url"` parameter to each file's metadata.
 

docs/guides/dataset/set_page_rank.md

@@ -11,15 +11,15 @@ Create a step-retrieval strategy using page rank.
 
 ## Scenario
 
-In an AI-powered chat, you can configure a chat assistant or an agent to respond using knowledge retrieved from multiple specified knowledge bases (datasets), provided that they employ the same embedding model. In situations where you prefer information from certain knowledge base(s) to take precedence or to be retrieved first, you can use RAGFlow's page rank feature to increase the ranking of chunks from these knowledge bases. For example, if you have configured a chat assistant to draw from two knowledge bases, knowledge base A for 2024 news and knowledge base B for 2023 news, but wish to prioritize news from year 2024, this feature is particularly useful.
+In an AI-powered chat, you can configure a chat assistant or an agent to respond using knowledge retrieved from multiple specified datasets (datasets), provided that they employ the same embedding model. In situations where you prefer information from certain dataset(s) to take precedence or to be retrieved first, you can use RAGFlow's page rank feature to increase the ranking of chunks from these datasets. For example, if you have configured a chat assistant to draw from two datasets, dataset A for 2024 news and dataset B for 2023 news, but wish to prioritize news from year 2024, this feature is particularly useful.
 
 :::info NOTE
-It is important to note that this 'page rank' feature operates at the level of the entire knowledge base rather than on individual files or documents.
+It is important to note that this 'page rank' feature operates at the level of the entire dataset rather than on individual files or documents.
 :::
 
 ## Configuration
 
-On the **Configuration** page of your knowledge base, drag the slider under **Page rank** to set the page rank value for your knowledge base. You are also allowed to input the intended page rank value in the field next to the slider.
+On the **Configuration** page of your dataset, drag the slider under **Page rank** to set the page rank value for your dataset. You are also allowed to input the intended page rank value in the field next to the slider.
 
 :::info NOTE
 The page rank value must be an integer. Range: [0,100]
@@ -36,4 +36,4 @@ If you set the page rank value to a non-integer, say 1.7, it will be rounded dow
 
 If you configure a chat assistant's **similarity threshold** to 0.2, only chunks with a hybrid score greater than 0.2 x 100 = 20 will be retrieved and sent to the chat model for content generation. This initial filtering step is crucial for narrowing down relevant information.
 
-If you have assigned a page rank of 1 to knowledge base A (2024 news) and 0 to knowledge base B (2023 news), the final hybrid scores of the retrieved chunks will be adjusted accordingly. A chunk retrieved from knowledge base A with an initial score of 50 will receive a boost of 1 x 100 = 100 points, resulting in a final score of 50 + 1 x 100 = 150. In this way, chunks retrieved from knowledge base A will always precede chunks from knowledge base B.
+If you have assigned a page rank of 1 to dataset A (2024 news) and 0 to dataset B (2023 news), the final hybrid scores of the retrieved chunks will be adjusted accordingly. A chunk retrieved from dataset A with an initial score of 50 will receive a boost of 1 x 100 = 100 points, resulting in a final score of 50 + 1 x 100 = 150. In this way, chunks retrieved from dataset A will always precede chunks from dataset B.

docs/guides/dataset/use_tag_sets.md

@@ -9,9 +9,9 @@ Use a tag set to auto-tag chunks in your datasets.
 
 ---
 
-Retrieval accuracy is the touchstone for a production-ready RAG framework. In addition to retrieval-enhancing approaches like auto-keyword, auto-question, and knowledge graph, RAGFlow introduces an auto-tagging feature to address semantic gaps. The auto-tagging feature automatically maps tags in the user-defined tag sets to relevant chunks within your knowledge base based on similarity with each chunk. This automation mechanism allows you to apply an additional "layer" of domain-specific knowledge to existing datasets, which is particularly useful when dealing with a large number of chunks.
+Retrieval accuracy is the touchstone for a production-ready RAG framework. In addition to retrieval-enhancing approaches like auto-keyword, auto-question, and knowledge graph, RAGFlow introduces an auto-tagging feature to address semantic gaps. The auto-tagging feature automatically maps tags in the user-defined tag sets to relevant chunks within your dataset based on similarity with each chunk. This automation mechanism allows you to apply an additional "layer" of domain-specific knowledge to existing datasets, which is particularly useful when dealing with a large number of chunks.
 
-To use this feature, ensure you have at least one properly configured tag set, specify the tag set(s) on the **Configuration** page of your knowledge base (dataset), and then re-parse your documents to initiate the auto-tagging process. During this process, each chunk in your dataset is compared with every entry in the specified tag set(s), and tags are automatically applied based on similarity.
+To use this feature, ensure you have at least one properly configured tag set, specify the tag set(s) on the **Configuration** page of your dataset, and then re-parse your documents to initiate the auto-tagging process. During this process, each chunk in your dataset is compared with every entry in the specified tag set(s), and tags are automatically applied based on similarity.
 
