Docs: How to use MinerU to parse pdf documents (#10763 )

### What problem does this PR solve? ### Type of change - [x] Documentation Update
Docs: Update version references to v0.21.1 in READMEs and docs (#10761 )
2026-01-04 03:25:30 +08:00 · 2025-10-23 18:56:09 +08:00 · 2025-10-23 18:55:41 +08:00 · 2025-10-23 17:38:54 +08:00 · 2025-10-23 17:25:29 +08:00 · 2025-10-23 16:17:43 +08:00
229 changed files with 9303 additions and 3081 deletions
--- a/README.md
+++ b/README.md
@ -22,7 +22,7 @@
        <img alt="Static Badge" src="https://img.shields.io/badge/Online-Demo-4e6b99">
    </a>
    <a href="https://hub.docker.com/r/infiniflow/ragflow" target="_blank">
-        <img src="https://img.shields.io/docker/pulls/infiniflow/ragflow?label=Docker%20Pulls&color=0db7ed&logo=docker&logoColor=white&style=flat-square" alt="docker pull infiniflow/ragflow:v0.21.0">
+        <img src="https://img.shields.io/docker/pulls/infiniflow/ragflow?label=Docker%20Pulls&color=0db7ed&logo=docker&logoColor=white&style=flat-square" alt="docker pull infiniflow/ragflow:v0.21.1">
    </a>
    <a href="https://github.com/infiniflow/ragflow/releases/latest">
        <img src="https://img.shields.io/github/v/release/infiniflow/ragflow?color=blue&label=Latest%20Release" alt="Latest Release">
@ -187,7 +187,7 @@ releases! 🌟
 > All Docker images are built for x86 platforms. We don't currently offer Docker images for ARM64.
 > If you are on an ARM64 platform, follow [this guide](https://ragflow.io/docs/dev/build_docker_image) to build a Docker image compatible with your system.

-   > The command below downloads the `v0.21.0-slim` edition of the RAGFlow Docker image. See the following table for descriptions of different RAGFlow editions. To download a RAGFlow edition different from `v0.21.0-slim`, update the `RAGFLOW_IMAGE` variable accordingly in **docker/.env** before using `docker compose` to start the server. For example: set `RAGFLOW_IMAGE=infiniflow/ragflow:v0.21.0` for the full edition `v0.21.0`.
+   > The command below downloads the `v0.21.1-slim` edition of the RAGFlow Docker image. See the following table for descriptions of different RAGFlow editions. To download a RAGFlow edition different from `v0.21.1-slim`, update the `RAGFLOW_IMAGE` variable accordingly in **docker/.env** before using `docker compose` to start the server. For example: set `RAGFLOW_IMAGE=infiniflow/ragflow:v0.21.1` for the full edition `v0.21.1`.

   ```bash
   $ cd ragflow/docker
@ -200,8 +200,8 @@ releases! 🌟

   | RAGFlow image tag | Image size (GB) | Has embedding models? | Stable?                  |
   |-------------------|-----------------|-----------------------|--------------------------|
-   | v0.21.0           | &approx;9       | :heavy_check_mark:    | Stable release           |
-   | v0.21.0-slim      | &approx;2       | ❌                   | Stable release            |
+   | v0.21.1           | &approx;9       | :heavy_check_mark:    | Stable release           |
+   | v0.21.1-slim      | &approx;2       | ❌                   | Stable release            |
   | nightly           | &approx;9       | :heavy_check_mark:    | _Unstable_ nightly build |
   | nightly-slim      | &approx;2       | ❌                   | _Unstable_ nightly build  |

--- a/README_id.md
+++ b/README_id.md
@ -22,7 +22,7 @@
        <img alt="Lencana Daring" src="https://img.shields.io/badge/Online-Demo-4e6b99">
    </a>
    <a href="https://hub.docker.com/r/infiniflow/ragflow" target="_blank">
-        <img src="https://img.shields.io/docker/pulls/infiniflow/ragflow?label=Docker%20Pulls&color=0db7ed&logo=docker&logoColor=white&style=flat-square" alt="docker pull infiniflow/ragflow:v0.21.0">
+        <img src="https://img.shields.io/docker/pulls/infiniflow/ragflow?label=Docker%20Pulls&color=0db7ed&logo=docker&logoColor=white&style=flat-square" alt="docker pull infiniflow/ragflow:v0.21.1">
    </a>
    <a href="https://github.com/infiniflow/ragflow/releases/latest">
        <img src="https://img.shields.io/github/v/release/infiniflow/ragflow?color=blue&label=Rilis%20Terbaru" alt="Rilis Terbaru">
@ -181,7 +181,7 @@ Coba demo kami di [https://demo.ragflow.io](https://demo.ragflow.io).
 > Semua gambar Docker dibangun untuk platform x86. Saat ini, kami tidak menawarkan gambar Docker untuk ARM64.
 > Jika Anda menggunakan platform ARM64, [silakan gunakan panduan ini untuk membangun gambar Docker yang kompatibel dengan sistem Anda](https://ragflow.io/docs/dev/build_docker_image).

-> Perintah di bawah ini mengunduh edisi v0.21.0-slim dari gambar Docker RAGFlow. Silakan merujuk ke tabel berikut untuk deskripsi berbagai edisi RAGFlow. Untuk mengunduh edisi RAGFlow yang berbeda dari v0.21.0-slim, perbarui variabel RAGFLOW_IMAGE di docker/.env sebelum menggunakan docker compose untuk memulai server. Misalnya, atur RAGFLOW_IMAGE=infiniflow/ragflow:v0.21.0 untuk edisi lengkap v0.21.0.
+> Perintah di bawah ini mengunduh edisi v0.21.1-slim dari gambar Docker RAGFlow. Silakan merujuk ke tabel berikut untuk deskripsi berbagai edisi RAGFlow. Untuk mengunduh edisi RAGFlow yang berbeda dari v0.21.1-slim, perbarui variabel RAGFLOW_IMAGE di docker/.env sebelum menggunakan docker compose untuk memulai server. Misalnya, atur RAGFLOW_IMAGE=infiniflow/ragflow:v0.21.1 untuk edisi lengkap v0.21.1.

 ```bash
 $ cd ragflow/docker
@ -194,8 +194,8 @@ $ docker compose -f docker-compose.yml up -d

 | RAGFlow image tag | Image size (GB) | Has embedding models? | Stable?                  |
 | ----------------- | --------------- | --------------------- | ------------------------ |
-| v0.21.0           | &approx;9       | :heavy_check_mark:    | Stable release           |
-| v0.21.0-slim      | &approx;2       | ❌                    | Stable release           |
+| v0.21.1           | &approx;9       | :heavy_check_mark:    | Stable release           |
+| v0.21.1-slim      | &approx;2       | ❌                    | Stable release           |
 | nightly           | &approx;9       | :heavy_check_mark:    | _Unstable_ nightly build |
 | nightly-slim      | &approx;2       | ❌                    | _Unstable_ nightly build |

--- a/README_ja.md
+++ b/README_ja.md
@ -22,7 +22,7 @@
        <img alt="Static Badge" src="https://img.shields.io/badge/Online-Demo-4e6b99">
    </a>
    <a href="https://hub.docker.com/r/infiniflow/ragflow" target="_blank">
-        <img src="https://img.shields.io/docker/pulls/infiniflow/ragflow?label=Docker%20Pulls&color=0db7ed&logo=docker&logoColor=white&style=flat-square" alt="docker pull infiniflow/ragflow:v0.21.0">
+        <img src="https://img.shields.io/docker/pulls/infiniflow/ragflow?label=Docker%20Pulls&color=0db7ed&logo=docker&logoColor=white&style=flat-square" alt="docker pull infiniflow/ragflow:v0.21.1">
    </a>
    <a href="https://github.com/infiniflow/ragflow/releases/latest">
        <img src="https://img.shields.io/github/v/release/infiniflow/ragflow?color=blue&label=Latest%20Release" alt="Latest Release">
@ -160,7 +160,7 @@
 > 現在、公式に提供されているすべての Docker イメージは x86 アーキテクチャ向けにビルドされており、ARM64 用の Docker イメージは提供されていません。
 > ARM64 アーキテクチャのオペレーティングシステムを使用している場合は、[このドキュメント](https://ragflow.io/docs/dev/build_docker_image)を参照して Docker イメージを自分でビルドしてください。

-   > 以下のコマンドは、RAGFlow Docker イメージの v0.21.0-slim エディションをダウンロードします。異なる RAGFlow エディションの説明については、以下の表を参照してください。v0.21.0-slim とは異なるエディションをダウンロードするには、docker/.env ファイルの RAGFLOW_IMAGE 変数を適宜更新し、docker compose を使用してサーバーを起動してください。例えば、完全版 v0.21.0 をダウンロードするには、RAGFLOW_IMAGE=infiniflow/ragflow:v0.21.0 と設定します。
+   > 以下のコマンドは、RAGFlow Docker イメージの v0.21.1-slim エディションをダウンロードします。異なる RAGFlow エディションの説明については、以下の表を参照してください。v0.21.1-slim とは異なるエディションをダウンロードするには、docker/.env ファイルの RAGFLOW_IMAGE 変数を適宜更新し、docker compose を使用してサーバーを起動してください。例えば、完全版 v0.21.1 をダウンロードするには、RAGFLOW_IMAGE=infiniflow/ragflow:v0.21.1 と設定します。

   ```bash
   $ cd ragflow/docker
@ -173,8 +173,8 @@

   | RAGFlow image tag | Image size (GB) | Has embedding models? | Stable?                  |
   | ----------------- | --------------- | --------------------- | ------------------------ |
-   | v0.21.0           | &approx;9       | :heavy_check_mark:    | Stable release           |
-   | v0.21.0-slim      | &approx;2       | ❌                    | Stable release           |
+   | v0.21.1           | &approx;9       | :heavy_check_mark:    | Stable release           |
+   | v0.21.1-slim      | &approx;2       | ❌                    | Stable release           |
   | nightly           | &approx;9       | :heavy_check_mark:    | _Unstable_ nightly build |
   | nightly-slim      | &approx;2       | ❌                     | _Unstable_ nightly build |

--- a/README_ko.md
+++ b/README_ko.md
@ -22,7 +22,7 @@
        <img alt="Static Badge" src="https://img.shields.io/badge/Online-Demo-4e6b99">
    </a>
    <a href="https://hub.docker.com/r/infiniflow/ragflow" target="_blank">
-        <img src="https://img.shields.io/docker/pulls/infiniflow/ragflow?label=Docker%20Pulls&color=0db7ed&logo=docker&logoColor=white&style=flat-square" alt="docker pull infiniflow/ragflow:v0.21.0">
+        <img src="https://img.shields.io/docker/pulls/infiniflow/ragflow?label=Docker%20Pulls&color=0db7ed&logo=docker&logoColor=white&style=flat-square" alt="docker pull infiniflow/ragflow:v0.21.1">
    </a>
    <a href="https://github.com/infiniflow/ragflow/releases/latest">
        <img src="https://img.shields.io/github/v/release/infiniflow/ragflow?color=blue&label=Latest%20Release" alt="Latest Release">
@ -160,7 +160,7 @@
 > 모든 Docker 이미지는 x86 플랫폼을 위해 빌드되었습니다. 우리는 현재 ARM64 플랫폼을 위한 Docker 이미지를 제공하지 않습니다.
 > ARM64 플랫폼을 사용 중이라면, [시스템과 호환되는 Docker 이미지를 빌드하려면 이 가이드를 사용해 주세요](https://ragflow.io/docs/dev/build_docker_image).

-   > 아래 명령어는 RAGFlow Docker 이미지의 v0.21.0-slim 버전을 다운로드합니다. 다양한 RAGFlow 버전에 대한 설명은 다음 표를 참조하십시오. v0.21.0-slim과 다른 RAGFlow 버전을 다운로드하려면, docker/.env 파일에서 RAGFLOW_IMAGE 변수를 적절히 업데이트한 후 docker compose를 사용하여 서버를 시작하십시오. 예를 들어, 전체 버전인 v0.21.0을 다운로드하려면 RAGFLOW_IMAGE=infiniflow/ragflow:v0.21.0로 설정합니다.
+   > 아래 명령어는 RAGFlow Docker 이미지의 v0.21.1-slim 버전을 다운로드합니다. 다양한 RAGFlow 버전에 대한 설명은 다음 표를 참조하십시오. v0.21.1-slim과 다른 RAGFlow 버전을 다운로드하려면, docker/.env 파일에서 RAGFLOW_IMAGE 변수를 적절히 업데이트한 후 docker compose를 사용하여 서버를 시작하십시오. 예를 들어, 전체 버전인 v0.21.1을 다운로드하려면 RAGFLOW_IMAGE=infiniflow/ragflow:v0.21.1로 설정합니다.

   ```bash
   $ cd ragflow/docker
@ -173,8 +173,8 @@

   | RAGFlow image tag | Image size (GB) | Has embedding models? | Stable?                  |
   | ----------------- | --------------- | --------------------- | ------------------------ |
-   | v0.21.0           | &approx;9       | :heavy_check_mark:    | Stable release           |
-   | v0.21.0-slim      | &approx;2       | ❌                    | Stable release           |
+   | v0.21.1           | &approx;9       | :heavy_check_mark:    | Stable release           |
+   | v0.21.1-slim      | &approx;2       | ❌                    | Stable release           |
   | nightly           | &approx;9       | :heavy_check_mark:    | _Unstable_ nightly build |
   | nightly-slim      | &approx;2       | ❌                     | _Unstable_ nightly build |

--- a/README_pt_br.md
+++ b/README_pt_br.md
@ -22,7 +22,7 @@
        <img alt="Badge Estático" src="https://img.shields.io/badge/Online-Demo-4e6b99">
    </a>
    <a href="https://hub.docker.com/r/infiniflow/ragflow" target="_blank">
-        <img src="https://img.shields.io/docker/pulls/infiniflow/ragflow?label=Docker%20Pulls&color=0db7ed&logo=docker&logoColor=white&style=flat-square" alt="docker pull infiniflow/ragflow:v0.21.0">
+        <img src="https://img.shields.io/docker/pulls/infiniflow/ragflow?label=Docker%20Pulls&color=0db7ed&logo=docker&logoColor=white&style=flat-square" alt="docker pull infiniflow/ragflow:v0.21.1">
    </a>
    <a href="https://github.com/infiniflow/ragflow/releases/latest">
        <img src="https://img.shields.io/github/v/release/infiniflow/ragflow?color=blue&label=Última%20Relese" alt="Última Versão">
@ -180,7 +180,7 @@ Experimente nossa demo em [https://demo.ragflow.io](https://demo.ragflow.io).
 > Todas as imagens Docker são construídas para plataformas x86. Atualmente, não oferecemos imagens Docker para ARM64.
 > Se você estiver usando uma plataforma ARM64, por favor, utilize [este guia](https://ragflow.io/docs/dev/build_docker_image) para construir uma imagem Docker compatível com o seu sistema.

-    > O comando abaixo baixa a edição `v0.21.0-slim` da imagem Docker do RAGFlow. Consulte a tabela a seguir para descrições de diferentes edições do RAGFlow. Para baixar uma edição do RAGFlow diferente da `v0.21.0-slim`, atualize a variável `RAGFLOW_IMAGE` conforme necessário no **docker/.env** antes de usar `docker compose` para iniciar o servidor. Por exemplo: defina `RAGFLOW_IMAGE=infiniflow/ragflow:v0.21.0` para a edição completa `v0.21.0`.
+    > O comando abaixo baixa a edição `v0.21.1-slim` da imagem Docker do RAGFlow. Consulte a tabela a seguir para descrições de diferentes edições do RAGFlow. Para baixar uma edição do RAGFlow diferente da `v0.21.1-slim`, atualize a variável `RAGFLOW_IMAGE` conforme necessário no **docker/.env** antes de usar `docker compose` para iniciar o servidor. Por exemplo: defina `RAGFLOW_IMAGE=infiniflow/ragflow:v0.21.1` para a edição completa `v0.21.1`.

    ```bash
    $ cd ragflow/docker
@ -193,8 +193,8 @@ Experimente nossa demo em [https://demo.ragflow.io](https://demo.ragflow.io).

    | Tag da imagem RAGFlow | Tamanho da imagem (GB) | Possui modelos de incorporação? | Estável?                 |
    | --------------------- | ---------------------- | ------------------------------- | ------------------------ |
-    | v0.21.0               | ~9                     | :heavy_check_mark:              | Lançamento estável       |
-    | v0.21.0-slim          | ~2                     | ❌                              | Lançamento estável       |
+    | v0.21.1               | ~9                     | :heavy_check_mark:              | Lançamento estável       |
+    | v0.21.1-slim          | ~2                     | ❌                              | Lançamento estável       |
    | nightly               | ~9                     | :heavy_check_mark:              | _Instável_ build noturno |
    | nightly-slim          | ~2                     | ❌                               | _Instável_ build noturno |

--- a/README_tzh.md
+++ b/README_tzh.md
@ -22,7 +22,7 @@
        <img alt="Static Badge" src="https://img.shields.io/badge/Online-Demo-4e6b99">
    </a>
    <a href="https://hub.docker.com/r/infiniflow/ragflow" target="_blank">
-        <img src="https://img.shields.io/docker/pulls/infiniflow/ragflow?label=Docker%20Pulls&color=0db7ed&logo=docker&logoColor=white&style=flat-square" alt="docker pull infiniflow/ragflow:v0.21.0">
+        <img src="https://img.shields.io/docker/pulls/infiniflow/ragflow?label=Docker%20Pulls&color=0db7ed&logo=docker&logoColor=white&style=flat-square" alt="docker pull infiniflow/ragflow:v0.21.1">
    </a>
    <a href="https://github.com/infiniflow/ragflow/releases/latest">
        <img src="https://img.shields.io/github/v/release/infiniflow/ragflow?color=blue&label=Latest%20Release" alt="Latest Release">
@ -183,7 +183,7 @@
 > 所有 Docker 映像檔都是為 x86 平台建置的。目前，我們不提供 ARM64 平台的 Docker 映像檔。
 > 如果您使用的是 ARM64 平台，請使用 [這份指南](https://ragflow.io/docs/dev/build_docker_image) 來建置適合您系統的 Docker 映像檔。

-   > 執行以下指令會自動下載 RAGFlow slim Docker 映像 `v0.21.0-slim`。請參考下表查看不同 Docker 發行版的說明。如需下載不同於 `v0.21.0-slim` 的 Docker 映像，請在執行 `docker compose` 啟動服務之前先更新 **docker/.env** 檔案內的 `RAGFLOW_IMAGE` 變數。例如，你可以透過設定 `RAGFLOW_IMAGE=infiniflow/ragflow:v0.21.0` 來下載 RAGFlow 鏡像的 `v0.21.0` 完整發行版。
+   > 執行以下指令會自動下載 RAGFlow slim Docker 映像 `v0.21.1-slim`。請參考下表查看不同 Docker 發行版的說明。如需下載不同於 `v0.21.1-slim` 的 Docker 映像，請在執行 `docker compose` 啟動服務之前先更新 **docker/.env** 檔案內的 `RAGFLOW_IMAGE` 變數。例如，你可以透過設定 `RAGFLOW_IMAGE=infiniflow/ragflow:v0.21.1` 來下載 RAGFlow 鏡像的 `v0.21.1` 完整發行版。