 ## Scenarios
 
@@ -19,7 +19,7 @@ Auto-tagging applies in situations where chunks are so similar to each other tha
 
 ## 1. Create tag set
 
-You can consider a tag set as a closed set, and the tags to attach to the chunks in your dataset (knowledge base) are *exclusively* from the specified tag set. You use a tag set to "inform" RAGFlow which chunks to tag and which tags to apply.
+You can consider a tag set as a closed set, and the tags to attach to the chunks in your dataset are *exclusively* from the specified tag set. You use a tag set to "inform" RAGFlow which chunks to tag and which tags to apply.
 
 ### Prepare a tag table file
 
@@ -41,8 +41,8 @@ As a rule of thumb, consider including the following entries in your tag table:
 A tag set is *not* involved in document indexing or retrieval. Do not specify a tag set when configuring your chat assistant or agent.
 :::
 
-1. Click **+ Create knowledge base** to create a knowledge base.
-2. Navigate to the **Configuration** page of the created knowledge base and choose **Tag** as the default chunking method.
+1. Click **+ Create dataset** to create a dataset.
+2. Navigate to the **Configuration** page of the created dataset and choose **Tag** as the default chunking method.
 3. Navigate to the **Dataset** page and upload and parse your table file in XLSX, CSV, or TXT formats.  
    _A tag cloud appears under the **Tag view** section, indicating the tag set is created:_  
    ![Image](https://github.com/user-attachments/assets/abefbcbf-c130-4abe-95e1-267b0d2a0505)
@@ -53,7 +53,7 @@ A tag set is *not* involved in document indexing or retrieval. Do not specify a
 
 Once a tag set is created, you can apply it to your dataset:
 
-1. Navigate to the **Configuration** page of your knowledge base (dataset).
+1. Navigate to the **Configuration** page of your dataset.
 2. Select the tag set from the **Tag sets** dropdown and click **Save** to confirm.
 
 :::tip NOTE
@@ -94,9 +94,9 @@ If you add new table files to your tag set, it is at your own discretion whether
 
 Yes, you can. Usually one tag set suffices. When using multiple tag sets, ensure they are independent of each other; otherwise, consider merging your tag sets.
 
-### Difference between a tag set and a standard knowledge base?
+### Difference between a tag set and a standard dataset?
 
-A standard knowledge base is a dataset. It will be searched by RAGFlow's document engine and the retrieved chunks will be fed to the LLM. In contrast, a tag set is used solely to attach tags to chunks within your dataset. It does not directly participate in the retrieval process, and you should not choose a tag set when selecting datasets for your chat assistant or agent.
+A standard dataset is a dataset. It will be searched by RAGFlow's document engine and the retrieved chunks will be fed to the LLM. In contrast, a tag set is used solely to attach tags to chunks within your dataset. It does not directly participate in the retrieval process, and you should not choose a tag set when selecting datasets for your chat assistant or agent.
 
 ### Difference between auto-tag and auto-keyword?
 

docs/guides/manage_files.md

@@ -5,10 +5,10 @@ slug: /manage_files
 
 # Files
 
-Knowledge base, hallucination-free chat, and file management are the three pillars of RAGFlow. RAGFlow's file management allows you to upload files individually or in bulk. You can then link an uploaded file to multiple target knowledge bases. This guide showcases some basic usages of the file management feature.
+RAGFlow's file management allows you to upload files individually or in bulk. You can then link an uploaded file to multiple target datasets. This guide showcases some basic usages of the file management feature.
 
 :::info IMPORTANT
-Compared to uploading files directly to various knowledge bases, uploading them to RAGFlow's file management and then linking them to different knowledge bases is *not* an unnecessary step, particularly when you want to delete some parsed files or an entire knowledge base but retain the original files.
+Compared to uploading files directly to various datasets, uploading them to RAGFlow's file management and then linking them to different datasets is *not* an unnecessary step, particularly when you want to delete some parsed files or an entire dataset but retain the original files.
 :::
 
 ## Create folder
@@ -18,7 +18,7 @@ RAGFlow's file management allows you to establish your file system with nested f
 ![create new folder](https://github.com/infiniflow/ragflow/assets/93570324/3a37a5f4-43a6-426d-a62a-e5cd2ff7a533)
 
 :::caution NOTE
-Each knowledge base in RAGFlow has a corresponding folder under the **root/.knowledgebase** directory. You are not allowed to create a subfolder within it.
+Each dataset in RAGFlow has a corresponding folder under the **root/.knowledgebase** directory. You are not allowed to create a subfolder within it.
 :::
 
 ## Upload file
@@ -39,13 +39,13 @@ RAGFlow's file management supports previewing files in the following formats:
 
 ![preview](https://github.com/infiniflow/ragflow/assets/93570324/2e931362-8bbf-482c-ac86-b68b09d331bc)
 
-## Link file to knowledge bases
+## Link file to datasets
 
-RAGFlow's file management allows you to *link* an uploaded file to multiple knowledge bases, creating a file reference in each target knowledge base. Therefore, deleting a file in your file management will AUTOMATICALLY REMOVE all related file references across the knowledge bases.
+RAGFlow's file management allows you to *link* an uploaded file to multiple datasets, creating a file reference in each target dataset. Therefore, deleting a file in your file management will AUTOMATICALLY REMOVE all related file references across the datasets.
 