   ```bash
   $ cd ragflow/docker
@ -196,8 +196,8 @@

   | RAGFlow image tag | Image size (GB) | Has embedding models? | Stable?                  |
   | ----------------- | --------------- | --------------------- | ------------------------ |
-   | v0.21.0           | &approx;9       | :heavy_check_mark:    | Stable release           |
-   | v0.21.0-slim      | &approx;2       | ❌                    | Stable release           |
+   | v0.21.1           | &approx;9       | :heavy_check_mark:    | Stable release           |
+   | v0.21.1-slim      | &approx;2       | ❌                    | Stable release           |
   | nightly           | &approx;9       | :heavy_check_mark:    | _Unstable_ nightly build |
   | nightly-slim      | &approx;2       | ❌                     | _Unstable_ nightly build |

--- a/README_zh.md
+++ b/README_zh.md
@ -22,7 +22,7 @@
        <img alt="Static Badge" src="https://img.shields.io/badge/Online-Demo-4e6b99">
    </a>
    <a href="https://hub.docker.com/r/infiniflow/ragflow" target="_blank">
-        <img src="https://img.shields.io/docker/pulls/infiniflow/ragflow?label=Docker%20Pulls&color=0db7ed&logo=docker&logoColor=white&style=flat-square" alt="docker pull infiniflow/ragflow:v0.21.0">
+        <img src="https://img.shields.io/docker/pulls/infiniflow/ragflow?label=Docker%20Pulls&color=0db7ed&logo=docker&logoColor=white&style=flat-square" alt="docker pull infiniflow/ragflow:v0.21.1">
    </a>
    <a href="https://github.com/infiniflow/ragflow/releases/latest">
        <img src="https://img.shields.io/github/v/release/infiniflow/ragflow?color=blue&label=Latest%20Release" alt="Latest Release">
@ -183,7 +183,7 @@
 > 请注意，目前官方提供的所有 Docker 镜像均基于 x86 架构构建，并不提供基于 ARM64 的 Docker 镜像。
 > 如果你的操作系统是 ARM64 架构，请参考[这篇文档](https://ragflow.io/docs/dev/build_docker_image)自行构建 Docker 镜像。

-   > 运行以下命令会自动下载 RAGFlow slim Docker 镜像 `v0.21.0-slim`。请参考下表查看不同 Docker 发行版的描述。如需下载不同于 `v0.21.0-slim` 的 Docker 镜像，请在运行 `docker compose` 启动服务之前先更新 **docker/.env** 文件内的 `RAGFLOW_IMAGE` 变量。比如，你可以通过设置 `RAGFLOW_IMAGE=infiniflow/ragflow:v0.21.0` 来下载 RAGFlow 镜像的 `v0.21.0` 完整发行版。
+   > 运行以下命令会自动下载 RAGFlow slim Docker 镜像 `v0.21.1-slim`。请参考下表查看不同 Docker 发行版的描述。如需下载不同于 `v0.21.1-slim` 的 Docker 镜像，请在运行 `docker compose` 启动服务之前先更新 **docker/.env** 文件内的 `RAGFLOW_IMAGE` 变量。比如，你可以通过设置 `RAGFLOW_IMAGE=infiniflow/ragflow:v0.21.1` 来下载 RAGFlow 镜像的 `v0.21.1` 完整发行版。

   ```bash
   $ cd ragflow/docker
@ -196,8 +196,8 @@

   | RAGFlow image tag | Image size (GB) | Has embedding models? | Stable?                  |
   | ----------------- | --------------- | --------------------- | ------------------------ |
-   | v0.21.0           | &approx;9       | :heavy_check_mark:    | Stable release           |
-   | v0.21.0-slim      | &approx;2       | ❌                    | Stable release           |
+   | v0.21.1           | &approx;9       | :heavy_check_mark:    | Stable release           |
+   | v0.21.1-slim      | &approx;2       | ❌                    | Stable release           |
   | nightly           | &approx;9       | :heavy_check_mark:    | _Unstable_ nightly build |
   | nightly-slim      | &approx;2       | ❌                     | _Unstable_ nightly build |

--- a/admin/client/README.md
+++ b/admin/client/README.md
@ -48,7 +48,7 @@ It consists of a server-side Service and a command-line client (CLI), both imple
 1.  Ensure the Admin Service is running.
 2.  Install ragflow-cli.
    ```bash
-    pip install ragflow-cli==0.21.0
+    pip install ragflow-cli==0.21.1
    ```
 3.  Launch the CLI client:
    ```bash
--- a/admin/client/admin_client.py
+++ b/admin/client/admin_client.py
@ -21,9 +21,8 @@ from cmd import Cmd
 from Cryptodome.PublicKey import RSA
 from Cryptodome.Cipher import PKCS1_v1_5 as Cipher_pkcs1_v1_5
 from typing import Dict, List, Any
-from lark import Lark, Transformer, Tree, Token
+from lark import Lark, Transformer, Tree
 import requests
-from requests.auth import HTTPBasicAuth

 GRAMMAR = r"""
 start: command
@ -43,6 +42,15 @@ sql_command: list_services
           | activate_user
           | list_datasets
           | list_agents
+           | create_role
+           | drop_role
+           | alter_role
+           | list_roles
+           | show_role
+           | grant_permission
+           | revoke_permission
+           | alter_user_role
+           | show_user_permission

 // meta command definition
 meta_command: "\\" meta_command_name [meta_args]
@ -71,6 +79,19 @@ PASSWORD: "PASSWORD"i
 DATASETS: "DATASETS"i
 OF: "OF"i
 AGENTS: "AGENTS"i
+ROLE: "ROLE"i
+ROLES: "ROLES"i
+DESCRIPTION: "DESCRIPTION"i
+GRANT: "GRANT"i
+REVOKE: "REVOKE"i
+ALL: "ALL"i
+PERMISSION: "PERMISSION"i
+TO: "TO"i
+FROM: "FROM"i
+FOR: "FOR"i
+RESOURCES: "RESOURCES"i
+ON: "ON"i
+SET: "SET"i

 list_services: LIST SERVICES ";"
 show_service: SHOW SERVICE NUMBER ";"
@ -88,6 +109,19 @@ activate_user: ALTER USER ACTIVE quoted_string status ";"
 list_datasets: LIST DATASETS OF quoted_string ";"
 list_agents: LIST AGENTS OF quoted_string ";"

+create_role: CREATE ROLE identifier [DESCRIPTION quoted_string] ";"
+drop_role: DROP ROLE identifier ";"
+alter_role: ALTER ROLE identifier SET DESCRIPTION quoted_string ";"
+list_roles: LIST ROLES ";"
+show_role: SHOW ROLE identifier ";"
+
+grant_permission: GRANT action_list ON identifier TO ROLE identifier ";"
+revoke_permission: REVOKE action_list ON identifier FROM ROLE identifier ";"
+alter_user_role: ALTER USER quoted_string SET ROLE identifier ";"
+show_user_permission: SHOW USER PERMISSION quoted_string ";"
+
+action_list: identifier ("," identifier)*
+
 identifier: WORD
 quoted_string: QUOTED_STRING
 status: WORD
@ -134,34 +168,86 @@ class AdminTransformer(Transformer):

    def show_user(self, items):
        user_name = items[2]
-        return {"type": "show_user", "username": user_name}
+        return {"type": "show_user", "user_name": user_name}

    def drop_user(self, items):
        user_name = items[2]
-        return {"type": "drop_user", "username": user_name}
+        return {"type": "drop_user", "user_name": 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}
+        return {"type": "alter_user", "user_name": user_name, "password": new_password}

    def create_user(self, items):
        user_name = items[2]
        password = items[3]
-        return {"type": "create_user", "username": user_name, "password": password, "role": "user"}
+        return {"type": "create_user", "user_name": user_name, "password": password, "role": "user"}

    def activate_user(self, items):
        user_name = items[3]
        activate_status = items[4]
-        return {"type": "activate_user", "activate_status": activate_status, "username": user_name}
+        return {"type": "activate_user", "activate_status": activate_status, "user_name": user_name}

    def list_datasets(self, items):
        user_name = items[3]
-        return {"type": "list_datasets", "username": user_name}
+        return {"type": "list_datasets", "user_name": user_name}

    def list_agents(self, items):
        user_name = items[3]
-        return {"type": "list_agents", "username": user_name}
+        return {"type": "list_agents", "user_name": user_name}
+
+    def create_role(self, items):
+        role_name = items[2]
+        if len(items) > 4:
+            description = items[4]
+            return {"type": "create_role", "role_name": role_name, "description": description}
+        else:
+            return {"type": "create_role", "role_name": role_name}
+
+    def drop_role(self, items):
+        role_name = items[2]
+        return {"type": "drop_role", "role_name": role_name}
+
+    def alter_role(self, items):
+        role_name = items[2]
+        description = items[5]
+        return {"type": "alter_role", "role_name": role_name, "description": description}
+
+    def list_roles(self, items):
+        return {"type": "list_roles"}
+
+    def show_role(self, items):
+        role_name = items[2]
+        return {"type": "show_role", "role_name": role_name}
+
+    def grant_permission(self, items):
+        action_list = items[1]
+        resource = items[3]
+        role_name = items[6]
+        return {"type": "grant_permission", "role_name": role_name, "resource": resource, "actions": action_list}
+
+    def revoke_permission(self, items):
+        action_list = items[1]
+        resource = items[3]
+        role_name = items[6]
+        return {
+            "type": "revoke_permission",
+            "role_name": role_name,
+            "resource": resource, "actions": action_list
+        }
+
+    def alter_user_role(self, items):
+        user_name = items[2]
+        role_name = items[5]
+        return {"type": "alter_user_role", "user_name": user_name, "role_name": role_name}
+
+    def show_user_permission(self, items):
+        user_name = items[3]
+        return {"type": "show_user_permission", "user_name": user_name}
+
+    def action_list(self, items):
+        return items

    def meta_command(self, items):
        command_name = str(items[0]).lower()
@ -205,6 +291,8 @@ class AdminCLI(Cmd):
        self.is_interactive = False
        self.admin_account = "admin@ragflow.io"
        self.admin_password: str = "admin"
+        self.session = requests.Session()
+        self.access_token: str = ""
        self.host: str = ""
        self.port: int = 0

@ -240,7 +328,7 @@ class AdminCLI(Cmd):
    def default(self, line: str) -> bool:
        return self.onecmd(line)

-    def parse_command(self, command_str: str) -> dict[str, str] | Tree[Token]:
+    def parse_command(self, command_str: str) -> dict[str, str]:
        if not command_str.strip():
            return {'type': 'empty'}

@ -252,32 +340,38 @@ class AdminCLI(Cmd):
        except Exception as e:
            return {'type': 'error', 'message': f'Parse error: {str(e)}'}

-    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']
+    def verify_admin(self, arguments: dict, single_command: bool):
+        self.host = arguments['host']
+        self.port = arguments['port']
        print(f"Attempt to access ip: {self.host}, port: {self.port}")
-        url = f'http://{self.host}:{self.port}/api/v1/admin/auth'
+        url = f"http://{self.host}:{self.port}/api/v1/admin/login"
+
+        attempt_count = 3
+        if single_command:
+            attempt_count = 1

        try_count = 0
        while True:
            try_count += 1
-            if try_count > 3:
+            if try_count > attempt_count:
                return False

-            admin_passwd = input(f"password for {self.admin_account}: ").strip()
+            if single_command:
+                admin_passwd = arguments['password']
+            else:
+                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))
+                self.admin_password = encrypt(admin_passwd)
+                response = self.session.post(url, json={'email': self.admin_account, 'password': self.admin_password})
                if response.status_code == 200:
                    res_json = response.json()
                    error_code = res_json.get('code', -1)
                    if error_code == 0:
+                        self.session.headers.update({
+                            'Content-Type': 'application/json',
+                            'Authorization': response.headers['Authorization'],
+                            'User-Agent': 'RAGFlow-CLI/0.21.1'
+                        })
                        print("Authentication successful.")
                        return True
                    else:
@ -285,8 +379,9 @@ class AdminCLI(Cmd):
                        print(f"Authentication failed: {error_message}, try again")
                        continue
                else:
-                    print(f"Bad response，status: {response.status_code}, try again")
-            except Exception:
+                    print(f"Bad response，status: {response.status_code}, password is wrong")
+            except Exception as e:
+                print(str(e))
                print(f"Can't access {self.host}, port: {self.port}")

    def _print_table_simple(self, data):
@ -371,23 +466,31 @@ class AdminCLI(Cmd):
                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 run_single_command(self, command: str):
+        result = self.parse_command(command)
+        self.execute_command(result)

-    def _parse_connection_args(self, args: List[str]) -> Dict[str, Any]:
+    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')
-
+        parser.add_argument('-w', '--password', default='admin', type=str, help='Superuser password')
+        parser.add_argument('command', nargs='?', help='Single command')
        try:
            parsed_args, remaining_args = parser.parse_known_args(args)
-            return {
-                'host': parsed_args.host,
-                'port': parsed_args.port,
-            }
+            if remaining_args:
+                command = remaining_args[0]
+                return {
+                    'host': parsed_args.host,
+                    'port': parsed_args.port,
+                    'password': parsed_args.password,
+                    'command': command
+                }
+            else:
+                return {
+                    'host': parsed_args.host,
+                    'port': parsed_args.port,
+                }
        except SystemExit:
            return {'error': 'Invalid connection arguments'}

@ -434,6 +537,24 @@ class AdminCLI(Cmd):
                self._handle_list_datasets(command_dict)
            case 'list_agents':
                self._handle_list_agents(command_dict)
+            case 'create_role':
+                self._create_role(command_dict)
+            case 'drop_role':
+                self._drop_role(command_dict)
+            case 'alter_role':
+                self._alter_role(command_dict)
+            case 'list_roles':
+                self._list_roles(command_dict)
+            case 'show_role':
+                self._show_role(command_dict)
+            case 'grant_permission':
+                self._grant_permission(command_dict)
+            case 'revoke_permission':
+                self._revoke_permission(command_dict)
+            case 'alter_user_role':
+                self._alter_user_role(command_dict)
+            case 'show_user_permission':
+                self._show_user_permission(command_dict)
            case 'meta':
                self._handle_meta_command(command_dict)
            case _:
@ -443,7 +564,7 @@ class AdminCLI(Cmd):
        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))
+        response = self.session.get(url)
        res_json = response.json()
        if response.status_code == 200:
            self._print_table_simple(res_json['data'])
@ -455,7 +576,7 @@ class AdminCLI(Cmd):
        print(f"Showing service: {service_id}")

        url = f'http://{self.host}:{self.port}/api/v1/admin/services/{service_id}'
-        response = requests.get(url, auth=HTTPBasicAuth(self.admin_account, self.admin_password))
+        response = self.session.get(url)
        res_json = response.json()
        if response.status_code == 200:
            res_data = res_json['data']
@ -486,7 +607,7 @@ class AdminCLI(Cmd):
        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))
+        response = self.session.get(url)
        res_json = response.json()
        if response.status_code == 200:
            self._print_table_simple(res_json['data'])
@ -494,23 +615,23 @@ class AdminCLI(Cmd):
            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}")
-        url = f'http://{self.host}:{self.port}/api/v1/admin/users/{username}'
-        response = requests.get(url, auth=HTTPBasicAuth(self.admin_account, self.admin_password))
+        username_tree: Tree = command['user_name']
+        user_name: str = username_tree.children[0].strip("'\"")
+        print(f"Showing user: {user_name}")
+        url = f'http://{self.host}:{self.port}/api/v1/admin/users/{user_name}'
+        response = self.session.get(url)
        res_json = response.json()
        if response.status_code == 200:
            self._print_table_simple(res_json['data'])
        else:
-            print(f"Fail to get user {username}, code: {res_json['code']}, message: {res_json['message']}")
+            print(f"Fail to get user {user_name}, code: {res_json['code']}, message: {res_json['message']}")

    def _handle_drop_user(self, command):
-        username_tree: Tree = command['username']
-        username: str = username_tree.children[0].strip("'\"")
-        print(f"Drop user: {username}")
-        url = f'http://{self.host}:{self.port}/api/v1/admin/users/{username}'
-        response = requests.delete(url, auth=HTTPBasicAuth(self.admin_account, self.admin_password))
+        username_tree: Tree = command['user_name']
+        user_name: str = username_tree.children[0].strip("'\"")
+        print(f"Drop user: {user_name}")
+        url = f'http://{self.host}:{self.port}/api/v1/admin/users/{user_name}'
+        response = self.session.delete(url)
        res_json = response.json()
        if response.status_code == 200:
            print(res_json["message"])
@ -518,14 +639,13 @@ class AdminCLI(Cmd):
            print(f"Fail to drop user, code: {res_json['code']}, message: {res_json['message']}")

    def _handle_alter_user(self, command):
-        username_tree: Tree = command['username']
-        username: str = username_tree.children[0].strip("'\"")
+        user_name_tree: Tree = command['user_name']
+        user_name: str = user_name_tree.children[0].strip("'\"")
        password_tree: Tree = command['password']
        password: str = password_tree.children[0].strip("'\"")
-        print(f"Alter user: {username}, password: {password}")
-        url = f'http://{self.host}:{self.port}/api/v1/admin/users/{username}/password'
-        response = requests.put(url, auth=HTTPBasicAuth(self.admin_account, self.admin_password),
-                                json={'new_password': encrypt(password)})
+        print(f"Alter user: {user_name}, password: {password}")
+        url = f'http://{self.host}:{self.port}/api/v1/admin/users/{user_name}/password'
+        response = self.session.put(url, json={'new_password': encrypt(password)})
        res_json = response.json()
        if response.status_code == 200:
            print(res_json["message"])
@ -533,34 +653,32 @@ class AdminCLI(Cmd):
            print(f"Fail to alter password, code: {res_json['code']}, message: {res_json['message']}")

    def _handle_create_user(self, command):
-        username_tree: Tree = command['username']
-        username: str = username_tree.children[0].strip("'\"")
+        user_name_tree: Tree = command['user_name']
+        user_name: str = user_name_tree.children[0].strip("'\"")
        password_tree: Tree = command['password']
        password: str = password_tree.children[0].strip("'\"")
        role: str = command['role']
-        print(f"Create user: {username}, password: {password}, role: {role}")
+        print(f"Create user: {user_name}, password: {password}, role: {role}")
        url = f'http://{self.host}:{self.port}/api/v1/admin/users'
-        response = requests.post(
+        response = self.session.post(
            url,
-            auth=HTTPBasicAuth(self.admin_account, self.admin_password),
-            json={'username': username, 'password': encrypt(password), 'role': role}
+            json={'user_name': user_name, 'password': encrypt(password), 'role': role}
        )
        res_json = response.json()
        if response.status_code == 200:
            self._print_table_simple(res_json['data'])
        else:
-            print(f"Fail to create user {username}, code: {res_json['code']}, message: {res_json['message']}")
+            print(f"Fail to create user {user_name}, code: {res_json['code']}, message: {res_json['message']}")

    def _handle_activate_user(self, command):
-        username_tree: Tree = command['username']
-        username: str = username_tree.children[0].strip("'\"")
+        user_name_tree: Tree = command['user_name']
+        user_name: str = user_name_tree.children[0].strip("'\"")
        activate_tree: Tree = command['activate_status']
        activate_status: str = activate_tree.children[0].strip("'\"")
        if activate_status.lower() in ['on', 'off']:
-            print(f"Alter user {username} activate status, turn {activate_status.lower()}.")
-            url = f'http://{self.host}:{self.port}/api/v1/admin/users/{username}/activate'
-            response = requests.put(url, auth=HTTPBasicAuth(self.admin_account, self.admin_password),
-                                    json={'activate_status': activate_status})
+            print(f"Alter user {user_name} activate status, turn {activate_status.lower()}.")
+            url = f'http://{self.host}:{self.port}/api/v1/admin/users/{user_name}/activate'
+            response = self.session.put(url, json={'activate_status': activate_status})
            res_json = response.json()
            if response.status_code == 200:
                print(res_json["message"])
@ -570,28 +688,178 @@ class AdminCLI(Cmd):
            print(f"Unknown activate status: {activate_status}.")

    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}")
-        url = f'http://{self.host}:{self.port}/api/v1/admin/users/{username}/datasets'
-        response = requests.get(url, auth=HTTPBasicAuth(self.admin_account, self.admin_password))
+        username_tree: Tree = command['user_name']
+        user_name: str = username_tree.children[0].strip("'\"")
+        print(f"Listing all datasets of user: {user_name}")
+        url = f'http://{self.host}:{self.port}/api/v1/admin/users/{user_name}/datasets'
+        response = self.session.get(url)
        res_json = response.json()
        if response.status_code == 200:
            self._print_table_simple(res_json['data'])
        else:
-            print(f"Fail to get all datasets of {username}, code: {res_json['code']}, message: {res_json['message']}")
+            print(f"Fail to get all datasets of {user_name}, code: {res_json['code']}, message: {res_json['message']}")

    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}")
-        url = f'http://{self.host}:{self.port}/api/v1/admin/users/{username}/agents'
-        response = requests.get(url, auth=HTTPBasicAuth(self.admin_account, self.admin_password))
+        username_tree: Tree = command['user_name']
+        user_name: str = username_tree.children[0].strip("'\"")
+        print(f"Listing all agents of user: {user_name}")
+        url = f'http://{self.host}:{self.port}/api/v1/admin/users/{user_name}/agents'
+        response = self.session.get(url)
        res_json = response.json()
        if response.status_code == 200:
            self._print_table_simple(res_json['data'])
        else:
-            print(f"Fail to get all agents of {username}, code: {res_json['code']}, message: {res_json['message']}")
+            print(f"Fail to get all agents of {user_name}, code: {res_json['code']}, message: {res_json['message']}")
+
+    def _create_role(self, command):
+        role_name_tree: Tree = command['role_name']
+        role_name: str = role_name_tree.children[0].strip("'\"")
+        desc_str: str = ''
+        if 'description' in command:
+            desc_tree: Tree = command['description']
+            desc_str = desc_tree.children[0].strip("'\"")
+
+        print(f"create role name: {role_name}, description: {desc_str}")
+        url = f'http://{self.host}:{self.port}/api/v1/admin/roles'
+        response = self.session.post(
+            url,
+            json={'role_name': role_name, 'description': desc_str}
+        )
+        res_json = response.json()
+        if response.status_code == 200:
+            self._print_table_simple(res_json['data'])
+        else:
+            print(f"Fail to create role {role_name}, code: {res_json['code']}, message: {res_json['message']}")
+
+    def _drop_role(self, command):
+        role_name_tree: Tree = command['role_name']
+        role_name: str = role_name_tree.children[0].strip("'\"")
+        print(f"drop role name: {role_name}")
+        url = f'http://{self.host}:{self.port}/api/v1/admin/roles/{role_name}'
+        response = self.session.delete(url)
+        res_json = response.json()
+        if response.status_code == 200:
+            self._print_table_simple(res_json['data'])
+        else:
+            print(f"Fail to drop role {role_name}, code: {res_json['code']}, message: {res_json['message']}")
+
+    def _alter_role(self, command):
+        role_name_tree: Tree = command['role_name']
+        role_name: str = role_name_tree.children[0].strip("'\"")
+        desc_tree: Tree = command['description']
+        desc_str: str = desc_tree.children[0].strip("'\"")
+
+        print(f"alter role name: {role_name}, description: {desc_str}")
+        url = f'http://{self.host}:{self.port}/api/v1/admin/roles/{role_name}'
+        response = self.session.put(
+            url,
+            json={'description': desc_str}
+        )
+        res_json = response.json()
+        if response.status_code == 200:
+            self._print_table_simple(res_json['data'])
+        else:
+            print(
+                f"Fail to update role {role_name} with description: {desc_str}, code: {res_json['code']}, message: {res_json['message']}")
+
+    def _list_roles(self, command):
+        print("Listing all roles")
+        url = f'http://{self.host}:{self.port}/api/v1/admin/roles'
+        response = self.session.get(url)
+        res_json = response.json()
+        if response.status_code == 200:
+            self._print_table_simple(res_json['data'])
+        else:
+            print(f"Fail to list roles, code: {res_json['code']}, message: {res_json['message']}")
+
+    def _show_role(self, command):
+        role_name_tree: Tree = command['role_name']
+        role_name: str = role_name_tree.children[0].strip("'\"")
+        print(f"show role: {role_name}")
+        url = f'http://{self.host}:{self.port}/api/v1/admin/roles/{role_name}/permission'
+        response = self.session.get(url)
+        res_json = response.json()
+        if response.status_code == 200:
+            self._print_table_simple(res_json['data'])
+        else:
+            print(f"Fail to list roles, code: {res_json['code']}, message: {res_json['message']}")
+
+    def _grant_permission(self, command):
+        role_name_tree: Tree = command['role_name']
+        role_name_str: str = role_name_tree.children[0].strip("'\"")
+        resource_tree: Tree = command['resource']
+        resource_str: str = resource_tree.children[0].strip("'\"")
+        action_tree_list: list = command['actions']
+        actions: list = []
+        for action_tree in action_tree_list:
+            action_str: str = action_tree.children[0].strip("'\"")
+            actions.append(action_str)
+        print(f"grant role_name: {role_name_str}, resource: {resource_str}, actions: {actions}")
+        url = f'http://{self.host}:{self.port}/api/v1/admin/roles/{role_name_str}/permission'
+        response = self.session.post(
+            url,
+            json={'actions': actions, 'resource': resource_str}
+        )
+        res_json = response.json()
+        if response.status_code == 200:
+            self._print_table_simple(res_json['data'])
+        else:
+            print(
+                f"Fail to grant role {role_name_str} with {actions} on {resource_str}, code: {res_json['code']}, message: {res_json['message']}")
+
+    def _revoke_permission(self, command):
+        role_name_tree: Tree = command['role_name']
+        role_name_str: str = role_name_tree.children[0].strip("'\"")
+        resource_tree: Tree = command['resource']
+        resource_str: str = resource_tree.children[0].strip("'\"")
+        action_tree_list: list = command['actions']
+        actions: list = []
+        for action_tree in action_tree_list:
+            action_str: str = action_tree.children[0].strip("'\"")
+            actions.append(action_str)
+        print(f"revoke role_name: {role_name_str}, resource: {resource_str}, actions: {actions}")
+        url = f'http://{self.host}:{self.port}/api/v1/admin/roles/{role_name_str}/permission'
+        response = self.session.delete(
+            url,
+            json={'actions': actions, 'resource': resource_str}
+        )
+        res_json = response.json()
+        if response.status_code == 200:
+            self._print_table_simple(res_json['data'])
+        else:
+            print(
+                f"Fail to revoke role {role_name_str} with {actions} on {resource_str}, code: {res_json['code']}, message: {res_json['message']}")
+
+    def _alter_user_role(self, command):
+        role_name_tree: Tree = command['role_name']
+        role_name_str: str = role_name_tree.children[0].strip("'\"")
+        user_name_tree: Tree = command['user_name']
+        user_name_str: str = user_name_tree.children[0].strip("'\"")
+        print(f"alter_user_role user_name: {user_name_str}, role_name: {role_name_str}")
+        url = f'http://{self.host}:{self.port}/api/v1/admin/users/{user_name_str}/role'
+        response = self.session.put(
+            url,
+            json={'role_name': role_name_str}
+        )
+        res_json = response.json()
+        if response.status_code == 200:
+            self._print_table_simple(res_json['data'])
+        else:
+            print(
+                f"Fail to alter user: {user_name_str} to role {role_name_str}, code: {res_json['code']}, message: {res_json['message']}")
+
+    def _show_user_permission(self, command):
+        user_name_tree: Tree = command['user_name']
+        user_name_str: str = user_name_tree.children[0].strip("'\"")
+        print(f"show_user_permission user_name: {user_name_str}")
+        url = f'http://{self.host}:{self.port}/api/v1/admin/users/{user_name_str}/permission'
+        response = self.session.get(url)
+        res_json = response.json()
+        if response.status_code == 200:
+            self._print_table_simple(res_json['data'])
+        else:
+            print(
+                f"Fail to show user: {user_name_str} permission, code: {res_json['code']}, message: {res_json['message']}")

    def _handle_meta_command(self, command):
        meta_command = command['command']
@ -634,27 +902,29 @@ def main():

    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.cmdloop()
+    args = cli.parse_connection_args(sys.argv)
+    if 'error' in args:
+        print(f"Error: {args['error']}")
+        return
+
+    if 'command' in args:
+        if 'password' not in args:
+            print("Error: password is missing")
+            return
+        if cli.verify_admin(args, single_command=True):
+            command: str = args['command']
+            print(f"Run single command: {command}")
+            cli.run_single_command(command)
    else:
-        print(r"""
-            ____  ___   ______________                 ___       __          _     
-           / __ \/   | / ____/ ____/ /___ _      __   /   | ____/ /___ ___  (_)___ 
-          / /_/ / /| |/ / __/ /_  / / __ \ | /| / /  / /| |/ __  / __ `__ \/ / __ \
-         / _, _/ ___ / /_/ / __/ / / /_/ / |/ |/ /  / ___ / /_/ / / / / / / / / / /
-        /_/ |_/_/  |_\____/_/   /_/\____/|__/|__/  /_/  |_\__,_/_/ /_/ /_/_/_/ /_/ 
-        """)
-        if cli.verify_admin(sys.argv):
+        if cli.verify_admin(args, single_command=False):
+            print(r"""
+                ____  ___   ______________                 ___       __          _     
+               / __ \/   | / ____/ ____/ /___ _      __   /   | ____/ /___ ___  (_)___ 
+              / /_/ / /| |/ / __/ /_  / / __ \ | /| / /  / /| |/ __  / __ `__ \/ / __ \
+             / _, _/ ___ / /_/ / __/ / / /_/ / |/ |/ /  / ___ / /_/ / / / / / / / / / /
+            /_/ |_/_/  |_\____/_/   /_/\____/|__/|__/  /_/  |_\__,_/_/ /_/ /_/_/_/ /_/ 
+            """)
            cli.cmdloop()
-            # cli.run_single_command(sys.argv[1:])


 if __name__ == '__main__':
--- a/admin/client/pyproject.toml
+++ b/admin/client/pyproject.toml
@ -1,6 +1,6 @@
 [project]
 name = "ragflow-cli"
-version = "0.21.0"
+version = "0.21.1"
 description = "Admin Service's client of [RAGFlow](https://github.com/infiniflow/ragflow). The Admin Service provides user management and system monitoring. "
 authors = [{ name = "Lynn", email = "lynn_inf@hotmail.com" }]
 license = { text = "Apache License, Version 2.0" }
--- a/admin/pyproject.toml
+++ b/admin/pyproject.toml
@ -1,24 +0,0 @@
-[project]
-name = "ragflow-cli"
-version = "0.21.0.dev2"
-description = "Admin Service's client of [RAGFlow](https://github.com/infiniflow/ragflow). The Admin Service provides user management and system monitoring. "
-authors = [{ name = "Lynn", email = "lynn_inf@hotmail.com" }]
-license = { text = "Apache License, Version 2.0" }
-readme = "README.md"
-requires-python = ">=3.10,<3.13"
-dependencies = [
-    "requests>=2.30.0,<3.0.0",
-    "beartype>=0.18.5,<0.19.0",
-    "pycryptodomex>=3.10.0",
-    "lark>=1.1.0",
-]
-
-[dependency-groups]
-test = [
-    "pytest>=8.3.5",
-    "requests>=2.32.3",
-    "requests-toolbelt>=1.0.0",
-]
-
-[project.scripts]
-ragflow-cli = "ragflow_cli.admin_client:main"
--- a/admin/server/admin_server.py
+++ b/admin/server/admin_server.py
@ -27,6 +27,9 @@ from api.utils.log_utils import init_root_logger
 from api.constants import SERVICE_CONF
 from api import settings
 from config import load_configurations, SERVICE_CONFIGS
+from auth import init_default_admin, setup_auth
+from flask_session import Session
+from flask_login import LoginManager

 stop_event = threading.Event()

@ -42,7 +45,17 @@ if __name__ == '__main__':

    app = Flask(__name__)
    app.register_blueprint(admin_bp)
+    app.config["SESSION_PERMANENT"] = False
+    app.config["SESSION_TYPE"] = "filesystem"
+    app.config["MAX_CONTENT_LENGTH"] = int(
+        os.environ.get("MAX_CONTENT_LENGTH", 1024 * 1024 * 1024)
+    )
+    Session(app)
+    login_manager = LoginManager()
+    login_manager.init_app(app)
    settings.init_settings()
+    setup_auth(login_manager)
+    init_default_admin()
    SERVICE_CONFIGS.configs = load_configurations(SERVICE_CONF)

    try:
--- a/admin/server/auth.py
+++ b/admin/server/auth.py
@ -18,11 +18,124 @@
 import logging
 import uuid
 from functools import wraps
+from datetime import datetime
 from flask import request, jsonify
+from flask_login import current_user, login_user
+from itsdangerous.url_safe import URLSafeTimedSerializer as Serializer

-from api.common.exceptions import AdminException
+from api import settings
+from api.common.exceptions import AdminException, UserNotFoundError
 from api.db.init_data import encode_to_base64
 from api.db.services import UserService
+from api.db import ActiveEnum, StatusEnum
+from api.utils.crypt import decrypt
+from api.utils import (
+    current_timestamp,
+    datetime_format,
+    get_format_time,
+    get_uuid,
+)
+from api.utils.api_utils import (
+    construct_response,
+)
+
+
+def setup_auth(login_manager):
+    @login_manager.request_loader
+    def load_user(web_request):
+        jwt = Serializer(secret_key=settings.SECRET_KEY)
+        authorization = web_request.headers.get("Authorization")
+        if authorization:
+            try:
+                access_token = str(jwt.loads(authorization))
+
+                if not access_token or not access_token.strip():
+                    logging.warning("Authentication attempt with empty access token")
+                    return None
+
+                # Access tokens should be UUIDs (32 hex characters)
+                if len(access_token.strip()) < 32:
+                    logging.warning(f"Authentication attempt with invalid token format: {len(access_token)} chars")
+                    return None
+
+                user = UserService.query(
+                    access_token=access_token, status=StatusEnum.VALID.value
+                )
+                if user:
+                    if not user[0].access_token or not user[0].access_token.strip():
+                        logging.warning(f"User {user[0].email} has empty access_token in database")
+                        return None
+                    return user[0]
+                else:
+                    return None
+            except Exception as e:
+                logging.warning(f"load_user got exception {e}")
+                return None
+        else:
+            return None
+
+
+def init_default_admin():
+    # Verify that at least one active admin user exists. If not, create a default one.
+    users = UserService.query(is_superuser=True)
+    if not users:
+        default_admin = {
+            "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(**default_admin):
+            raise AdminException("Can't init admin.", 500)
+    elif not any([u.is_active == ActiveEnum.ACTIVE.value for u in users]):
+        raise AdminException("No active admin. Please update 'is_active' in db manually.", 500)
+
+
+def check_admin_auth(func):
+    @wraps(func)
+    def wrapper(*args, **kwargs):
+        user = UserService.filter_by_id(current_user.id)
+        if not user:
+            raise UserNotFoundError(current_user.email)
+        if not user.is_superuser:
+            raise AdminException("Not admin", 403)
+        if user.is_active == ActiveEnum.INACTIVE.value:
+            raise AdminException(f"User {current_user.email} inactive", 403)
+
+        return func(*args, **kwargs)
+
+    return wrapper
+
+
+def login_admin(email: str, password: str):
+    """
+    :param email: admin email
+    :param password: string before decrypt
+    """
+    users = UserService.query(email=email)
+    if not users:
+        raise UserNotFoundError(email)
+    psw = decrypt(password)
+    user = UserService.query_user(email, psw)
+    if not user:
+        raise AdminException("Email and password do not match!")
+    if not user.is_superuser:
+        raise AdminException("Not admin", 403)
+    if user.is_active == ActiveEnum.INACTIVE.value:
+        raise AdminException(f"User {email} inactive", 403)
+
+    resp = user.to_json()
+    user.access_token = get_uuid()
+    login_user(user)
+    user.update_time = (current_timestamp(),)
+    user.update_date = (datetime_format(datetime.now()),)
+    user.last_login_time = get_format_time()
+    user.save()
+    msg = "Welcome back!"
+    return construct_response(data=resp, auth=user.get_id(), message=msg)


 def check_admin(username: str, password: str):
@ -61,12 +174,18 @@ def login_verify(f):

        username = auth.parameters['username']
        password = auth.parameters['password']
-        # TODO: to check the username and password from DB
-        if check_admin(username, password) is False:
+        try:
+            if check_admin(username, password) is False:
+                return jsonify({
+                    "code": 500,
+                    "message": "Access denied",
+                    "data": None
+                }), 200
+        except Exception as e:
+            error_msg = str(e)
            return jsonify({
-                "code": 403,
-                "message": "Access denied",
-                "data": None
+                "code": 500,
+                "message": error_msg
            }), 200

        return f(*args, **kwargs)
--- a/admin/server/roles.py
+++ b/admin/server/roles.py
@ -0,0 +1,76 @@
+#
+#  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 logging
+
+from typing import Dict, Any
+
+from api.common.exceptions import AdminException
+
+
+class RoleMgr:
+    @staticmethod
+    def create_role(role_name: str, description: str):
+        error_msg = f"not implement: create role: {role_name}, description: {description}"
+        logging.error(error_msg)
+        raise AdminException(error_msg)
+
+    @staticmethod
+    def update_role_description(role_name: str, description: str) -> Dict[str, Any]:
+        error_msg = f"not implement: update role: {role_name} with description: {description}"
+        logging.error(error_msg)
+        raise AdminException(error_msg)
+
+    @staticmethod
+    def delete_role(role_name: str) -> Dict[str, Any]:
+        error_msg = f"not implement: drop role: {role_name}"
+        logging.error(error_msg)
+        raise AdminException(error_msg)
+
+    @staticmethod
+    def list_roles() -> Dict[str, Any]:
+        error_msg = "not implement: list roles"
+        logging.error(error_msg)
+        raise AdminException(error_msg)
+
+    @staticmethod
+    def get_role_permission(role_name: str) -> Dict[str, Any]:
+        error_msg = f"not implement: show role {role_name}"
+        logging.error(error_msg)
+        raise AdminException(error_msg)
+
+    @staticmethod
+    def grant_role_permission(role_name: str, actions: list, resource: str) -> Dict[str, Any]:
+        error_msg = f"not implement: grant role {role_name} actions: {actions} on {resource}"
+        logging.error(error_msg)
+        raise AdminException(error_msg)
+
+    @staticmethod
+    def revoke_role_permission(role_name: str, actions: list, resource: str) -> Dict[str, Any]:
+        error_msg = f"not implement: revoke role {role_name} actions: {actions} on {resource}"
+        logging.error(error_msg)
+        raise AdminException(error_msg)
+
+    @staticmethod
+    def update_user_role(user_name: str, role_name: str) -> Dict[str, Any]:
+        error_msg = f"not implement: update user role: {user_name} to role {role_name}"
+        logging.error(error_msg)
+        raise AdminException(error_msg)
+
+    @staticmethod
+    def get_user_permission(user_name: str) -> Dict[str, Any]:
+        error_msg = f"not implement: get user permission: {user_name}"
+        logging.error(error_msg)
+        raise AdminException(error_msg)
--- a/admin/server/routes.py
+++ b/admin/server/routes.py
@ -14,17 +14,44 @@
 #  limitations under the License.
 #