 ![link knowledgebase](https://github.com/infiniflow/ragflow/assets/93570324/6c6b8db4-3269-4e35-9434-6089887e3e3f)
 
-You can link your file to one knowledge base or multiple knowledge bases at one time: 
+You can link your file to one dataset or multiple datasets at one time: 
 
 ![link multiple kb](https://github.com/infiniflow/ragflow/assets/93570324/6c508803-fb1f-435d-b688-683066fd7fff)
 
@@ -79,7 +79,7 @@ To bulk delete files or folders:
 ![bulk delete](https://github.com/infiniflow/ragflow/assets/93570324/519b99ab-ec7f-4c8a-8cea-e0b6dcb3cb46)
 
 > - You are not allowed to delete the **root/.knowledgebase** folder. 
-> - Deleting files that have been linked to knowledge bases will **AUTOMATICALLY REMOVE** all associated file references across the knowledge bases.
+> - Deleting files that have been linked to datasets will **AUTOMATICALLY REMOVE** all associated file references across the datasets.
 
 ## Download uploaded file
 

docs/guides/models/deploy_local_llm.mdx

@@ -91,7 +91,7 @@ In RAGFlow, click on your logo on the top right of the page **>** **Model provid
 In the popup window, complete basic settings for Ollama:
 
 1. Ensure that your model name and type match those been pulled at step 1 (Deploy Ollama using Docker). For example, (`llama3.2` and `chat`) or (`bge-m3` and `embedding`).
-2. In Ollama base URL, put the URL you found in step 2 followed by `/v1`, i.e. `http://host.docker.internal:11434/v1`, `http://localhost:11434/v1` or `http://${IP_OF_OLLAMA_MACHINE}:11434/v1`.
+2. Put in the Ollama base URL, i.e. `http://host.docker.internal:11434`, `http://localhost:11434` or `http://${IP_OF_OLLAMA_MACHINE}:11434`.
 3. OPTIONAL: Switch on the toggle under **Does it support Vision?** if your model includes an image-to-text model.
 
 
@@ -164,7 +164,7 @@ Click on your logo **>** **Model providers** **>** **System Model Settings** to
 
 Update your chat model accordingly in **Chat Configuration**:
 
-> If your local model is an embedding model, update it on the configuration page of your knowledge base.
+> If your local model is an embedding model, update it on the configuration page of your dataset.
 
 ## Deploy a local model using IPEX-LLM
 

docs/guides/run_health_check.md

@@ -31,3 +31,79 @@ You can click on a specific 30-second time interval to view the details of compl
 ![done_tasks](https://github.com/user-attachments/assets/49b25ec4-03af-48cf-b2e5-c892f6eaa261)
 
 ![done_vs_failed](https://github.com/user-attachments/assets/eaa928d0-a31c-4072-adea-046091e04599)
+
+## API Health Check
+
+In addition to checking the system dependencies from the **avatar > System** page in the UI, you can directly query the backend health check endpoint:
+
+```bash
+http://IP_OF_YOUR_MACHINE/v1/system/healthz
+```
+
+Here `<port>` refers to the actual port of your backend service (e.g., `7897`, `9222`, etc.).
+
+Key points:
+- **No login required** (no `@login_required` decorator)
+- Returns results in JSON format
+- If all dependencies are healthy → HTTP **200 OK**
+- If any dependency fails → HTTP **500 Internal Server Error**
+
+### Example 1: All services healthy (HTTP 200)
+
+```bash
+http://127.0.0.1/v1/system/healthz
+```
+
+Response:
+
+```http
+HTTP/1.1 200 OK
+Content-Type: application/json
+Content-Length: 120
+
+```
+
+Explanation:
+- Database (MySQL/Postgres), Redis, document engine (Elasticsearch/Infinity), and object storage (MinIO) are all healthy.
+- The `status` field returns `"ok"`.
+
+### Example 2: One service unhealthy (HTTP 500)
+
+For example, if Redis is down:
+
+Response:
+
+```http
+HTTP/1.1 500 INTERNAL SERVER ERROR
+Content-Type: application/json
+Content-Length: 300
+
+```
+
+Explanation:
+- `redis` is marked as `"nok"`, with detailed error info under `_meta.redis.error`.
+- The overall `status` is `"nok"`, so the endpoint returns 500.
+
+---
+
+This endpoint allows you to monitor RAGFlow’s core dependencies programmatically in scripts or external monitoring systems, without relying on the frontend UI.
+  "redis": "nok",
+  "doc_engine": "ok",
+  "storage": "ok",
+  "status": "nok",
+  "_meta": {
+    "redis": {
+      "elapsed": "5.2",
+      "error": "Lost connection!"
+    }
+  }
+}
+```
+
+Explanation:
+- `redis` is marked as `"nok"`, with detailed error info under `_meta.redis.error`.
+- The overall `status` is `"nok"`, so the endpoint returns 500.
+
+---
+
+This endpoint allows you to monitor RAGFlow’s core dependencies programmatically in scripts or external monitoring systems, without relying on the frontend UI.

docs/guides/team/join_or_leave_team.md

@@ -11,7 +11,7 @@ Accept an invite to join a team, decline an invite, or leave a team.
 
 Once you join a team, you can do the following:
 
-- Upload documents to the team owner's shared datasets (knowledge bases).
+- Upload documents to the team owner's shared datasets.
 - Parse documents in the team owner's shared datasets.
 - Use the team owner's shared Agents.
 
@@ -22,7 +22,7 @@ You cannot invite users to a team unless you are its owner.
 ## Prerequisites
 
 1. Ensure that your Email address that received the team invitation is associated with a RAGFlow user account.
-2. The team owner should share his knowledge bases by setting their **Permission** to **Team**.
+2. The team owner should share his datasets by setting their **Permission** to **Team**.
 
 ## Accept or decline team invite
 
@@ -32,6 +32,6 @@ You cannot invite users to a team unless you are its owner.
 
 _On the **Team** page, you can view the information about members of your team and the teams you have joined._
 
-_After accepting the team invite, you should be able to view and update the team owner's knowledge bases whose **Permissions** is set to **Team**._
+_After accepting the team invite, you should be able to view and update the team owner's datasets whose **Permissions** is set to **Team**._
 
 ## Leave a joined team

docs/guides/team/manage_team_members.md

@@ -11,7 +11,7 @@ Invite or remove team members.
 
 By default, each RAGFlow user is assigned a single team named after their name. RAGFlow allows you to invite RAGFlow users to your team. Your team members can help you:
 
-- Upload documents to your shared datasets (knowledge bases).
+- Upload documents to your shared datasets.
 - Parse documents in your shared datasets.
 - Use your shared Agents.
 
@@ -23,7 +23,7 @@ By default, each RAGFlow user is assigned a single team named after their name.
 ## Prerequisites
 
 1. Ensure that the invited team member is a RAGFlow user and that the Email address used is associated with a RAGFlow user account.
-2. To allow your team members to view and update your knowledge base, ensure that you set **Permissions** on its **Configuration** page from **Only me** to **Team**.
+2. To allow your team members to view and update your dataset, ensure that you set **Permissions** on its **Configuration** page from **Only me** to **Team**.
 
 ## Invite team members
 

docs/guides/team/share_knowledge_bases.md

@@ -3,16 +3,16 @@ sidebar_position: 4
 slug: /share_datasets
 ---
 
-# Share knowledge base
+# Share dataset
 
-Share a knowledge base with team members.
+Share a dataset with team members.
 
 ---
 
-When ready, you may share your knowledge bases with your team members so that they can upload and parse files in them. Please note that your knowledge bases are not shared automatically; you must manually enable sharing by selecting the appropriate **Permissions** radio button:
+When ready, you may share your datasets with your team members so that they can upload and parse files in them. Please note that your datasets are not shared automatically; you must manually enable sharing by selecting the appropriate **Permissions** radio button:
 
-1. Navigate to the knowledge base's **Configuration** page.
+1. Navigate to the dataset's **Configuration** page.
 2. Change **Permissions** from **Only me** to **Team**.
 3. Click **Save** to apply your changes.
 
-*Once completed, your team members will see your shared knowledge bases.*
+*Once completed, your team members will see your shared datasets.*

docs/guides/upgrade_ragflow.mdx

@@ -105,9 +105,9 @@ RAGFLOW_IMAGE=infiniflow/ragflow:v0.20.5
 
 ## Frequently asked questions
 
-### Do I need to back up my knowledge bases before upgrading RAGFlow?
+### Do I need to back up my datasets before upgrading RAGFlow?
 
-No, you do not need to. Upgrading RAGFlow in itself will *not* remove your uploaded data or knowledge base settings. However, be aware that `docker compose -f docker/docker-compose.yml down -v` will remove Docker container volumes, resulting in data loss.
+No, you do not need to. Upgrading RAGFlow in itself will *not* remove your uploaded data or dataset settings. However, be aware that `docker compose -f docker/docker-compose.yml down -v` will remove Docker container volumes, resulting in data loss.
 
 ### Upgrade RAGFlow in an offline environment (without Internet access)
 

docs/quickstart.mdx

@@ -13,7 +13,7 @@ RAGFlow is an open-source RAG (Retrieval-Augmented Generation) engine based on d
 This quick start guide describes a general process from:
 
 - Starting up a local RAGFlow server, 
-- Creating a knowledge base, 
+- Creating a dataset, 
 - Intervening with file parsing, to 
 - Establishing an AI chat based on your datasets.
 
@@ -280,29 +280,29 @@ To add and configure an LLM:
 
 > Some models, such as the image-to-text model **qwen-vl-max**, are subsidiary to a specific LLM. And you may need to update your API key to access these models. 
 