+import secrets

 from flask import Blueprint, request
+from flask_login import current_user, logout_user, login_required

-from auth import login_verify
+from auth import login_verify, login_admin, check_admin_auth
 from responses import success_response, error_response
 from services import UserMgr, ServiceMgr, UserServiceMgr
+from roles import RoleMgr
 from api.common.exceptions import AdminException

 admin_bp = Blueprint('admin', __name__, url_prefix='/api/v1/admin')


+@admin_bp.route('/login', methods=['POST'])
+def login():
+    if not request.json:
+        return error_response('Authorize admin failed.' ,400)
+    try:
+        email = request.json.get("email", "")
+        password = request.json.get("password", "")
+        return login_admin(email, password)
+    except Exception as e:
+        return error_response(str(e), 500)
+
+
+@admin_bp.route('/logout', methods=['GET'])
+@login_required
+def logout():
+    try:
+        current_user.access_token = f"INVALID_{secrets.token_hex(16)}"
+        current_user.save()
+        logout_user()
+        return success_response(True)
+    except Exception as e:
+        return error_response(str(e), 500)
+
+
@admin_bp.route('/auth', methods=['GET'])
@login_verify
 def auth_admin():
@ -35,7 +62,8 @@ def auth_admin():


@admin_bp.route('/users', methods=['GET'])
-@login_verify
+@login_required
+@check_admin_auth
 def list_users():
    try:
        users = UserMgr.get_all_users()
@ -45,7 +73,8 @@ def list_users():


@admin_bp.route('/users', methods=['POST'])
-@login_verify
+@login_required
+@check_admin_auth
 def create_user():
    try:
        data = request.get_json()
@ -71,7 +100,8 @@ def create_user():


@admin_bp.route('/users/<username>', methods=['DELETE'])
-@login_verify
+@login_required
+@check_admin_auth
 def delete_user(username):
    try:
        res = UserMgr.delete_user(username)
@ -87,7 +117,8 @@ def delete_user(username):


@admin_bp.route('/users/<username>/password', methods=['PUT'])
-@login_verify
+@login_required
+@check_admin_auth
 def change_password(username):
    try:
        data = request.get_json()
@ -105,7 +136,8 @@ def change_password(username):


@admin_bp.route('/users/<username>/activate', methods=['PUT'])
-@login_verify
+@login_required
+@check_admin_auth
 def alter_user_activate_status(username):
    try:
        data = request.get_json()
@ -121,7 +153,8 @@ def alter_user_activate_status(username):


@admin_bp.route('/users/<username>', methods=['GET'])
-@login_verify
+@login_required
+@check_admin_auth
 def get_user_details(username):
    try:
        user_details = UserMgr.get_user_details(username)
@ -134,7 +167,8 @@ def get_user_details(username):


@admin_bp.route('/users/<username>/datasets', methods=['GET'])
-@login_verify
+@login_required
+@check_admin_auth
 def get_user_datasets(username):
    try:
        datasets_list = UserServiceMgr.get_user_datasets(username)
@ -147,7 +181,8 @@ def get_user_datasets(username):


@admin_bp.route('/users/<username>/agents', methods=['GET'])
-@login_verify
+@login_required
+@check_admin_auth
 def get_user_agents(username):
    try:
        agents_list = UserServiceMgr.get_user_agents(username)
@ -160,7 +195,8 @@ def get_user_agents(username):


@admin_bp.route('/services', methods=['GET'])
-@login_verify
+@login_required
+@check_admin_auth
 def get_services():
    try:
        services = ServiceMgr.get_all_services()
@ -170,7 +206,8 @@ def get_services():


@admin_bp.route('/service_types/<service_type>', methods=['GET'])
-@login_verify
+@login_required
+@check_admin_auth
 def get_services_by_type(service_type_str):
    try:
        services = ServiceMgr.get_services_by_type(service_type_str)
@ -180,7 +217,8 @@ def get_services_by_type(service_type_str):


@admin_bp.route('/services/<service_id>', methods=['GET'])
-@login_verify
+@login_required
+@check_admin_auth
 def get_service(service_id):
    try:
        services = ServiceMgr.get_service_details(service_id)
@ -190,7 +228,8 @@ def get_service(service_id):


@admin_bp.route('/services/<service_id>', methods=['DELETE'])
-@login_verify
+@login_required
+@check_admin_auth
 def shutdown_service(service_id):
    try:
        services = ServiceMgr.shutdown_service(service_id)
@ -200,10 +239,133 @@ def shutdown_service(service_id):


@admin_bp.route('/services/<service_id>', methods=['PUT'])
-@login_verify
+@login_required
+@check_admin_auth
 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_bp.route('/roles', methods=['POST'])
+@login_required
+@check_admin_auth
+def create_role():
+    try:
+        data = request.get_json()
+        if not data or 'role_name' not in data:
+            return error_response("Role name is required", 400)
+        role_name: str = data['role_name']
+        description: str = data['description']
+        res = RoleMgr.create_role(role_name, description)
+        return success_response(res)
+    except Exception as e:
+        return error_response(str(e), 500)
+
+
+@admin_bp.route('/roles/<role_name>', methods=['PUT'])
+@login_required
+@check_admin_auth
+def update_role(role_name: str):
+    try:
+        data = request.get_json()
+        if not data or 'description' not in data:
+            return error_response("Role description is required", 400)
+        description: str = data['description']
+        res = RoleMgr.update_role_description(role_name, description)
+        return success_response(res)
+    except Exception as e:
+        return error_response(str(e), 500)
+
+
+@admin_bp.route('/roles/<role_name>', methods=['DELETE'])
+@login_required
+@check_admin_auth
+def delete_role(role_name: str):
+    try:
+        res = RoleMgr.delete_role(role_name)
+        return success_response(res)
+    except Exception as e:
+        return error_response(str(e), 500)
+
+
+@admin_bp.route('/roles', methods=['GET'])
+@login_required
+@check_admin_auth
+def list_roles():
+    try:
+        res = RoleMgr.list_roles()
+        return success_response(res)
+    except Exception as e:
+        return error_response(str(e), 500)
+
+
+@admin_bp.route('/roles/<role_name>/permission', methods=['GET'])
+@login_required
+@check_admin_auth
+def get_role_permission(role_name: str):
+    try:
+        res = RoleMgr.get_role_permission(role_name)
+        return success_response(res)
+    except Exception as e:
+        return error_response(str(e), 500)
+
+
+@admin_bp.route('/roles/<role_name>/permission', methods=['POST'])
+@login_required
+@check_admin_auth
+def grant_role_permission(role_name: str):
+    try:
+        data = request.get_json()
+        if not data or 'actions' not in data or 'resource' not in data:
+            return error_response("Permission is required", 400)
+        actions: list = data['actions']
+        resource: str = data['resource']
+        res = RoleMgr.grant_role_permission(role_name, actions, resource)
+        return success_response(res)
+    except Exception as e:
+        return error_response(str(e), 500)
+
+
+@admin_bp.route('/roles/<role_name>/permission', methods=['DELETE'])
+@login_required
+@check_admin_auth
+def revoke_role_permission(role_name: str):
+    try:
+        data = request.get_json()
+        if not data or 'actions' not in data or 'resource' not in data:
+            return error_response("Permission is required", 400)
+        actions: list = data['actions']
+        resource: str = data['resource']
+        res = RoleMgr.revoke_role_permission(role_name, actions, resource)
+        return success_response(res)
+    except Exception as e:
+        return error_response(str(e), 500)
+
+
+@admin_bp.route('/users/<user_name>/role', methods=['PUT'])
+@login_required
+@check_admin_auth
+def update_user_role(user_name: str):
+    try:
+        data = request.get_json()
+        if not data or 'role_name' not in data:
+            return error_response("Role name is required", 400)
+        role_name: str = data['role_name']
+        res = RoleMgr.update_user_role(user_name, role_name)
+        return success_response(res)
+    except Exception as e:
+        return error_response(str(e), 500)
+
+
+@admin_bp.route('/users/<user_name>/permission', methods=['GET'])
+@login_required
+@check_admin_auth
+def get_user_permission(user_name: str):
+    try:
+        res = RoleMgr.get_user_permission(user_name)
+        return success_response(res)
+    except Exception as e:
+        return error_response(str(e), 500)
--- a/admin/server/services.py
+++ b/admin/server/services.py
@ -36,8 +36,13 @@ class UserMgr:
        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})
+            result.append({
+                'email': user.email,
+                'nickname': user.nickname,
+                'create_date': user.create_date,
+                'is_active': user.is_active,
+                'is_superuser': user.is_superuser,
+            })
        return result

    @staticmethod
@ -50,7 +55,6 @@ class UserMgr:
                'email': user.email,
                'language': user.language,
                'last_login_time': user.last_login_time,
-                'is_authenticated': user.is_authenticated,
                'is_active': user.is_active,
                'is_anonymous': user.is_anonymous,
                'login_channel': user.login_channel,
@ -166,8 +170,7 @@ class UserServiceMgr:
        return [{
            'title': r['title'],
            'permission': r['permission'],
-            'canvas_type': r['canvas_type'],
-            'canvas_category': r['canvas_category']
+            'canvas_category': r['canvas_category'].split('_')[0]
        } for r in res]


--- a/agent/templates/chunk_summary.json
+++ b/agent/templates/chunk_summary.json
@ -350,7 +350,7 @@
                            ]
                        },
                        "label": "Tokenizer",
-                        "name": "Tokenizer"
+                        "name": "Indexer"
                    },
                    "dragging": false,
                    "id": "Tokenizer:EightRocketsAppear",
--- a/agent/tools/retrieval.py
+++ b/agent/tools/retrieval.py
@ -18,12 +18,14 @@ import re
 from abc import ABC
 from agent.tools.base import ToolParamBase, ToolBase, ToolMeta
 from api.db import LLMType
+from api.db.services.document_service import DocumentService
+from api.db.services.dialog_service import meta_filter
 from api.db.services.knowledgebase_service import KnowledgebaseService
 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.generator import cross_languages, kb_prompt
+from rag.prompts.generator import cross_languages, kb_prompt, gen_meta_filter


 class RetrievalParam(ToolParamBase):
@ -58,6 +60,7 @@ class RetrievalParam(ToolParamBase):
        self.use_kg = False
        self.cross_languages = []
        self.toc_enhance = False
+        self.meta_data_filter={}

    def check(self):
        self.check_decimal_float(self.similarity_threshold, "[Retrieval] Similarity threshold")
@ -117,6 +120,21 @@ class Retrieval(ToolBase, ABC):
        vars = self.get_input_elements_from_text(kwargs["query"])
        vars = {k:o["value"] for k,o in vars.items()}
        query = self.string_format(kwargs["query"], vars)
+        
+        doc_ids=[]
+        if self._param.meta_data_filter!={}:
+            metas = DocumentService.get_meta_by_kbs(kb_ids)
+            if self._param.meta_data_filter.get("method") == "auto":
+                chat_mdl = LLMBundle(self._canvas.get_tenant_id(), LLMType.CHAT)
+                filters = gen_meta_filter(chat_mdl, metas, query)
+                doc_ids.extend(meta_filter(metas, filters))
+                if not doc_ids:
+                    doc_ids = None
+            elif self._param.meta_data_filter.get("method") == "manual":
+                doc_ids.extend(meta_filter(metas, self._param.meta_data_filter["manual"]))
+                if not doc_ids:
+                    doc_ids = None
+
        if self._param.cross_languages:
            query = cross_languages(kbs[0].tenant_id, None, query, self._param.cross_languages)

@ -131,6 +149,7 @@ class Retrieval(ToolBase, ABC):
                self._param.top_n,
                self._param.similarity_threshold,
                1 - self._param.keywords_similarity_weight,
+                doc_ids=doc_ids,
                aggs=False,
                rerank_mdl=rerank_mdl,
                rank_feature=label_question(query, kbs),
--- a/api/apps/chunk_app.py
+++ b/api/apps/chunk_app.py
@ -350,7 +350,8 @@ def retrieval_test():
                               float(req.get("similarity_threshold", 0.0)),
                               float(req.get("vector_similarity_weight", 0.3)),
                               top,
-                               doc_ids, rerank_mdl=rerank_mdl, highlight=req.get("highlight"),
+                               doc_ids, rerank_mdl=rerank_mdl,
+                                             highlight=req.get("highlight", False),
                               rank_feature=labels
                               )
        if use_kg:
--- a/api/apps/document_app.py
+++ b/api/apps/document_app.py
@ -45,7 +45,7 @@ from api.utils.api_utils import (
 from api.utils.file_utils import filename_type, get_project_base_directory, thumbnail
 from api.utils.web_utils import CONTENT_TYPE_MAP, html2pdf, is_valid_url
 from deepdoc.parser.html_parser import RAGFlowHtmlParser
-from rag.nlp import search
+from rag.nlp import search, rag_tokenizer
 from rag.utils.storage_factory import STORAGE_IMPL


@ -524,6 +524,21 @@ def rename():
            e, file = FileService.get_by_id(informs[0].file_id)
            FileService.update_by_id(file.id, {"name": req["name"]})

+        tenant_id = DocumentService.get_tenant_id(req["doc_id"])
+        title_tks = rag_tokenizer.tokenize(req["name"])
+        es_body = {
+            "docnm_kwd": req["name"],
+            "title_tks": title_tks,
+            "title_sm_tks": rag_tokenizer.fine_grained_tokenize(title_tks),
+        }
+        if settings.docStoreConn.indexExist(search.index_name(tenant_id), doc.kb_id):
+            settings.docStoreConn.update(
+                {"doc_id": req["doc_id"]},
+                es_body,
+                search.index_name(tenant_id),
+                doc.kb_id,
+            )
+
        return get_json_result(data=True)
    except Exception as e:
        return server_error_response(e)
--- a/api/apps/kb_app.py
+++ b/api/apps/kb_app.py
@ -70,6 +70,7 @@ def create():
        e, t = TenantService.get_by_id(current_user.id)
        if not e:
            return get_data_error_result(message="Tenant not found.")
+
        req["parser_config"] = {
            "layout_recognize": "DeepDOC",
            "chunk_token_num": 512,
@ -579,7 +580,7 @@ def run_graphrag():
    sample_document = documents[0]
    document_ids = [document["id"] for document in documents]

-    task_id = queue_raptor_o_graphrag_tasks(doc=sample_document, ty="graphrag", priority=0, fake_doc_id=GRAPH_RAPTOR_FAKE_DOC_ID, doc_ids=list(document_ids))
+    task_id = queue_raptor_o_graphrag_tasks(sample_doc_id=sample_document, ty="graphrag", priority=0, fake_doc_id=GRAPH_RAPTOR_FAKE_DOC_ID, doc_ids=list(document_ids))

    if not KnowledgebaseService.update_by_id(kb.id, {"graphrag_task_id": task_id}):
        logging.warning(f"Cannot save graphrag_task_id for kb {kb_id}")
@ -648,7 +649,7 @@ def run_raptor():
    sample_document = documents[0]
    document_ids = [document["id"] for document in documents]

-    task_id = queue_raptor_o_graphrag_tasks(doc=sample_document, ty="raptor", priority=0, fake_doc_id=GRAPH_RAPTOR_FAKE_DOC_ID, doc_ids=list(document_ids))
+    task_id = queue_raptor_o_graphrag_tasks(sample_doc_id=sample_document, ty="raptor", priority=0, fake_doc_id=GRAPH_RAPTOR_FAKE_DOC_ID, doc_ids=list(document_ids))

    if not KnowledgebaseService.update_by_id(kb.id, {"raptor_task_id": task_id}):
        logging.warning(f"Cannot save raptor_task_id for kb {kb_id}")
@ -717,7 +718,7 @@ def run_mindmap():
    sample_document = documents[0]
    document_ids = [document["id"] for document in documents]

-    task_id = queue_raptor_o_graphrag_tasks(doc=sample_document, ty="mindmap", priority=0, fake_doc_id=GRAPH_RAPTOR_FAKE_DOC_ID, doc_ids=list(document_ids))
+    task_id = queue_raptor_o_graphrag_tasks(sample_doc_id=sample_document, ty="mindmap", priority=0, fake_doc_id=GRAPH_RAPTOR_FAKE_DOC_ID, doc_ids=list(document_ids))

    if not KnowledgebaseService.update_by_id(kb.id, {"mindmap_task_id": task_id}):
        logging.warning(f"Cannot save mindmap_task_id for kb {kb_id}")
--- a/api/apps/sdk/doc.py
+++ b/api/apps/sdk/doc.py
@ -470,6 +470,20 @@ def list_docs(dataset_id, tenant_id):
        required: false
        default: 0
        description: Unix timestamp for filtering documents created before this time. 0 means no filter.
+      - in: query
+        name: suffix
+        type: array
+        items:
+          type: string
+        required: false
+        description: Filter by file suffix (e.g., ["pdf", "txt", "docx"]).
+      - in: query
+        name: run
+        type: array
+        items:
+          type: string
+        required: false
+        description: Filter by document run status. Supports both numeric ("0", "1", "2", "3", "4") and text formats ("UNSTART", "RUNNING", "CANCEL", "DONE", "FAIL").
      - in: header
        name: Authorization
        type: string
@ -512,63 +526,62 @@ def list_docs(dataset_id, tenant_id):
                    description: Processing status.
    """
    if not KnowledgebaseService.accessible(kb_id=dataset_id, user_id=tenant_id):
-        return get_error_data_result(message=f"You don't own the dataset {dataset_id}. ")
-    id = request.args.get("id")
-    name = request.args.get("name")
+      return get_error_data_result(message=f"You don't own the dataset {dataset_id}. ")

-    if id and not DocumentService.query(id=id, kb_id=dataset_id):
-        return get_error_data_result(message=f"You don't own the document {id}.")
+    q = request.args
+    document_id = q.get("id")  
+    name        = q.get("name")
+
+    if document_id and not DocumentService.query(id=document_id, kb_id=dataset_id):
+        return get_error_data_result(message=f"You don't own the document {document_id}.")
    if name and not DocumentService.query(name=name, kb_id=dataset_id):
        return get_error_data_result(message=f"You don't own the document {name}.")