-## Create your first knowledge base
+## Create your first dataset
 
-You are allowed to upload files to a knowledge base in RAGFlow and parse them into datasets. A knowledge base is virtually a collection of datasets. Question answering in RAGFlow can be based on a particular knowledge base or multiple knowledge bases. File formats that RAGFlow supports include documents (PDF, DOC, DOCX, TXT, MD, MDX), tables (CSV, XLSX, XLS), pictures (JPEG, JPG, PNG, TIF, GIF), and slides (PPT, PPTX).
+You are allowed to upload files to a dataset in RAGFlow and parse them into datasets. A dataset is virtually a collection of datasets. Question answering in RAGFlow can be based on a particular dataset or multiple datasets. File formats that RAGFlow supports include documents (PDF, DOC, DOCX, TXT, MD, MDX), tables (CSV, XLSX, XLS), pictures (JPEG, JPG, PNG, TIF, GIF), and slides (PPT, PPTX).
 
-To create your first knowledge base:
+To create your first dataset:
 
 1. Click the **Dataset** tab in the top middle of the page **>** **Create dataset**.
 
-2. Input the name of your knowledge base and click **OK** to confirm your changes.
+2. Input the name of your dataset and click **OK** to confirm your changes.
 
-   _You are taken to the **Configuration** page of your knowledge base._
+   _You are taken to the **Configuration** page of your dataset._
 
-   ![knowledge base configuration](https://raw.githubusercontent.com/infiniflow/ragflow-docs/main/images/configure_knowledge_base.jpg)
+   ![dataset configuration](https://raw.githubusercontent.com/infiniflow/ragflow-docs/main/images/configure_knowledge_base.jpg)
 
-3. RAGFlow offers multiple chunk templates that cater to different document layouts and file formats. Select the embedding model and chunking method (template) for your knowledge base. 
+3. RAGFlow offers multiple chunk templates that cater to different document layouts and file formats. Select the embedding model and chunking method (template) for your dataset. 
 
 :::danger IMPORTANT 
-Once you have selected an embedding model and used it to parse a file, you are no longer allowed to change it. The obvious reason is that we must ensure that all files in a specific knowledge base are parsed using the *same* embedding model (ensure that they are being compared in the same embedding space). 
+Once you have selected an embedding model and used it to parse a file, you are no longer allowed to change it. The obvious reason is that we must ensure that all files in a specific dataset are parsed using the *same* embedding model (ensure that they are being compared in the same embedding space). 
 :::
 
-   _You are taken to the **Dataset** page of your knowledge base._
+   _You are taken to the **Dataset** page of your dataset._
 
-4. Click **+ Add file** **>** **Local files** to start uploading a particular file to the knowledge base. 
+4. Click **+ Add file** **>** **Local files** to start uploading a particular file to the dataset. 
 
 5. In the uploaded file entry, click the play button to start file parsing:
 
@@ -341,17 +341,17 @@ You can add keywords or questions to a file chunk to improve its ranking for que
 
 ## Set up an AI chat
 
-Conversations in RAGFlow are based on a particular knowledge base or multiple knowledge bases. Once you have created your knowledge base and finished file parsing, you can go ahead and start an AI conversation.
+Conversations in RAGFlow are based on a particular dataset or multiple datasets. Once you have created your dataset and finished file parsing, you can go ahead and start an AI conversation.
 
 1. Click the **Chat** tab in the middle top of the mage **>** **Create an assistant** to show the **Chat Configuration** dialogue *of your next dialogue*.
    > RAGFlow offer the flexibility of choosing a different chat model for each dialogue, while allowing you to set the default models in **System Model Settings**.
 
 2. Update **Assistant settings**:
 
-   - Name your assistant and specify your knowledge bases.
+   - Name your assistant and specify your datasets.
    - **Empty response**:
-     - If you wish to *confine* RAGFlow's answers to your knowledge bases, leave a response here. Then when it doesn't retrieve an answer, it *uniformly* responds with what you set here.
-     - If you wish RAGFlow to *improvise* when it doesn't retrieve an answer from your knowledge bases, leave it blank, which may give rise to hallucinations.
+     - If you wish to *confine* RAGFlow's answers to your datasets, leave a response here. Then when it doesn't retrieve an answer, it *uniformly* responds with what you set here.
+     - If you wish RAGFlow to *improvise* when it doesn't retrieve an answer from your datasets, leave it blank, which may give rise to hallucinations.
 
 3. Update **Prompt engine** or leave it as is for the beginning.
 

docs/references/http_api_reference.md

@@ -1856,7 +1856,7 @@ curl --request POST \
   - `false`: Disable highlighting of matched terms (default).
 - `"cross_languages"`: (*Body parameter*) `list[string]`  
   The languages that should be translated into, in order to achieve keywords retrievals in different languages.
-- `"metadata_condition"`: (*Body parameter*), `object`
+- `"metadata_condition"`: (*Body parameter*), `object`  
   The metadata condition for filtering chunks.
 #### Response
 
@@ -4102,3 +4102,77 @@ Failure:
 ```
 
 ---
+
+### System
+---
+### Check system health
+
+**GET** `/v1/system/healthz`
+
+Check the health status of RAGFlow’s dependencies (database, Redis, document engine, object storage).
+
+#### Request
+
+- Method: GET
+- URL: `/v1/system/healthz`
+- Headers:
+  - 'Content-Type: application/json'
+  (no Authorization required)
+
+##### Request example
+
+```bash
+curl --request GET
+     --url http://{address}/v1/system/healthz
+     --header 'Content-Type: application/json'
+```
+
+##### Request parameters
+
+- `address`: (*Path parameter*), string  
+  The host and port of the backend service (e.g., `localhost:7897`).
+
+---
+
+#### Responses
+
+- **200 OK** – All services healthy
+
+```http
+HTTP/1.1 200 OK
+Content-Type: application/json
+
+{
+  "db": "ok",
+  "redis": "ok",
+  "doc_engine": "ok",
+  "storage": "ok",
+  "status": "ok"
+}
+```
+
+- **500 Internal Server Error** – At least one service unhealthy
+
+```http
+HTTP/1.1 500 INTERNAL SERVER ERROR
+Content-Type: application/json
+
+{
+  "db": "ok",
+  "redis": "nok",
+  "doc_engine": "ok",
+  "storage": "ok",
+  "status": "nok",
+  "_meta": {
+    "redis": {
+      "elapsed": "5.2",
+      "error": "Lost connection!"
+    }
+  }
+}
+```
+
+Explanation:  
+- Each service is reported as "ok" or "nok".  
+- The top-level `status` reflects overall health.  
+- If any service is "nok", detailed error info appears in `_meta`.  

docs/references/python_api_reference.md

@@ -85,11 +85,11 @@ completion = client.chat.completions.create(
 )
 
 if stream:
-for chunk in completion:
-    print(chunk)
-    if reference and chunk.choices[0].finish_reason == "stop":
-        print(f"Reference:\n{chunk.choices[0].delta.reference}")
-        print(f"Final content:\n{chunk.choices[0].delta.final_content}")
+    for chunk in completion:
+        print(chunk)
+        if reference and chunk.choices[0].finish_reason == "stop":
+            print(f"Reference:\n{chunk.choices[0].delta.reference}")
+            print(f"Final content:\n{chunk.choices[0].delta.final_content}")
 else:
     print(completion.choices[0].message.content)
     if reference:
@@ -977,7 +977,7 @@ The languages that should be translated into, in order to achieve keywords retri
 
 ##### metadata_condition: `dict`
 
-filter condition for meta_fields
+filter condition for `meta_fields`.
 
 #### Returns
 

docs/references/supported_models.mdx

@@ -65,6 +65,7 @@ A complete list of models supported by RAGFlow, which will continue to expand.
 | 01.AI                 | :heavy_check_mark: |                    |                    |                    |                    |                    |
 | DeepInfra             | :heavy_check_mark: | :heavy_check_mark: |                    |                    | :heavy_check_mark: | :heavy_check_mark: |
 | 302.AI                | :heavy_check_mark: | :heavy_check_mark: | :heavy_check_mark: | :heavy_check_mark: |                    |                    |
+| CometAPI              | :heavy_check_mark: | :heavy_check_mark: | :heavy_check_mark: | :heavy_check_mark: |                    |                    |
 
 ```mdx-code-block
 </APITable>

docs/release_notes.md

@@ -28,11 +28,11 @@ Released on September 10, 2025.
 
 ### Improvements
 
-- Agent Performance Optimized: Improved planning and reflection speed for simple tasks; optimized concurrent tool calls for parallelizable scenarios, significantly reducing overall response time.
-- Agent Prompt Framework exposed: Developers can now customize and override framework-level prompts in the system prompt section, enhancing flexibility and control.
-- Execute SQL Component Enhanced: Replaced the original variable reference component with a text input field, allowing free-form SQL writing with variable support.
-- Chat: Re-enabled Reasoning and Cross-language search.
-- Retrieval API Enhanced: Added metadata filtering support to the [Retrieve chunks](https://ragflow.io/docs/dev/http_api_reference#retrieve-chunks) method.
+- Agent:
+  - Agent Performance Optimized: Improves planning and reflection speed for simple tasks; optimizes concurrent tool calls for parallelizable scenarios, significantly reducing overall response time.
+  - Four framework-level prompt blocks are available in the **System prompt** section, enabling customization and overriding of prompts at the framework level, thereby enhancing flexibility and control. See [here](./guides/agent/agent_component_reference/agent.mdx#system-prompt).
+  - **Execute SQL** component enhanced: Replaces the original variable reference component with a text input field, allowing users to write free-form SQL queries and reference variables. See [here](./guides/agent/agent_component_reference/execute_sql.md).
+- Chat: Re-enables **Reasoning** and **Cross-language search**.
 
 ### Added models
 
@@ -44,8 +44,22 @@ Released on September 10, 2025.
 ### Fixed issues
 
 - Dataset: Deleted files remained searchable.
-- Chat: Unable to chat with an Ollama model. 
-- Agent: Resolved issues including cite toggle failure, task mode requiring dialogue triggers, repeated answers in multi-turn dialogues, and duplicate summarization of parallel execution results.
+- Chat: Unable to chat with an Ollama model.
+- Agent:
+  - A **Cite** toggle failure.
+  - An Agent in task mode still required a dialogue to trigger.
+  - Repeated answers in multi-turn dialogues.
+  - Duplicate summarization of parallel execution results.
+
+### API changes
+
+#### HTTP APIs
+
+- Adds a body parameter `"metadata_condition"` to the [Retrieve chunks](./references/http_api_reference.md#retrieve-chunks) method, enabling metadata-based chunk filtering during retrieval. [#9877](https://github.com/infiniflow/ragflow/pull/9877)
+
+#### Python APIs
+
+- Adds a parameter `metadata_condition` to the [Retrieve chunks](./references/python_api_reference.md#retrieve-chunks) method, enabling metadata-based chunk filtering during retrieval. [#9877](https://github.com/infiniflow/ragflow/pull/9877)
 