-    page = int(request.args.get("page", 1))
-    keywords = request.args.get("keywords", "")
-    page_size = int(request.args.get("page_size", 30))
-    orderby = request.args.get("orderby", "create_time")
-    if request.args.get("desc") == "False":
-        desc = False
-    else:
-        desc = True
-    docs, tol = DocumentService.get_list(dataset_id, page, page_size, orderby, desc, keywords, id, name)
+    page        = int(q.get("page", 1))
+    page_size   = int(q.get("page_size", 30))  
+    orderby     = q.get("orderby", "create_time")
+    desc        = str(q.get("desc", "true")).strip().lower() != "false"
+    keywords    = q.get("keywords", "")

-    create_time_from = int(request.args.get("create_time_from", 0))
-    create_time_to = int(request.args.get("create_time_to", 0))
+    # filters - align with OpenAPI parameter names
+    suffix               = q.getlist("suffix") 
+    run_status           = q.getlist("run")   
+    create_time_from     = int(q.get("create_time_from", 0))  
+    create_time_to       = int(q.get("create_time_to", 0))    

+    # map run status (accept text or numeric) - align with API parameter
+    run_status_text_to_numeric = {"UNSTART": "0", "RUNNING": "1", "CANCEL": "2", "DONE": "3", "FAIL": "4"}
+    run_status_converted = [run_status_text_to_numeric.get(v, v) for v in run_status]
+
+    docs, total = DocumentService.get_list(
+        dataset_id, page, page_size, orderby, desc, keywords, document_id, name, suffix, run_status_converted
+    )
+
+    # time range filter (0 means no bound)
    if create_time_from or create_time_to:
-        filtered_docs = []
-        for doc in docs:
-            doc_create_time = doc.get("create_time", 0)
-            if (create_time_from == 0 or doc_create_time >= create_time_from) and (create_time_to == 0 or doc_create_time <= create_time_to):
-                filtered_docs.append(doc)
-        docs = filtered_docs
+        docs = [
+            d for d in docs
+            if (create_time_from == 0 or d.get("create_time", 0) >= create_time_from)
+            and (create_time_to == 0 or d.get("create_time", 0) <= create_time_to)
+        ]

-    # rename key's name
-    renamed_doc_list = []
+    # rename keys + map run status back to text for output
    key_mapping = {
        "chunk_num": "chunk_count",
-        "kb_id": "dataset_id",
+        "kb_id": "dataset_id", 
        "token_num": "token_count",
        "parser_id": "chunk_method",
    }
-    run_mapping = {
-        "0": "UNSTART",
-        "1": "RUNNING",
-        "2": "CANCEL",
-        "3": "DONE",
-        "4": "FAIL",
-    }
-    for doc in docs:
-        renamed_doc = {}
-        for key, value in doc.items():
-            if key == "run":
-                renamed_doc["run"] = run_mapping.get(str(value))
-            new_key = key_mapping.get(key, key)
-            renamed_doc[new_key] = value
-            if key == "run":
-                renamed_doc["run"] = run_mapping.get(value)
-        renamed_doc_list.append(renamed_doc)
-    return get_result(data={"total": tol, "docs": renamed_doc_list})
+    run_status_numeric_to_text = {"0": "UNSTART", "1": "RUNNING", "2": "CANCEL", "3": "DONE", "4": "FAIL"}

+    output_docs = []
+    for d in docs:
+        renamed_doc = {key_mapping.get(k, k): v for k, v in d.items()}
+        if "run" in d:
+            renamed_doc["run"] = run_status_numeric_to_text.get(str(d["run"]), d["run"])
+        output_docs.append(renamed_doc)
+
+    return get_result(data={"total": total, "docs": output_docs})

@manager.route("/datasets/<dataset_id>/documents", methods=["DELETE"])  # noqa: F821
@token_required
--- a/api/apps/user_app.py
+++ b/api/apps/user_app.py
@ -22,7 +22,7 @@ import secrets
 import time
 from datetime import datetime

-from flask import redirect, request, session, Response
+from flask import redirect, request, session, make_response
 from flask_login import current_user, login_required, login_user, logout_user
 from werkzeug.security import check_password_hash, generate_password_hash

@ -866,7 +866,9 @@ def forget_get_captcha():
    from captcha.image import ImageCaptcha
    image = ImageCaptcha(width=300, height=120, font_sizes=[50, 60, 70])
    img_bytes = image.generate(captcha_text).read()
-    return Response(img_bytes, mimetype="image/png")
+    response = make_response(img_bytes)
+    response.headers.set("Content-Type", "image/JPEG")
+    return response


@manager.route("/forget/otp", methods=["POST"])  # noqa: F821
--- a/api/common/exceptions.py
+++ b/api/common/exceptions.py
@ -36,3 +36,8 @@ class UserAlreadyExistsError(AdminException):
 class CannotDeleteAdminError(AdminException):
    def __init__(self):
        super().__init__("Cannot delete admin account", 403)
+
+
+class NotAdminError(AdminException):
+    def __init__(self, username):
+        super().__init__(f"User '{username}' is not admin", 403)
--- a/api/db/services/document_service.py
+++ b/api/db/services/document_service.py
@ -79,7 +79,7 @@ class DocumentService(CommonService):
    @classmethod
    @DB.connection_context()
    def get_list(cls, kb_id, page_number, items_per_page,
-                 orderby, desc, keywords, id, name):
+                 orderby, desc, keywords, id, name, suffix=None, run = None):
        fields = cls.get_cls_model_fields()
        docs = cls.model.select(*[*fields, UserCanvas.title]).join(File2Document, on = (File2Document.document_id == cls.model.id))\
            .join(File, on = (File.id == File2Document.file_id))\
@ -96,6 +96,10 @@ class DocumentService(CommonService):
            docs = docs.where(
                fn.LOWER(cls.model.name).contains(keywords.lower())
            )
+        if suffix:
+            docs = docs.where(cls.model.suffix.in_(suffix))
+        if run:
+            docs = docs.where(cls.model.run.in_(run))
        if desc:
            docs = docs.order_by(cls.model.getter_by(orderby).desc())
        else:
@ -667,9 +671,11 @@ class DocumentService(CommonService):
    @classmethod
    @DB.connection_context()
    def _sync_progress(cls, docs:list[dict]):
+        from api.db.services.task_service import TaskService
+
        for d in docs:
            try:
-                tsks = Task.query(doc_id=d["id"], order_by=Task.create_time)
+                tsks = TaskService.query(doc_id=d["id"], order_by=Task.create_time)
                if not tsks:
                    continue
                msg = []
@ -787,21 +793,23 @@ class DocumentService(CommonService):
            "cancelled": int(cancelled),
        }

-def queue_raptor_o_graphrag_tasks(doc, ty, priority, fake_doc_id="", doc_ids=[]):
+def queue_raptor_o_graphrag_tasks(sample_doc_id, ty, priority, fake_doc_id="", doc_ids=[]):
    """
    You can provide a fake_doc_id to bypass the restriction of tasks at the knowledgebase level.
    Optionally, specify a list of doc_ids to determine which documents participate in the task.
    """
-    chunking_config = DocumentService.get_chunking_config(doc["id"])
+    assert ty in ["graphrag", "raptor", "mindmap"], "type should be graphrag, raptor or mindmap"
+
+    chunking_config = DocumentService.get_chunking_config(sample_doc_id["id"])
    hasher = xxhash.xxh64()
    for field in sorted(chunking_config.keys()):
        hasher.update(str(chunking_config[field]).encode("utf-8"))

    def new_task():
-        nonlocal doc
+        nonlocal sample_doc_id
        return {
            "id": get_uuid(),
-            "doc_id": fake_doc_id if fake_doc_id else doc["id"],
+            "doc_id": sample_doc_id["id"],
            "from_page": 100000000,
            "to_page": 100000000,
            "task_type": ty,
@ -816,9 +824,9 @@ def queue_raptor_o_graphrag_tasks(doc, ty, priority, fake_doc_id="", doc_ids=[])
    task["digest"] = hasher.hexdigest()
    bulk_insert_into_db(Task, [task], True)

-    if ty in ["graphrag", "raptor", "mindmap"]:
-        task["doc_ids"] = doc_ids
-        DocumentService.begin2parse(doc["id"])
+    task["doc_id"] = fake_doc_id
+    task["doc_ids"] = doc_ids
+    DocumentService.begin2parse(sample_doc_id["id"])
    assert REDIS_CONN.queue_product(get_svr_queue_name(priority), message=task), "Can't access Redis. Please check the Redis' status."
    return task["id"]

--- a/api/db/services/llm_service.py
+++ b/api/db/services/llm_service.py
@ -205,32 +205,31 @@ class LLMBundle(LLM4Tenant):
            return txt

        return txt[last_think_end + len("</think>") :]
-    
+
    @staticmethod
    def _clean_param(chat_partial, **kwargs):
        func = chat_partial.func
        sig = inspect.signature(func)
-        keyword_args = []
        support_var_args = False
-        for param in sig.parameters.values():
-            if param.kind == inspect.Parameter.VAR_KEYWORD or param.kind == inspect.Parameter.VAR_POSITIONAL:
-                support_var_args = True
-            elif param.kind == inspect.Parameter.KEYWORD_ONLY:
-                keyword_args.append(param.name)
+        allowed_params = set()

-        use_kwargs = kwargs
-        if not support_var_args:
-            use_kwargs = {k: v for k, v in kwargs.items() if k in keyword_args}
-        return use_kwargs
-        
+        for param in sig.parameters.values():
+            if param.kind == inspect.Parameter.VAR_KEYWORD:
+                support_var_args = True
+            elif param.kind in (inspect.Parameter.POSITIONAL_OR_KEYWORD, inspect.Parameter.KEYWORD_ONLY):
+                allowed_params.add(param.name)
+        if support_var_args:
+            return kwargs
+        else:
+            return {k: v for k, v in kwargs.items() if k in allowed_params}
    def chat(self, system: str, history: list, gen_conf: dict = {}, **kwargs) -> str:
        if self.langfuse:
            generation = self.langfuse.start_generation(trace_context=self.trace_context, name="chat", model=self.llm_name, input={"system": system, "history": history})

-        chat_partial = partial(self.mdl.chat, system, history, gen_conf)
+        chat_partial = partial(self.mdl.chat, system, history, gen_conf, **kwargs)
        if self.is_tools and self.mdl.is_tools:
-            chat_partial = partial(self.mdl.chat_with_tools, system, history, gen_conf)
-            
+            chat_partial = partial(self.mdl.chat_with_tools, system, history, gen_conf, **kwargs)
+
        use_kwargs = self._clean_param(chat_partial, **kwargs)
        txt, used_tokens = chat_partial(**use_kwargs)
        txt = self._remove_reasoning_content(txt)
@ -266,7 +265,7 @@ class LLMBundle(LLM4Tenant):
                break

            if txt.endswith("</think>"):
-                ans = ans.rstrip("</think>")
+                ans = ans[: -len("</think>")]

            if not self.verbose_tool_use:
                txt = re.sub(r"<tool_call>.*?</tool_call>", "", txt, flags=re.DOTALL)
--- a/api/db/services/task_service.py
+++ b/api/db/services/task_service.py
@ -351,7 +351,7 @@ def queue_tasks(doc: dict, bucket: str, name: str, priority: int):
            "progress": 0.0,
            "from_page": 0,
            "to_page": 100000000,
-            "begin_at": datetime.now(),
+            "begin_at": datetime.now().strftime("%Y-%m-%d %H:%M:%S"),
        }

    parse_task_array = []
@ -503,7 +503,7 @@ def queue_dataflow(tenant_id:str, flow_id:str, task_id:str, doc_id:str=CANVAS_DE
        to_page=100000000,
        task_type="dataflow" if not rerun else "dataflow_rerun",
        priority=priority,
-        begin_at=datetime.now(),
+        begin_at= datetime.now().strftime("%Y-%m-%d %H:%M:%S"),
    )
    if doc_id not in [CANVAS_DEBUG_DOC_ID, GRAPH_RAPTOR_FAKE_DOC_ID]:
        TaskService.model.delete().where(TaskService.model.doc_id == doc_id).execute()
--- a/api/utils/file_utils.py
+++ b/api/utils/file_utils.py
@ -13,7 +13,12 @@
 #  See the License for the specific language governing permissions and
 #  limitations under the License.
 #
+
+
+# Standard library imports
 import base64
+import hashlib
+import io
 import json
 import os
 import re
@ -22,13 +27,20 @@ import subprocess
 import sys
 import tempfile
 import threading
+import zipfile
 from io import BytesIO

+# Typing
+from typing import List, Union, Tuple
+
+# Third-party imports
+import olefile
 import pdfplumber
 from cachetools import LRUCache, cached
 from PIL import Image
 from ruamel.yaml import YAML

+# Local imports
 from api.constants import IMG_BASE64_PREFIX
 from api.db import FileType

@ -161,7 +173,7 @@ def filename_type(filename):
    if re.match(r".*\.(wav|flac|ape|alac|wavpack|wv|mp3|aac|ogg|vorbis|opus)$", filename):
        return FileType.AURAL.value

-    if re.match(r".*\.(jpg|jpeg|png|tif|gif|pcx|tga|exif|fpx|svg|psd|cdr|pcd|dxf|ufo|eps|ai|raw|WMF|webp|avif|apng|icon|ico|mpg|mpeg|avi|rm|rmvb|mov|wmv|asf|dat|asx|wvx|mpe|mpa|mp4)$", filename):
+    if re.match(r".*\.(jpg|jpeg|png|tif|gif|pcx|tga|exif|fpx|svg|psd|cdr|pcd|dxf|ufo|eps|ai|raw|WMF|webp|avif|apng|icon|ico|mpg|mpeg|avi|rm|rmvb|mov|wmv|asf|dat|asx|wvx|mpe|mpa|mp4|avi|mkv)$", filename):
        return FileType.VISUAL.value

    return FileType.OTHER.value
@ -284,3 +296,125 @@ def read_potential_broken_pdf(blob):
        return repaired

    return blob
+
+
+
+def _is_zip(h: bytes) -> bool:
+    return h.startswith(b"PK\x03\x04") or h.startswith(b"PK\x05\x06") or h.startswith(b"PK\x07\x08")
+
+def _is_pdf(h: bytes) -> bool:
+    return h.startswith(b"%PDF-")
+
+def _is_ole(h: bytes) -> bool:
+    return h.startswith(b"\xD0\xCF\x11\xE0\xA1\xB1\x1A\xE1")
+
+def _sha10(b: bytes) -> str:
+    return hashlib.sha256(b).hexdigest()[:10]
+
+def _guess_ext(b: bytes) -> str:
+    h = b[:8]
+    if _is_zip(h):
+        try:
+            with zipfile.ZipFile(io.BytesIO(b), "r") as z:
+                names = [n.lower() for n in z.namelist()]
+                if any(n.startswith("word/") for n in names):
+                    return ".docx"
+                if any(n.startswith("ppt/") for n in names):
+                    return ".pptx"
+                if any(n.startswith("xl/") for n in names):
+                    return ".xlsx"
+        except Exception:
+            pass
+        return ".zip"
+    if _is_pdf(h):
+        return ".pdf"
+    if _is_ole(h):
+        return ".doc"
+    return ".bin"
+
+# Try to extract the real embedded payload from OLE's Ole10Native
+def _extract_ole10native_payload(data: bytes) -> bytes:
+    try:
+        pos = 0
+        if len(data) < 4:
+            return data
+        _ = int.from_bytes(data[pos:pos+4], "little")
+        pos += 4
+        # filename/src/tmp (NUL-terminated ANSI)
+        for _ in range(3):
+            z = data.index(b"\x00", pos)
+            pos = z + 1
+        # skip unknown 4 bytes
+        pos += 4
+        if pos + 4 > len(data):
+            return data
+        size = int.from_bytes(data[pos:pos+4], "little")
+        pos += 4
+        if pos + size <= len(data):
+            return data[pos:pos+size]
+    except Exception:
+        pass
+    return data
+
+def extract_embed_file(target: Union[bytes, bytearray]) -> List[Tuple[str, bytes]]:
+    """
+    Only extract the 'first layer' of embedding, returning raw (filename, bytes).
+    """
+    top = bytes(target)
+    head = top[:8]
+    out: List[Tuple[str, bytes]] = []
+    seen = set()
+
+    def push(b: bytes, name_hint: str = ""):
+        h10 = _sha10(b)
+        if h10 in seen:
+            return
+        seen.add(h10)
+        ext = _guess_ext(b)
+        # If name_hint has an extension use its basename; else fallback to guessed ext
+        if "." in name_hint:
+            fname = name_hint.split("/")[-1]
+        else:
+            fname = f"{h10}{ext}"
+        out.append((fname, b))
+
+    # OOXML/ZIP container (docx/xlsx/pptx)
+    if _is_zip(head):
+        try:
+            with zipfile.ZipFile(io.BytesIO(top), "r") as z:
+                embed_dirs = (
+                    "word/embeddings/", "word/objects/", "word/activex/",
+                    "xl/embeddings/", "ppt/embeddings/"
+                )
+                for name in z.namelist():
+                    low = name.lower()
+                    if any(low.startswith(d) for d in embed_dirs):
+                        try:
+                            b = z.read(name)
+                            push(b, name)
+                        except Exception:
+                            pass
+        except Exception:
+            pass
+        return out
+
+    # OLE container (doc/ppt/xls)
+    if _is_ole(head):
+        try:
+            with olefile.OleFileIO(io.BytesIO(top)) as ole:
+                for entry in ole.listdir():
+                    p = "/".join(entry)
+                    try:
+                        data = ole.openstream(entry).read()
+                    except Exception:
+                        continue
+                    if not data:
+                        continue
+                    if "Ole10Native" in p or "ole10native" in p.lower():
+                        data = _extract_ole10native_payload(data)
+                    push(data, p)
+        except Exception:
+            pass
+        return out
+
+    return out
--- a/conf/llm_factories.json
+++ b/conf/llm_factories.json
@ -971,31 +971,9 @@
        {
            "name": "VolcEngine",
            "logo": "",
-            "tags": "LLM, TEXT EMBEDDING",
+            "tags": "LLM, TEXT EMBEDDING, IMAGE2TEXT",
            "status": "1",
-            "llm": [
-                {
-                    "llm_name": "Doubao-pro-128k",
-                    "tags": "LLM,CHAT,128k",
-                    "max_tokens": 131072,
-                    "model_type": "chat",
-                    "is_tools": true
-                },
-                {
-                    "llm_name": "Doubao-pro-32k",
-                    "tags": "LLM,CHAT,32k",
-                    "max_tokens": 32768,
-                    "model_type": "chat",
-                    "is_tools": true
-                },
-                {
-                    "llm_name": "Doubao-pro-4k",
-                    "tags": "LLM,CHAT,4k",
-                    "max_tokens": 4096,
-                    "model_type": "chat",
-                    "is_tools": true
-                }
-            ]
+            "llm": []
        },
        {
            "name": "BaiChuan",
@ -1367,35 +1345,35 @@
                    "llm_name": "gemini-2.5-flash",
                    "tags": "LLM,CHAT,1024K,IMAGE2TEXT",
                    "max_tokens": 1048576,
-                    "model_type": "chat",
+                    "model_type": "image2text",
                    "is_tools": true
                },
                {
                    "llm_name": "gemini-2.5-pro",
                    "tags": "LLM,CHAT,IMAGE2TEXT,1024K",
                    "max_tokens": 1048576,
-                    "model_type": "chat",
+                    "model_type": "image2text",
                    "is_tools": true
                },
                {
                    "llm_name": "gemini-2.5-flash-lite",
                    "tags": "LLM,CHAT,1024K,IMAGE2TEXT",
                    "max_tokens": 1048576,
-                    "model_type": "chat",
+                    "model_type": "image2text",
                    "is_tools": true
                },
                {
                    "llm_name": "gemini-2.0-flash",
                    "tags": "LLM,CHAT,1024K",
                    "max_tokens": 1048576,
-                    "model_type": "chat",
+                    "model_type": "image2text",
                    "is_tools": true
                },
                {
                    "llm_name": "gemini-2.0-flash-lite",
                    "tags": "LLM,CHAT,1024K",
                    "max_tokens": 1048576,
-                    "model_type": "chat",
+                    "model_type": "image2text",
                    "is_tools": true
                },
                {
@ -3009,7 +2987,7 @@
                    "tags": "LLM,CHAT,IMAGE2TEXT,32k",
                    "max_tokens": 32000,
                    "model_type": "image2text",
-                    "is_tools": true
+                    "is_tools": false
                },
                {
                    "llm_name": "THUDM/GLM-Z1-32B-0414",
--- a/deepdoc/parser/excel_parser.py
+++ b/deepdoc/parser/excel_parser.py
@ -54,8 +54,8 @@ class RAGFlowExcelParser:
            try:
                file_like_object.seek(0)
                try:
-                    df = pd.read_excel(file_like_object)
-                    return RAGFlowExcelParser._dataframe_to_workbook(df)
+                    dfs = pd.read_excel(file_like_object, sheet_name=None)
+                    return RAGFlowExcelParser._dataframe_to_workbook(dfs)
                except Exception as ex:
                    logging.info(f"pandas with default engine load error: {ex}, try calamine instead")
                    file_like_object.seek(0)
@ -75,6 +75,10 @@ class RAGFlowExcelParser:

    @staticmethod
    def _dataframe_to_workbook(df):
+        # if contains multiple sheets use _dataframes_to_workbook
+        if isinstance(df, dict) and len(df) > 1:
+            return RAGFlowExcelParser._dataframes_to_workbook(df)
+
        df = RAGFlowExcelParser._clean_dataframe(df)
        wb = Workbook()
        ws = wb.active
@ -88,6 +92,22 @@ class RAGFlowExcelParser:
                ws.cell(row=row_num, column=col_num, value=value)

        return wb
+    
+    @staticmethod
+    def _dataframes_to_workbook(dfs: dict):
+        wb = Workbook()
+        default_sheet = wb.active
+        wb.remove(default_sheet)
+        
+        for sheet_name, df in dfs.items():
+            df = RAGFlowExcelParser._clean_dataframe(df)
+            ws = wb.create_sheet(title=sheet_name)
+            for col_num, column_name in enumerate(df.columns, 1):
+                ws.cell(row=1, column=col_num, value=column_name)
+            for row_num, row in enumerate(df.values, 2):
+                for col_num, value in enumerate(row, 1):
+                    ws.cell(row=row_num, column=col_num, value=value)
+        return wb

    def html(self, fnm, chunk_rows=256):
        from html import escape
--- a/deepdoc/parser/figure_parser.py
+++ b/deepdoc/parser/figure_parser.py
@ -17,6 +17,8 @@ from concurrent.futures import ThreadPoolExecutor, as_completed

 from PIL import Image

+from api.db import LLMType
+from api.db.services.llm_service import LLMBundle
 from api.utils.api_utils import timeout
 from rag.app.picture import vision_llm_chunk as picture_vision_llm_chunk
 from rag.prompts.generator import vision_llm_figure_describe_prompt
@ -32,6 +34,43 @@ def vision_figure_parser_figure_data_wrapper(figures_data_without_positions):
        if isinstance(figure_data[1], Image.Image)
    ]

+def vision_figure_parser_docx_wrapper(sections,tbls,callback=None,**kwargs):
+    try:
+        vision_model = LLMBundle(kwargs["tenant_id"], LLMType.IMAGE2TEXT)
+        callback(0.7, "Visual model detected. Attempting to enhance figure extraction...")
+    except Exception:
+        vision_model = None
+    if vision_model:
+        figures_data = vision_figure_parser_figure_data_wrapper(sections)
+        try:
+            docx_vision_parser = VisionFigureParser(vision_model=vision_model, figures_data=figures_data, **kwargs)
+            boosted_figures = docx_vision_parser(callback=callback)
+            tbls.extend(boosted_figures)
+        except Exception as e:
+            callback(0.8, f"Visual model error: {e}. Skipping figure parsing enhancement.")
+    return tbls
+
+def vision_figure_parser_pdf_wrapper(tbls,callback=None,**kwargs):
+    try:
+        vision_model = LLMBundle(kwargs["tenant_id"], LLMType.IMAGE2TEXT)
+        callback(0.7, "Visual model detected. Attempting to enhance figure extraction...")
+    except Exception:
+        vision_model = None
+    if vision_model:
+        def is_figure_item(item):
+            return (
+                isinstance(item[0][0], Image.Image) and
+                isinstance(item[0][1], list)
+            )
+        figures_data = [item for item in tbls if is_figure_item(item)]
+        try:
+            docx_vision_parser = VisionFigureParser(vision_model=vision_model, figures_data=figures_data, **kwargs)
+            boosted_figures = docx_vision_parser(callback=callback)
+            tbls = [item for item in tbls if not is_figure_item(item)]
+            tbls.extend(boosted_figures)
+        except Exception as e:
+            callback(0.8, f"Visual model error: {e}. Skipping figure parsing enhancement.")
+    return tbls

 shared_executor = ThreadPoolExecutor(max_workers=10)

--- a/deepdoc/parser/mineru_parser.py
+++ b/deepdoc/parser/mineru_parser.py
@ -0,0 +1,344 @@
+#
+#  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 json
+import logging
+import platform
+import re
+import subprocess
+import sys
+import tempfile
+import threading
+import time
+from io import BytesIO
+from os import PathLike
+from pathlib import Path
+from queue import Empty, Queue
+from typing import Any, Callable, Optional
+
+import numpy as np
+import pdfplumber
+from PIL import Image
+from strenum import StrEnum
+
+from deepdoc.parser.pdf_parser import RAGFlowPdfParser
+
+LOCK_KEY_pdfplumber = "global_shared_lock_pdfplumber"
+if LOCK_KEY_pdfplumber not in sys.modules:
+    sys.modules[LOCK_KEY_pdfplumber] = threading.Lock()
+
+
+class MinerUContentType(StrEnum):
+    IMAGE = "image"
+    TABLE = "table"
+    TEXT = "text"
+    EQUATION = "equation"
+
+
+class MinerUParser(RAGFlowPdfParser):
+    def __init__(self, mineru_path: str = "mineru"):
+        self.mineru_path = Path(mineru_path)
+        self.logger = logging.getLogger(self.__class__.__name__)
+
+    def check_installation(self) -> bool:
+        subprocess_kwargs = {
+            "capture_output": True,
+            "text": True,
+            "check": True,
+            "encoding": "utf-8",
+            "errors": "ignore",
+        }
+
+        if platform.system() == "Windows":
+            subprocess_kwargs["creationflags"] = getattr(subprocess, "CREATE_NO_WINDOW", 0)
+
+        try:
+            result = subprocess.run([str(self.mineru_path), "--version"], **subprocess_kwargs)
+            version_info = result.stdout.strip()
+            if version_info:
+                logging.info(f"[MinerU] Detected version: {version_info}")
+            else:
+                logging.info("[MinerU] Detected MinerU, but version info is empty.")
+            return True
+        except subprocess.CalledProcessError as e:
+            logging.warning(f"[MinerU] Execution failed (exit code {e.returncode}).")
+        except FileNotFoundError:
+            logging.warning("[MinerU] MinerU not found. Please install it via: pip install -U 'mineru[core]'")
+        except Exception as e:
+            logging.error(f"[MinerU] Unexpected error during installation check: {e}")
+        return False
+
+    def _run_mineru(self, input_path: Path, output_dir: Path, method: str = "auto", lang: Optional[str] = None):
+        cmd = [str(self.mineru_path), "-p", str(input_path), "-o", str(output_dir), "-m", method]
+        if lang:
+            cmd.extend(["-l", lang])
+
+        self.logger.info(f"[MinerU] Running command: {' '.join(cmd)}")
+
+        subprocess_kwargs = {
+            "stdout": subprocess.PIPE,
+            "stderr": subprocess.PIPE,
+            "text": True,
+            "encoding": "utf-8",
+            "errors": "ignore",
+            "bufsize": 1,
+        }
+
+        if platform.system() == "Windows":
+            subprocess_kwargs["creationflags"] = getattr(subprocess, "CREATE_NO_WINDOW", 0)
+
+        process = subprocess.Popen(cmd, **subprocess_kwargs)
+        stdout_queue, stderr_queue = Queue(), Queue()
+
+        def enqueue_output(pipe, queue, prefix):
+            for line in iter(pipe.readline, ""):
+                if line.strip():
+                    queue.put((prefix, line.strip()))
+            pipe.close()
+
+        threading.Thread(target=enqueue_output, args=(process.stdout, stdout_queue, "STDOUT"), daemon=True).start()
+        threading.Thread(target=enqueue_output, args=(process.stderr, stderr_queue, "STDERR"), daemon=True).start()
+
+        while process.poll() is None:
+            for q in (stdout_queue, stderr_queue):
+                try:
+                    while True:
+                        prefix, line = q.get_nowait()
+                        if prefix == "STDOUT":
+                            self.logger.info(f"[MinerU] {line}")
+                        else:
+                            self.logger.warning(f"[MinerU] {line}")
+                except Empty:
+                    pass
+            time.sleep(0.1)
+
+        return_code = process.wait()
+        if return_code != 0:
+            raise RuntimeError(f"[MinerU] Process failed with exit code {return_code}")
+        self.logger.info("[MinerU] Command completed successfully.")
+
+    def __images__(self, fnm, zoomin: int = 1, page_from=0, page_to=600, callback=None):
+        self.page_from = page_from
+        self.page_to = page_to
+        try:
+            with pdfplumber.open(fnm) if isinstance(fnm, (str, PathLike)) else pdfplumber.open(BytesIO(fnm)) as pdf:
+                self.pdf = pdf
+                self.page_images = [p.to_image(resolution=72 * zoomin, antialias=True).original for _, p in enumerate(self.pdf.pages[page_from:page_to])]
+        except Exception as e:
+            self.page_images = None
+            self.total_page = 0
+            logging.exception(e)
+
+    def _line_tag(self, bx):
+        pn = [bx["page_idx"] + 1]
+        positions = bx["bbox"]
+        x0, top, x1, bott = positions
+
+        if hasattr(self, "page_images") and self.page_images and len(self.page_images) > bx["page_idx"]:
+            page_width, page_height = self.page_images[bx["page_idx"]].size
+            x0 = (x0 / 1000.0) * page_width
+            x1 = (x1 / 1000.0) * page_width
+            top = (top / 1000.0) * page_height
+            bott = (bott / 1000.0) * page_height
+
+        return "@@{}\t{:.1f}\t{:.1f}\t{:.1f}\t{:.1f}##".format("-".join([str(p) for p in pn]), x0, x1, top, bott)
+
+    def crop(self, text, ZM=1, need_position=False):
+        imgs = []
+        poss = self.extract_positions(text)
+        if not poss:
+            if need_position:
+                return None, None
+            return
+
+        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)))
+        pos = poss[-1]
+        poss.append(([pos[0][-1]], pos[1], pos[2], min(self.page_images[pos[0][-1]].size[1], pos[4] + GAP), min(self.page_images[pos[0][-1]].size[1], pos[4] + 120)))
+
+        positions = []
+        for ii, (pns, left, right, top, bottom) in enumerate(poss):
+            right = left + max_width
+
+            if bottom <= top:
+                bottom = top + 2
+
+            for pn in pns[1:]:
+                bottom += self.page_images[pn - 1].size[1]
+
+            img0 = self.page_images[pns[0]]
+            x0, y0, x1, y1 = int(left), int(top), int(right), int(min(bottom, img0.size[1]))
+            crop0 = img0.crop((x0, y0, x1, y1))
+            imgs.append(crop0)
+            if 0 < ii < len(poss) - 1:
+                positions.append((pns[0] + self.page_from, x0, x1, y0, y1))
+
+            bottom -= img0.size[1]
+            for pn in pns[1:]:
+                page = self.page_images[pn]
+                x0, y0, x1, y1 = int(left), 0, int(right), int(min(bottom, page.size[1]))
+                cimgp = page.crop((x0, y0, x1, y1))
+                imgs.append(cimgp)
+                if 0 < ii < len(poss) - 1:
+                    positions.append((pn + self.page_from, x0, x1, y0, y1))
+                bottom -= page.size[1]
+
+        if not imgs:
+            if need_position:
+                return None, None
+            return
+
+        height = 0
+        for img in imgs:
+            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))
+        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))
+                overlay.putalpha(128)
+                img = Image.alpha_composite(img, overlay).convert("RGB")
+            pic.paste(img, (0, int(height)))
+            height += img.size[1] + GAP
+
+        if need_position:
+            return pic, positions
+        return pic
+
+    @staticmethod
+    def extract_positions(txt: str):
+        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))
+        return poss
+
+    def _read_output(self, output_dir: Path, file_stem: str, method: str = "auto") -> list[dict[str, Any]]:
+        subdir = output_dir / file_stem / method
+        json_file = subdir / f"{file_stem}_content_list.json"
+
+        if not json_file.exists():
+            raise FileNotFoundError(f"[MinerU] Missing output file: {json_file}")
+
+        with open(json_file, "r", encoding="utf-8") as f:
+            data = json.load(f)
+
+        for item in data:
+            for key in ("img_path", "table_img_path", "equation_img_path"):
+                if key in item and item[key]:
+                    item[key] = str((subdir / item[key]).resolve())
+        return data
+
+    def _transfer_to_sections(self, outputs: list[dict[str, Any]]):
+        sections = []
+        for output in outputs:
+            match output["type"]:
+                case MinerUContentType.TEXT:
+                    section = output["text"]
+                case MinerUContentType.TABLE:
+                    section = output["table_body"] + "\n".join(output["table_caption"]) + "\n".join(output["table_footnote"])
+                case MinerUContentType.IMAGE:
+                    section = "".join(output["image_caption"]) + "\n" + "".join(output["image_footnote"])
+                case MinerUContentType.EQUATION:
+                    section = output["text"]
+
+            if section:
+                sections.append((section, self._line_tag(output)))
+        return sections
+
+    def _transfer_to_tables(self, outputs: list[dict[str, Any]]):
+        return []
+
+    def parse_pdf(
+        self,
+        filepath: str | PathLike[str],
+        binary: BytesIO | bytes,
+        callback: Optional[Callable] = None,
+        *,
+        output_dir: Optional[str] = None,
+        lang: Optional[str] = None,
+        method: str = "auto",
+        delete_output: bool = True,
+    ) -> tuple:
+        import shutil
+
+        temp_pdf = None
+        created_tmp_dir = False
+
+        if binary:
+            temp_dir = Path(tempfile.mkdtemp(prefix="mineru_bin_pdf_"))
+            temp_pdf = temp_dir / Path(filepath).name
+            with open(temp_pdf, "wb") as f:
+                f.write(binary)
+            pdf = temp_pdf
+            self.logger.info(f"[MinerU] Received binary PDF -> {temp_pdf}")
+            if callback:
+                callback(0.15, f"[MinerU] Received binary PDF -> {temp_pdf}")
+        else:
+            pdf = Path(filepath)
+            if not pdf.exists():
+                if callback:
+                    callback(-1, f"[MinerU] PDF not found: {pdf}")
+                raise FileNotFoundError(f"[MinerU] PDF not found: {pdf}")
+
+        if output_dir:
+            out_dir = Path(output_dir)
+            out_dir.mkdir(parents=True, exist_ok=True)
+        else:
+            out_dir = Path(tempfile.mkdtemp(prefix="mineru_pdf_"))
+            created_tmp_dir = True
+
+        self.logger.info(f"[MinerU] Output directory: {out_dir}")
+        if callback:
+            callback(0.15, f"[MinerU] Output directory: {out_dir}")
+
+        self.__images__(pdf, zoomin=1)
+
+        try:
+            self._run_mineru(pdf, out_dir, method=method, lang=lang)
+            outputs = self._read_output(out_dir, pdf.stem, method=method)
+            self.logger.info(f"[MinerU] Parsed {len(outputs)} blocks from PDF.")
+            if callback:
+                callback(0.75, f"[MinerU] Parsed {len(outputs)} blocks from PDF.")
+            return self._transfer_to_sections(outputs), self._transfer_to_tables(outputs)
+        finally:
+            if temp_pdf and temp_pdf.exists():
+                try:
+                    temp_pdf.unlink()
+                    temp_pdf.parent.rmdir()
+                except Exception:
+                    pass
+            if delete_output and created_tmp_dir and out_dir.exists():
+                try:
+                    shutil.rmtree(out_dir)
+                except Exception:
+                    pass
+
+
+if __name__ == "__main__":
+    parser = MinerUParser("mineru")
+    print("MinerU available:", parser.check_installation())
+
+    filepath = ""
+    with open(filepath, "rb") as file:
+        outputs = parser.parse_pdf(filepath=filepath, binary=file.read())
+        for output in outputs:
+            print(output)
--- a/docker/.env
+++ b/docker/.env
@ -97,13 +97,13 @@ SVR_HTTP_PORT=9380
 ADMIN_SVR_HTTP_PORT=9381