 ## v0.20.4
 
@@ -65,7 +79,7 @@ ZHIPU GLM-4.5
 
 ### New Agent templates
 
-Ecommerce Customer Service Workflow: A template designed to handle enquiries about product features and multi-product comparisons using the internal knowledge base, as well as to manage installation appointment bookings.
+Ecommerce Customer Service Workflow: A template designed to handle enquiries about product features and multi-product comparisons using the internal dataset, as well as to manage installation appointment bookings.
 
 ### Fixed issues
 
@@ -117,7 +131,7 @@ Released on August 8, 2025.
 
 ### New Features
 
-- The **Retrieval** component now supports the dynamic specification of knowledge base names using variables.
+- The **Retrieval** component now supports the dynamic specification of dataset names using variables.
 - The user interface now includes a French language option.
 
 ### Added Models
@@ -128,7 +142,7 @@ Released on August 8, 2025.
 ### New agent templates (both workflow and agentic)
 
 - SQL Assistant Workflow: Empowers non-technical teams (e.g., operations, product) to independently query business data.
-- Choose Your Knowledge Base Workflow: Lets users select a knowledge base to query during conversations. [#9325](https://github.com/infiniflow/ragflow/pull/9325)
+- Choose Your Knowledge Base Workflow: Lets users select a dataset to query during conversations. [#9325](https://github.com/infiniflow/ragflow/pull/9325)
 - Choose Your Knowledge Base Agent: Delivers higher-quality responses with extended reasoning time, suited for complex queries. [#9325](https://github.com/infiniflow/ragflow/pull/9325)
 
 ### Fixed Issues
@@ -161,14 +175,14 @@ From v0.20.0 onwards, Agents are no longer compatible with earlier versions, and
 ### New agent templates introduced
 
 - Multi-Agent based Deep Research: Collaborative Agent teamwork led by a Lead Agent with multiple Subagents, distinct from traditional workflow orchestration.
-- An intelligent Q&A chatbot leveraging internal knowledge bases, designed for customer service and training scenarios.
+- An intelligent Q&A chatbot leveraging internal datasets, designed for customer service and training scenarios.
 - A resume analysis template used by the RAGFlow team to screen, analyze, and record candidate information.
 - A blog generation workflow that transforms raw ideas into SEO-friendly blog content.
 - An intelligent customer service workflow.
 - A user feedback analysis template that directs user feedback to appropriate teams through semantic analysis.
 - Trip Planner: Uses web search and map MCP servers to assist with travel planning.
 - Image Lingo: Translates content from uploaded photos.
-- An information search assistant that retrieves answers from both internal knowledge bases and the web.
+- An information search assistant that retrieves answers from both internal datasets and the web.
 
 ## v0.19.1
 
@@ -181,7 +195,7 @@ Released on June 23, 2025.
 - A context error occurring when using Sandbox in standalone mode. [#8340](https://github.com/infiniflow/ragflow/pull/8340)
 - An excessive CPU usage issue caused by Ollama. [#8216](https://github.com/infiniflow/ragflow/pull/8216)
 - A bug in the Code Component. [#7949](https://github.com/infiniflow/ragflow/pull/7949)
-- Added support for models installed via Ollama or VLLM when creating a knowledge base through the API. [#8069](https://github.com/infiniflow/ragflow/pull/8069)
+- Added support for models installed via Ollama or VLLM when creating a dataset through the API. [#8069](https://github.com/infiniflow/ragflow/pull/8069)
 - Enabled role-based authentication for S3 bucket access. [#8149](https://github.com/infiniflow/ragflow/pull/8149)
 
 ### Added models
@@ -195,7 +209,7 @@ Released on May 26, 2025.
 
 ### New features
 
-- [Cross-language search](./references/glossary.mdx#cross-language-search) is supported in the Knowledge and Chat modules, enhancing search accuracy and user experience in multilingual environments, such as in Chinese-English knowledge bases.
+- [Cross-language search](./references/glossary.mdx#cross-language-search) is supported in the Knowledge and Chat modules, enhancing search accuracy and user experience in multilingual environments, such as in Chinese-English datasets.
 - Agent component: A new Code component supports Python and JavaScript scripts, enabling developers to handle more complex tasks like dynamic data processing.
 - Enhanced image display: Images in Chat and Search now render directly within responses, rather than as external references. Knowledge retrieval testing can retrieve images directly, instead of texts extracted from images.
 - Claude 4 and ChatGPT o3: Developers can now use the newly released, most advanced Claude model and OpenAI’s latest ChatGPT o3 inference model.
@@ -224,7 +238,7 @@ From this release onwards, built-in rerank models have been removed because they
 
 ### New features
 
-- MCP server: enables access to RAGFlow's knowledge bases via MCP.
+- MCP server: enables access to RAGFlow's datasets via MCP.
 - DeepDoc supports adopting VLM model as a processing pipeline during document layout recognition, enabling in-depth analysis of images in PDF and DOCX files.
 - OpenAI-compatible APIs: Agents can be called via OpenAI-compatible APIs.
 - User registration control: administrators can enable or disable user registration through an environment variable.
@@ -316,7 +330,7 @@ Released on March 3, 2025.
 