 # The RAGFlow Docker image to download.
-# Defaults to the v0.21.0-slim edition, which is the RAGFlow Docker image without embedding models.
-RAGFLOW_IMAGE=infiniflow/ragflow:v0.21.0-slim
+# Defaults to the v0.21.1-slim edition, which is the RAGFlow Docker image without embedding models.
+RAGFLOW_IMAGE=infiniflow/ragflow:v0.21.1-slim
 #
 # To download the RAGFlow Docker image with embedding models, uncomment the following line instead:
-# RAGFLOW_IMAGE=infiniflow/ragflow:v0.21.0
+# RAGFLOW_IMAGE=infiniflow/ragflow:v0.21.1
 #
-# The Docker image of the v0.21.0 edition includes built-in embedding models:
+# The Docker image of the v0.21.1 edition includes built-in embedding models:
 #   - BAAI/bge-large-zh-v1.5
 #   - maidalun1020/bce-embedding-base_v1
 #
--- a/docker/README.md
+++ b/docker/README.md
@ -79,8 +79,8 @@ The [.env](./.env) file contains important environment variables for Docker.
 - `RAGFLOW-IMAGE`  
  The Docker image edition. Available editions:  
  
-  - `infiniflow/ragflow:v0.21.0-slim` (default): The RAGFlow Docker image without embedding models.  
-  - `infiniflow/ragflow:v0.21.0`: The RAGFlow Docker image with embedding models including:
+  - `infiniflow/ragflow:v0.21.1-slim` (default): The RAGFlow Docker image without embedding models.  
+  - `infiniflow/ragflow:v0.21.1`: The RAGFlow Docker image with embedding models including:
    - Built-in embedding models:
      - `BAAI/bge-large-zh-v1.5` 
      - `maidalun1020/bce-embedding-base_v1`
--- a/docker/docker-compose-base.yml
+++ b/docker/docker-compose-base.yml
@ -77,7 +77,7 @@ services:
    container_name: ragflow-infinity
    profiles:
      - infinity
-    image: infiniflow/infinity:v0.6.0
+    image: infiniflow/infinity:v0.6.1
    volumes:
      - infinity_data:/var/infinity
      - ./infinity_conf.toml:/infinity_conf.toml
--- a/docker/infinity_conf.toml
+++ b/docker/infinity_conf.toml
@ -1,5 +1,5 @@
 [general]
-version                  = "0.6.0"
+version                  = "0.6.1"
 time_zone                = "utc-8"

 [network]
--- a/docs/configurations.md
+++ b/docs/configurations.md
@ -99,8 +99,8 @@ RAGFlow utilizes MinIO as its object storage solution, leveraging its scalabilit
 - `RAGFLOW-IMAGE`  
  The Docker image edition. Available editions:  
  
-  - `infiniflow/ragflow:v0.21.0-slim` (default): The RAGFlow Docker image without embedding models.  
-  - `infiniflow/ragflow:v0.21.0`: The RAGFlow Docker image with embedding models including:
+  - `infiniflow/ragflow:v0.21.1-slim` (default): The RAGFlow Docker image without embedding models.  
+  - `infiniflow/ragflow:v0.21.1`: The RAGFlow Docker image with embedding models including:
    - Built-in embedding models:
      - `BAAI/bge-large-zh-v1.5` 
      - `maidalun1020/bce-embedding-base_v1`
--- a/docs/develop/build_docker_image.mdx
+++ b/docs/develop/build_docker_image.mdx
@ -77,7 +77,7 @@ After building the infiniflow/ragflow:nightly-slim image, you are ready to launc

 1. Edit Docker Compose Configuration

-Open the `docker/.env` file. Find the `RAGFLOW_IMAGE` setting and change the image reference from `infiniflow/ragflow:v0.21.0-slim` to `infiniflow/ragflow:nightly-slim` to use the pre-built image.
+Open the `docker/.env` file. Find the `RAGFLOW_IMAGE` setting and change the image reference from `infiniflow/ragflow:v0.21.1-slim` to `infiniflow/ragflow:nightly-slim` to use the pre-built image.


 2. Launch the Service
--- a/docs/faq.mdx
+++ b/docs/faq.mdx
@ -30,17 +30,17 @@ The "garbage in garbage out" status quo remains unchanged despite the fact that

 Each RAGFlow release is available in two editions:

- **Slim edition**: excludes built-in embedding models and is identified by a **-slim** suffix added to the version name. Example: `infiniflow/ragflow:v0.21.0-slim`
- **Full edition**: includes built-in embedding models and has no suffix added to the version name. Example: `infiniflow/ragflow:v0.21.0`
+- **Slim edition**: excludes built-in embedding models and is identified by a **-slim** suffix added to the version name. Example: `infiniflow/ragflow:v0.21.1-slim`
+- **Full edition**: includes built-in embedding models and has no suffix added to the version name. Example: `infiniflow/ragflow:v0.21.1`

 ---

 ### Which embedding models can be deployed locally?

-RAGFlow offers two Docker image editions, `v0.21.0-slim` and `v0.21.0`:  
+RAGFlow offers two Docker image editions, `v0.21.1-slim` and `v0.21.1`:  
  
- `infiniflow/ragflow:v0.21.0-slim` (default): The RAGFlow Docker image without embedding models.  
- `infiniflow/ragflow:v0.21.0`: The RAGFlow Docker image with the following built-in embedding models:
+- `infiniflow/ragflow:v0.21.1-slim` (default): The RAGFlow Docker image without embedding models.  
+- `infiniflow/ragflow:v0.21.1`: The RAGFlow Docker image with the following built-in embedding models:
  - `BAAI/bge-large-zh-v1.5`
  - `maidalun1020/bce-embedding-base_v1`

@ -510,3 +510,27 @@ See [here](./guides/agent/best_practices/accelerate_agent_question_answering.md)

 ---

+### How to use MinerU to parse PDF documents?
+
+MinerU PDF document parsing is available starting from v0.21.1. To use this feature, follow these steps:
+
+1. Before deploying ragflow-server, update your **docker/.env** file:  
+   - Enable `HF_ENDPOINT=https://hf-mirror.com`
+   - Add a MinerU entry: `MINERU_EXECUTABLE=/ragflow/uv_tools/.venv/bin/mineru`
+
+2. Start the ragflow-server and run the following commands inside the container:  
+
+```bash
+mkdir uv_tools
+cd uv_tools
+uv venv .venv
+source .venv/bin/activate
+uv pip install -U "mineru[core]" -i https://mirrors.aliyun.com/pypi/simple
+```
+
+3. Restart the ragflow-server.
+4. In the web UI, navigate to the **Configuration** page of your dataset. Click **Built-in** in the **Ingestion pipeline** section, select a chunking method from the **Built-in** dropdown, which supports PDF parsing, and slect **MinerU** in **PDF parser**.
+5. If you use a custom ingestion pipeline instead, you must also complete the first three steps before selecting **MinerU** in the **Parsing method** section of the **Parser** component.
+
+
+
--- a/docs/guides/agent/agent_component_reference/agent.mdx
+++ b/docs/guides/agent/agent_component_reference/agent.mdx
@ -9,7 +9,7 @@ The component equipped with reasoning, tool usage, and multi-agent collaboration

 ---

-An **Agent** component fine-tunes the LLM and sets its prompt. From v0.21.0 onwards, an **Agent** component is able to work independently and with the following capabilities:
+An **Agent** component fine-tunes the LLM and sets its prompt. From v0.20.5 onwards, an **Agent** component is able to work independently and with the following capabilities:

 - Autonomous reasoning with reflection and adjustment based on environmental feedback.
 - Use of tools or subagents to complete tasks.
@ -24,7 +24,7 @@ An **Agent** component is essential when you need the LLM to assist with summari

 ![Set default models](https://raw.githubusercontent.com/infiniflow/ragflow-docs/main/images/set_default_models.jpg)

-2. If your Agent involves dataset retrieval, ensure you [have properly configured your target knowledge base(s)](../../dataset/configure_knowledge_base.md).
+2. If your Agent involves dataset retrieval, ensure you [have properly configured your target dataset(s)](../../dataset/configure_knowledge_base.md).

 ## Quickstart

@ -113,7 +113,7 @@ Click the dropdown menu of **Model** to show the model configuration window.
 - **Model**: The chat model to use.  
  - Ensure you set the chat model correctly on the **Model providers** page.
  - You can use different models for different components to increase flexibility or improve overall performance.
- **Freedom**: A shortcut to **Temperature**, **Top P**, **Presence penalty**, and **Frequency penalty** settings, indicating the freedom level of the model. From **Improvise**, **Precise**, to **Balance**, each preset configuration corresponds to a unique combination of **Temperature**, **Top P**, **Presence penalty**, and **Frequency penalty**.   
+- **Creavity**: A shortcut to **Temperature**, **Top P**, **Presence penalty**, and **Frequency penalty** settings, indicating the freedom level of the model. From **Improvise**, **Precise**, to **Balance**, each preset configuration corresponds to a unique combination of **Temperature**, **Top P**, **Presence penalty**, and **Frequency penalty**.   
  This parameter has three options:
  - **Improvise**: Produces more creative responses.
  - **Precise**: (Default) Produces more conservative responses.
@ -132,11 +132,12 @@ Click the dropdown menu of **Model** to show the model configuration window.
 - **Frequency penalty**: Discourages the model from repeating the same words or phrases too frequently in the generated text.  
  - A higher **frequency penalty** value results in the model being more conservative in its use of repeated tokens.
  - Defaults to 0.7.
- **Max tokens**: 
+- **Max tokens**:  
+  This sets the maximum length of the model's output, measured in the number of tokens (words or pieces of words). It is disabled by default, allowing the model to determine the number of tokens in its responses.

 :::tip NOTE
 - It is not necessary to stick with the same model for all components. If a specific model is not performing well for a particular task, consider using a different one.
- If you are uncertain about the mechanism behind **Temperature**, **Top P**, **Presence penalty**, and **Frequency penalty**, simply choose one of the three options of **Preset configurations**.
+- If you are uncertain about the mechanism behind **Temperature**, **Top P**, **Presence penalty**, and **Frequency penalty**, simply choose one of the three options of **Creavity**.
 :::

 ### System prompt
@ -147,7 +148,7 @@ An **Agent** component relies on keys (variables) to specify its data inputs. It

 #### Advanced usage

-From v0.21.0 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.
+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.
--- a/docs/guides/agent/agent_component_reference/categorize.mdx
+++ b/docs/guides/agent/agent_component_reference/categorize.mdx
@ -42,7 +42,7 @@ Click the dropdown menu of **Model** to show the model configuration window.
 - **Model**: The chat model to use.  
  - Ensure you set the chat model correctly on the **Model providers** page.
  - You can use different models for different components to increase flexibility or improve overall performance.
- **Freedom**: A shortcut to **Temperature**, **Top P**, **Presence penalty**, and **Frequency penalty** settings, indicating the freedom level of the model. From **Improvise**, **Precise**, to **Balance**, each preset configuration corresponds to a unique combination of **Temperature**, **Top P**, **Presence penalty**, and **Frequency penalty**.  
+- **Creavity**: A shortcut to **Temperature**, **Top P**, **Presence penalty**, and **Frequency penalty** settings, indicating the freedom level of the model. From **Improvise**, **Precise**, to **Balance**, each preset configuration corresponds to a unique combination of **Temperature**, **Top P**, **Presence penalty**, and **Frequency penalty**.  
  This parameter has three options:  
  - **Improvise**: Produces more creative responses.
  - **Precise**: (Default) Produces more conservative responses.
@ -61,10 +61,12 @@ Click the dropdown menu of **Model** to show the model configuration window.
 - **Frequency penalty**: Discourages the model from repeating the same words or phrases too frequently in the generated text.  
  - A higher **frequency penalty** value results in the model being more conservative in its use of repeated tokens.
  - Defaults to 0.7.
+- **Max tokens**:  
+  This sets the maximum length of the model's output, measured in the number of tokens (words or pieces of words). It is disabled by default, allowing the model to determine the number of tokens in its responses.

 :::tip NOTE
 - It is not necessary to stick with the same model for all components. If a specific model is not performing well for a particular task, consider using a different one.
- If you are uncertain about the mechanism behind **Temperature**, **Top P**, **Presence penalty**, and **Frequency penalty**, simply choose one of the three options of **Preset configurations**.
+- If you are uncertain about the mechanism behind **Temperature**, **Top P**, **Presence penalty**, and **Frequency penalty**, simply choose one of the three options of **Creavity**.
 :::

 ### Message window size
--- a/docs/guides/agent/agent_component_reference/chunker_title.md
+++ b/docs/guides/agent/agent_component_reference/chunker_title.md
@ -0,0 +1,40 @@
+---
+sidebar_position: 31
+slug: /chunker_title_component
+---
+
+# Title chunker component
+
+A component that splits texts into chunks by heading level.
+
+---
+
+A **Token chunker** component is a text splitter that uses specified heading level as delimiter to define chunk boundaries and create chunks.
+
+## Scenario
+
+A **Title chunker** component is optional, usually placed immediately after **Parser**.
+
+:::caution WARNING
+Placing a **Title chunker** after a **Token chunker** is invalid and will cause an error. Please note that this restriction is not currently system-enforced and requires your attention.
+:::
+
+## Configurations
+
+### Hierarchy
+
+Specifies the heading level to define chunk boundaries: 
+
+- H1
+- H2
+- H3 (Default)
+- H4
+
+Click **+ Add** to add heading levels here or update the corresponding **Regular Expressions** fields for custom heading patterns.
+
+### Output
+
+The global variable name for the output of the **Title chunker** component, which can be referenced by subsequent components in the ingestion pipeline.
+
+- Default: `chunks`
+- Type: `Array<Object>`
--- a/docs/guides/agent/agent_component_reference/chunker_token.md
+++ b/docs/guides/agent/agent_component_reference/chunker_token.md
@ -0,0 +1,43 @@
+---
+sidebar_position: 32
+slug: /chunker_token_component
+---
+
+# Token chunker component
+
+A component that splits texts into chunks, respecting a maximum token limit and using delimiters to find optimal breakpoints.
+
+---
+
+A **Token chunker** component is a text splitter that creates chunks by respecting a recommended maximum token length, using delimiters to ensure logical chunk breakpoints. It splits long texts into appropriately-sized, semantically related chunks.
+
+
+## Scenario
+
+A **Token chunker** component is optional, usually placed immediately after **Parser** or **Title chunker**.
+
+## Configurations
+
+### Recommended chunk size
+
+The recommended maximum token limit for each created chunk. The **Token chunker** component creates chunks at specified delimiters. If this token limit is reached before a delimiter, a chunk is created at that point.
+
+### Overlapped percent (%)
+
+This defines the overlap percentage between chunks. An appropriate degree of overlap ensures semantic coherence without creating excessive, redundant tokens for the LLM.
+
+- Default: 0
+- Maximum: 30%
+
+
+### Delimiters
+
+Defaults to `\n`. Click the right-hand **Recycle bin** button to remove it, or click **+ Add** to add a delimiter.
+
+
+### Output
+
+The global variable name for the output of the **Token chunker** component, which can be referenced by subsequent components in the ingestion pipeline.
+
+- Default: `chunks`
+- Type: `Array<Object>`
--- a/docs/guides/agent/agent_component_reference/indexer.md
+++ b/docs/guides/agent/agent_component_reference/indexer.md
@ -0,0 +1,29 @@
+---
+sidebar_position: 40
+slug: /indexer_component
+---
+
+# Indexer component
+
+A component that defines how chunks are indexed.
+
+---
+
+An **Indexer** component indexes chunks and configures their storage formats in the document engine.
+
+## Scenario
+
+An **Indexer** component is the mandatory ending component for all ingestion pipelines.
+
+## Configurations
+
+### Search method
+
+This setting configures how chunks are stored in the document engine: as full-text, embeddings, or both.
+
+### Filename embedding weight
+
+This setting defines the filename's contribution to the final embedding, which is a weighted combination of both the chunk content and the filename. Essentially, a higher value gives the filename more influence in the final *composite* embedding.
+
+- 0.1: Filename contributes 10% (chunk content 90%)
+- 0.5 (maximum): Filename contributes 50% (chunk content 90%)
--- a/docs/guides/agent/agent_component_reference/parser.md
+++ b/docs/guides/agent/agent_component_reference/parser.md
@ -0,0 +1,17 @@
+---
+sidebar_position: 30
+slug: /parser_component
+---
+
+# Parser component
+
+A component that sets the parsing rules for your dataset.
+
+---
+
+A **Parser** component defines how various file types should be parsed, including parsing methods for PDFs , fields to parse for Emails, and OCR methods for images.
+
+
+## Scenario
+
+A **Parser** component is auto-populated on the ingestion pipeline canvas and required in all ingestion pipeline workflows.
--- a/docs/guides/agent/agent_component_reference/retrieval.mdx
+++ b/docs/guides/agent/agent_component_reference/retrieval.mdx
@ -87,9 +87,9 @@ RAGFlow employs a combination of weighted keyword similarity and weighted vector

 Defaults to 0.2.

-### Keyword similarity weight
+### Vector similarity weight

-This parameter sets the weight of keyword similarity in the combined similarity score. The total of the two weights must equal 1.0. Its default value is 0.7, which means the weight of vector similarity in the combined search is 1 - 0.7 = 0.3.
+This parameter sets the weight of vector similarity in the composite similarity score. The total of the two weights must equal 1.0. Its default value is 0.3, which means the weight of keyword similarity in a combined search is 1 - 0.3 = 0.7.