 - AI chat: Implements Deep Research for agentic reasoning. To activate this, enable the **Reasoning** toggle under the **Prompt engine** tab of your chat assistant dialogue.
 - AI chat: Leverages Tavily-based web search to enhance contexts in agentic reasoning. To activate this, enter the correct Tavily API key under the **Assistant settings** tab of your chat assistant dialogue.
-- AI chat: Supports starting a chat without specifying knowledge bases.
+- AI chat: Supports starting a chat without specifying datasets.
 - AI chat: HTML files can also be previewed and referenced, in addition to PDF files.
 - Dataset: Adds a **PDF parser**, aka **Document parser**, dropdown menu to dataset configurations. This includes a DeepDoc model option, which is time-consuming, a much faster **naive** option (plain text), which skips DLA (Document Layout Analysis), OCR (Optical Character Recognition), and TSR (Table Structure Recognition) tasks, and several currently *experimental* large model options. See [here](./guides/dataset/select_pdf_parser.md).
 - Agent component: **(x)** or a forward slash `/` can be used to insert available keys (variables) in the system prompt field of the **Generate** or **Template** component.
@@ -355,16 +369,16 @@ Released on February 6, 2025.
 ### New features
 
 - Supports DeepSeek R1 and DeepSeek V3.
-- GraphRAG refactor: Knowledge graph is dynamically built on an entire knowledge base (dataset) rather than on an individual file, and automatically updated when a newly uploaded file starts parsing. See [here](https://ragflow.io/docs/dev/construct_knowledge_graph).
+- GraphRAG refactor: Knowledge graph is dynamically built on an entire dataset rather than on an individual file, and automatically updated when a newly uploaded file starts parsing. See [here](https://ragflow.io/docs/dev/construct_knowledge_graph).
 - Adds an **Iteration** agent component and a **Research report generator** agent template. See [here](./guides/agent/agent_component_reference/iteration.mdx).
 - New UI language: Portuguese.
-- Allows setting metadata for a specific file in a knowledge base to enhance AI-powered chats. See [here](./guides/dataset/set_metadata.md).
+- Allows setting metadata for a specific file in a dataset to enhance AI-powered chats. See [here](./guides/dataset/set_metadata.md).
 - Upgrades RAGFlow's document engine [Infinity](https://github.com/infiniflow/infinity) to v0.6.0.dev3.
 - Supports GPU acceleration for DeepDoc (see [docker-compose-gpu.yml](https://github.com/infiniflow/ragflow/blob/main/docker/docker-compose-gpu.yml)).
-- Supports creating and referencing a **Tag** knowledge base as a key milestone towards bridging the semantic gap between query and response.
+- Supports creating and referencing a **Tag** dataset as a key milestone towards bridging the semantic gap between query and response.
 
 :::danger IMPORTANT
-The **Tag knowledge base** feature is *unavailable* on the [Infinity](https://github.com/infiniflow/infinity) document engine.
+The **Tag dataset** feature is *unavailable* on the [Infinity](https://github.com/infiniflow/infinity) document engine.
 :::
 
 ### Documentation
@@ -401,7 +415,7 @@ Released on December 25, 2024.
 This release fixes the following issues:
 
 - The `SCORE not found` and `position_int` errors returned by [Infinity](https://github.com/infiniflow/infinity).
-- Once an embedding model in a specific knowledge base is changed, embedding models in other knowledge bases can no longer be changed.
+- Once an embedding model in a specific dataset is changed, embedding models in other datasets can no longer be changed.
 - Slow response in question-answering and AI search due to repetitive loading of the embedding model.
 - Fails to parse documents with RAPTOR.
 - Using the **Table** parsing method results in information loss.
@@ -428,7 +442,7 @@ Released on December 18, 2024.
 ### New features
 
 - Introduces additional Agent-specific APIs.
-- Supports using page rank score to improve retrieval performance when searching across multiple knowledge bases.
+- Supports using page rank score to improve retrieval performance when searching across multiple datasets.
 - Offers an iframe in Chat and Agent to facilitate the integration of RAGFlow into your webpage.
 - Adds a Helm chart for deploying RAGFlow on Kubernetes.
 - Supports importing or exporting an agent in JSON format.

graphrag/general/extractor.py

@@ -37,7 +37,7 @@ from graphrag.utils import (
     split_string_by_multi_markers,
 )
 from rag.llm.chat_model import Base as CompletionLLM
-from rag.prompts import message_fit_in
+from rag.prompts.generator import message_fit_in
 from rag.utils import truncate
 
 GRAPH_FIELD_SEP = "<SEP>"