 ### Top N

--- a/docs/guides/agent/agent_component_reference/transformer.md
+++ b/docs/guides/agent/agent_component_reference/transformer.md
@ -0,0 +1,80 @@
+---
+sidebar_position: 37
+slug: /transformer_component
+---
+
+# Transformer component
+
+A component that uses an LLM to extract insights from the chunks.
+
+---
+
+A **Transformer** component indexes chunks and configures their storage formats in the document engine. It *typically* precedes the **Indexer** in the ingestion pipeline, but you can also chain multiple **Transformer** components in sequence.
+
+## Scenario
+
+A **Transformer** component is essential when you need the LLM to extract new information, such as keywords, questions, metadata, and summaries, from the original chunks. 
+
+## Configurations
+
+### Model
+
+Click the dropdown menu of **Model** to show the model configuration window.
+
+- **Model**: The chat model to use.  
+  - Ensure you set the chat model correctly on the **Model providers** page.
+  - You can use different models for different components to increase flexibility or improve overall performance.
+- **Creavity**: A shortcut to **Temperature**, **Top P**, **Presence penalty**, and **Frequency penalty** settings, indicating the freedom level of the model. From **Improvise**, **Precise**, to **Balance**, each preset configuration corresponds to a unique combination of **Temperature**, **Top P**, **Presence penalty**, and **Frequency penalty**.   
+  This parameter has three options:
+  - **Improvise**: Produces more creative responses.
+  - **Precise**: (Default) Produces more conservative responses.
+  - **Balance**: A middle ground between **Improvise** and **Precise**.
+- **Temperature**: The randomness level of the model's output.  
+  Defaults to 0.1.
+  - Lower values lead to more deterministic and predictable outputs.
+  - Higher values lead to more creative and varied outputs.
+  - A temperature of zero results in the same output for the same prompt.
+- **Top P**: Nucleus sampling.  
+  - Reduces the likelihood of generating repetitive or unnatural text by setting a threshold *P* and restricting the sampling to tokens with a cumulative probability exceeding *P*.
+  - Defaults to 0.3.
+- **Presence penalty**: Encourages the model to include a more diverse range of tokens in the response.  
+  - A higher **presence penalty** value results in the model being more likely to generate tokens not yet been included in the generated text.
+  - Defaults to 0.4.
+- **Frequency penalty**: Discourages the model from repeating the same words or phrases too frequently in the generated text.  
+  - A higher **frequency penalty** value results in the model being more conservative in its use of repeated tokens.
+  - Defaults to 0.7.
+- **Max tokens**:  
+  This sets the maximum length of the model's output, measured in the number of tokens (words or pieces of words). It is disabled by default, allowing the model to determine the number of tokens in its responses.
+
+:::tip NOTE
+- It is not necessary to stick with the same model for all components. If a specific model is not performing well for a particular task, consider using a different one.
+- If you are uncertain about the mechanism behind **Temperature**, **Top P**, **Presence penalty**, and **Frequency penalty**, simply choose one of the three options of **Creativity**.
+:::
+
+### Result destination
+
+Select the type of output to be generated by the LLM:
+
+- Summary
+- Keywords
+- Questions
+- Metadata
+
+### System prompt
+
+Typically, you use the system prompt to describe the task for the LLM, specify how it should respond, and outline other miscellaneous requirements. We do not plan to elaborate on this topic, as it can be as extensive as prompt engineering. 
+
+:::tip NOTE
+The system prompt here automatically updates to match your selected **Result destination**.
+:::
+
+### User prompt
+
+The user-defined prompt. For example, you can type `/` or click **(x)** to insert variables of preceding components in the ingestion pipeline as the LLM's input.
+
+### Output
+
+The global variable name for the output of the **Transformer** component, which can be referenced by subsequent **Transformer** components in the ingestion pipeline.
+
+- Default: `chunks`
+- Type: `Array<Object>`
--- a/docs/guides/chat/start_chat.md
+++ b/docs/guides/chat/start_chat.md
@ -19,7 +19,7 @@ You start an AI conversation by creating an assistant.

   > RAGFlow offers you 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**:
+2. Update Assistant-specific 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**:
@ -28,12 +28,12 @@ You start an AI conversation by creating an assistant.
   - **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 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**:
+3. Update Prompt-specific settings:

   - In **System**, you fill in the prompts for your LLM, you can also leave the default prompt as-is for the beginning.
   - **Similarity threshold** sets the similarity "bar" for each chunk of text. The default is 0.2. Text chunks with lower similarity scores are filtered out of the final response.
-   - **Keyword similarity weight** is set to 0.7 by default. RAGFlow uses a hybrid score system to evaluate the relevance of different text chunks. This value sets the weight assigned to the keyword similarity component in the hybrid score.
-     - 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.
+   - **Vector similarity weight** is set to 0.3 by default. RAGFlow uses a hybrid score system to evaluate the relevance of different text chunks. This value sets the weight assigned to the vector similarity component in the hybrid score.
+     - If **Rerank model** is left empty, the hybrid score system uses keyword similarity and vector similarity, and the default weight assigned to the keyword similarity component is 1-0.3=0.7.
     - 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.
@ -48,14 +48,14 @@ You start an AI conversation by creating an assistant.
     - 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*.
-      - As of v0.21.0, if you add custom variables here, the only way you can pass in their values is to call:
+      - As of v0.21.1, if you add custom variables here, the only way you can pass in their values is to call:
         - HTTP method [Converse with chat assistant](../../references/http_api_reference.md#converse-with-chat-assistant), or
         - Python method [Converse with chat assistant](../../references/python_api_reference.md#converse-with-chat-assistant).

-4. Update **Model Setting**:
+4. Update Model-specific Settings:

   - In **Model**: you select the chat model. Though you have selected the default chat model in **System Model Settings**, RAGFlow allows you to choose an alternative chat model for your dialogue.
-   - **Freedom**: A shortcut to **Temperature**, **Top P**, **Presence penalty**, and **Frequency penalty** settings, indicating the freedom level of the model. From **Improvise**, **Precise**, to **Balance**, each preset configuration corresponds to a unique combination of **Temperature**, **Top P**, **Presence penalty**, and **Frequency penalty**.   
+   - **Creavity**: A shortcut to **Temperature**, **Top P**, **Presence penalty**, and **Frequency penalty** settings, indicating the freedom level of the model. From **Improvise**, **Precise**, to **Balance**, each preset configuration corresponds to a unique combination of **Temperature**, **Top P**, **Presence penalty**, and **Frequency penalty**.   
   This parameter has three options:
      - **Improvise**: Produces more creative responses.
      - **Precise**: (Default) Produces more conservative responses.
--- a/docs/guides/dataset/configure_knowledge_base.md
+++ b/docs/guides/dataset/configure_knowledge_base.md
@ -59,7 +59,7 @@ You can also change a file's chunking method on the **Files** page.
 ![change chunking method](https://raw.githubusercontent.com/infiniflow/ragflow-docs/main/images/change_chunking_method.jpg)

 <details>
-  <summary>From v0.21.0 onward, RAGFlow supports ingestion pipeline for customized data ingestion and cleansing workflows.</summary>
+  <summary>From v0.21.1 onward, RAGFlow supports ingestion pipeline for customized data ingestion and cleansing workflows.</summary>
   
  To use a customized data pipeline:

@ -138,7 +138,7 @@ See [Run retrieval test](./run_retrieval_test.md) for details.

 ## Search for dataset

-As of RAGFlow v0.21.0, the search feature is still in a rudimentary form, supporting only dataset search by name.
+As of RAGFlow v0.21.1, the search feature is still in a rudimentary form, supporting only dataset search by name.

 ![search dataset](https://raw.githubusercontent.com/infiniflow/ragflow-docs/main/images/search_datasets.jpg)

--- a/docs/guides/dataset/run_retrieval_test.md
+++ b/docs/guides/dataset/run_retrieval_test.md
@ -29,9 +29,9 @@ In contrast, chunks created from [knowledge graph construction](./construct_know

 This sets the bar for retrieving chunks: chunks with similarities below the threshold will be filtered out. By default, the threshold is set to 0.2. This means that only chunks with hybrid similarity score of 20 or higher will be retrieved.

-### Keyword similarity weight
+### Vector similarity weight

-This sets the weight of keyword similarity in the combined similarity score, whether used with vector cosine similarity or a reranking score. By default, it is set to 0.7, making the weight of the other component 0.3 (1 - 0.7).
+This sets the weight of vector similarity in the composite similarity score, whether used with vector cosine similarity or a reranking score. By default, it is set to 0.3, making the weight of the other component 0.7 (1 - 0.3).

 ### Rerank model

--- a/docs/guides/dataset/select_pdf_parser.md
+++ b/docs/guides/dataset/select_pdf_parser.md
@ -35,8 +35,31 @@ RAGFlow isn't one-size-fits-all. It is built for flexibility and supports deeper

  - DeepDoc: (Default) The default visual model performing OCR, TSR, and DLR tasks on PDFs, which can be time-consuming.
  - Naive: Skip OCR, TSR, and DLR tasks if *all* your PDFs are plain text.
+  - MinerU: An experimental feature.
  - A third-party visual model provided by a specific model provider.

+:::danger IMPORTANG
+MinerU PDF document parsing is available starting from v0.21.1. To use this feature, follow these steps:
+
+1. Before deploying ragflow-server, update your **docker/.env** file:  
+   - Enable `HF_ENDPOINT=https://hf-mirror.com`
+   - Add a MinerU entry: `MINERU_EXECUTABLE=/ragflow/uv_tools/.venv/bin/mineru`
+
+2. Start the ragflow-server and run the following commands inside the container:  
+
+```bash
+mkdir uv_tools
+cd uv_tools
+uv venv .venv
+source .venv/bin/activate
+uv pip install -U "mineru[core]" -i https://mirrors.aliyun.com/pypi/simple
+```
+
+3. Restart the ragflow-server.
+4. In the web UI, navigate to the **Configuration** page of your dataset. Click **Built-in** in the **Ingestion pipeline** section, select a chunking method from the **Built-in** dropdown, which supports PDF parsing, and slect **MinerU** in **PDF parser**.
+5. If you use a custom ingestion pipeline instead, you must also complete the first three steps before selecting **MinerU** in the **Parsing method** section of the **Parser** component.
+:::
+
 :::caution WARNING
 Third-party visual models are marked **Experimental**, because we have not fully tested these models for the aforementioned data extraction tasks.
 :::
--- a/docs/guides/manage_files.md
+++ b/docs/guides/manage_files.md
@ -87,4 +87,4 @@ RAGFlow's file management allows you to download an uploaded file:

 ![download_file](https://github.com/infiniflow/ragflow/assets/93570324/cf3b297f-7d9b-4522-bf5f-4f45743e4ed5)

-> As of RAGFlow v0.21.0, bulk download is not supported, nor can you download an entire folder. 
+> As of RAGFlow v0.21.1, bulk download is not supported, nor can you download an entire folder. 
--- a/docs/guides/manage_users_and_services.md
+++ b/docs/guides/manage_users_and_services.md
@ -46,7 +46,7 @@ The Admin CLI and Admin Service form a client-server architectural suite for RAG
 2. Install ragflow-cli.

   ```bash
-   pip install ragflow-cli==0.21.0
+   pip install ragflow-cli==0.21.1
   ```

 3. Launch the CLI client:
@ -348,7 +348,7 @@ Listing all agents of user: lynn_inf@hotmail.com
 +-----------------+-------------+------------+-----------------+
 | canvas_category | canvas_type | permission | title           |
 +-----------------+-------------+------------+-----------------+
-| agent_canvas    | None        | team       | research_helper |
+| agent           | None        | team       | research_helper |
 +-----------------+-------------+------------+-----------------+
 ```

--- a/docs/guides/tracing.mdx
+++ b/docs/guides/tracing.mdx
@ -18,7 +18,7 @@ RAGFlow ships with a built-in [Langfuse](https://langfuse.com) integration so th
 Langfuse stores traces, spans and prompt payloads in a purpose-built observability backend and offers filtering and visualisations on top.  

 :::info NOTE
-• RAGFlow **≥ 0.21.0** (contains the Langfuse connector)  
+• RAGFlow **≥ 0.21.1** (contains the Langfuse connector)  
 • A Langfuse workspace (cloud or self-hosted) with a _Project Public Key_ and _Secret Key_
 :::

--- a/docs/guides/upgrade_ragflow.mdx
+++ b/docs/guides/upgrade_ragflow.mdx
@ -66,10 +66,10 @@ To upgrade RAGFlow, you must upgrade **both** your code **and** your Docker imag
   git clone https://github.com/infiniflow/ragflow.git
   ```

-2. Switch to the latest, officially published release, e.g., `v0.21.0`:
+2. Switch to the latest, officially published release, e.g., `v0.21.1`:

   ```bash
-   git checkout -f v0.21.0
+   git checkout -f v0.21.1
   ```

 3. Update **ragflow/docker/.env**:
@ -83,14 +83,14 @@ To upgrade RAGFlow, you must upgrade **both** your code **and** your Docker imag
  <TabItem value="slim">

 ```bash
-RAGFLOW_IMAGE=infiniflow/ragflow:v0.21.0-slim
+RAGFLOW_IMAGE=infiniflow/ragflow:v0.21.1-slim
 ```

  </TabItem>
  <TabItem value="full">

 ```bash
-RAGFLOW_IMAGE=infiniflow/ragflow:v0.21.0
+RAGFLOW_IMAGE=infiniflow/ragflow:v0.21.1
 ```

  </TabItem>
@ -114,10 +114,10 @@ No, you do not need to. Upgrading RAGFlow in itself will *not* remove your uploa
 1. From an environment with Internet access, pull the required Docker image.
 2. Save the Docker image to a **.tar** file.
   ```bash
-   docker save -o ragflow.v0.21.0.tar infiniflow/ragflow:v0.21.0
+   docker save -o ragflow.v0.21.1.tar infiniflow/ragflow:v0.21.1
   ```
 3. Copy the **.tar** file to the target server.
 4. Load the **.tar** file into Docker:
   ```bash
-   docker load -i ragflow.v0.21.0.tar
+   docker load -i ragflow.v0.21.1.tar
   ```
--- a/docs/quickstart.mdx
+++ b/docs/quickstart.mdx
@ -44,7 +44,7 @@ This section provides instructions on setting up the RAGFlow server on Linux. If

   `vm.max_map_count`. This value sets the maximum number of memory map areas a process may have. Its default value is 65530. While most applications require fewer than a thousand maps, reducing this value can result in abnormal behaviors, and the system will throw out-of-memory errors when a process reaches the limitation.

-   RAGFlow v0.21.0 uses Elasticsearch or [Infinity](https://github.com/infiniflow/infinity) for multiple recall. Setting the value of `vm.max_map_count` correctly is crucial to the proper functioning of the Elasticsearch component.
+   RAGFlow v0.21.1 uses Elasticsearch or [Infinity](https://github.com/infiniflow/infinity) for multiple recall. Setting the value of `vm.max_map_count` correctly is crucial to the proper functioning of the Elasticsearch component.

 <Tabs
  defaultValue="linux"
@ -184,13 +184,13 @@ This section provides instructions on setting up the RAGFlow server on Linux. If
   ```bash
   $ git clone https://github.com/infiniflow/ragflow.git
   $ cd ragflow/docker
-   $ git checkout -f v0.21.0
+   $ git checkout -f v0.21.1
   ```

 3. Use the pre-built Docker images and start up the server:

   :::tip NOTE
-   The command below downloads the `v0.21.0-slim` edition of the RAGFlow Docker image. Refer to the following table for descriptions of different RAGFlow editions. To download a RAGFlow edition different from `v0.21.0-slim`, update the `RAGFLOW_IMAGE` variable accordingly in **docker/.env** before using `docker compose` to start the server. For example: set `RAGFLOW_IMAGE=infiniflow/ragflow:v0.21.0` for the full edition `v0.21.0`.
+   The command below downloads the `v0.21.1-slim` edition of the RAGFlow Docker image. Refer to the following table for descriptions of different RAGFlow editions. To download a RAGFlow edition different from `v0.21.1-slim`, update the `RAGFLOW_IMAGE` variable accordingly in **docker/.env** before using `docker compose` to start the server. For example: set `RAGFLOW_IMAGE=infiniflow/ragflow:v0.21.1` for the full edition `v0.21.1`.
   :::

   ```bash
@ -207,8 +207,8 @@ This section provides instructions on setting up the RAGFlow server on Linux. If

 | RAGFlow image tag   | Image size (GB) | Has embedding models and Python packages? | Stable?                  |
 | ------------------- | --------------- | ----------------------------------------- | ------------------------ |
-| `v0.21.0`           | &approx;9       | :heavy_check_mark:                        | Stable release           |
-| `v0.21.0-slim`      | &approx;2       | ❌                                        | Stable release           |
+| `v0.21.1`           | &approx;9       | :heavy_check_mark:                        | Stable release           |
+| `v0.21.1-slim`      | &approx;2       | ❌                                        | Stable release           |
 | `nightly`           | &approx;9       | :heavy_check_mark:                        | *Unstable* nightly build |
 | `nightly-slim`      | &approx;2       | ❌                                        | *Unstable* nightly build |

@ -217,7 +217,7 @@ This section provides instructions on setting up the RAGFlow server on Linux. If
 ```

 :::danger IMPORTANT
-The embedding models included in `v0.21.0` and `nightly` are:
+The embedding models included in `v0.21.1` and `nightly` are:

 - BAAI/bge-large-zh-v1.5
 - maidalun1020/bce-embedding-base_v1
--- a/docs/references/glossary.mdx
+++ b/docs/references/glossary.mdx
@ -19,7 +19,7 @@ import TOCInline from '@theme/TOCInline';

 ### Cross-language search

-Cross-language search (also known as cross-lingual retrieval) is a feature introduced in version 0.21.0. It enables users to submit queries in one language (for example, English) and retrieve relevant documents written in other languages such as Chinese or Spanish. This feature is enabled by the system’s default chat model, which translates queries to ensure accurate matching of semantic meaning across languages.
+Cross-language search (also known as cross-lingual retrieval) is a feature introduced in version 0.21.1. It enables users to submit queries in one language (for example, English) and retrieve relevant documents written in other languages such as Chinese or Spanish. This feature is enabled by the system’s default chat model, which translates queries to ensure accurate matching of semantic meaning across languages.

 By enabling cross-language search, users can effortlessly access a broader range of information regardless of language barriers, significantly enhancing the system’s usability and inclusiveness.

--- a/docs/references/http_api_reference.md
+++ b/docs/references/http_api_reference.md
@ -1198,23 +1198,24 @@ Failure:

 ### List documents

-**GET** `/api/v1/datasets/{dataset_id}/documents?page={page}&page_size={page_size}&orderby={orderby}&desc={desc}&keywords={keywords}&id={document_id}&name={document_name}&create_time_from={timestamp}&create_time_to={timestamp}`
+**GET** `/api/v1/datasets/{dataset_id}/documents?page={page}&page_size={page_size}&orderby={orderby}&desc={desc}&keywords={keywords}&id={document_id}&name={document_name}&create_time_from={timestamp}&create_time_to={timestamp}&suffix={file_suffix}&run={run_status}`

 Lists documents in a specified dataset.

 #### Request

 - Method: GET
- URL: `/api/v1/datasets/{dataset_id}/documents?page={page}&page_size={page_size}&orderby={orderby}&desc={desc}&keywords={keywords}&id={document_id}&name={document_name}&create_time_from={timestamp}&create_time_to={timestamp}`
+- URL: `/api/v1/datasets/{dataset_id}/documents?page={page}&page_size={page_size}&orderby={orderby}&desc={desc}&keywords={keywords}&id={document_id}&name={document_name}&create_time_from={timestamp}&create_time_to={timestamp}&suffix={file_suffix}&run={run_status}`
 - Headers:
  - `'content-Type: application/json'`
  - `'Authorization: Bearer <YOUR_API_KEY>'`

-##### Request example
+##### Request examples

+**A basic request with pagination:**
 ```bash
 curl --request GET \
-     --url http://{address}/api/v1/datasets/{dataset_id}/documents?page={page}&page_size={page_size}&orderby={orderby}&desc={desc}&keywords={keywords}&id={document_id}&name={document_name}&create_time_from={timestamp}&create_time_to={timestamp} \
+     --url http://{address}/api/v1/datasets/{dataset_id}/documents?page=1&page_size=10 \
     --header 'Authorization: Bearer <YOUR_API_KEY>'
 ```

@ -1236,10 +1237,34 @@ curl --request GET \
  Indicates whether the retrieved documents should be sorted in descending order. Defaults to `true`.
 - `id`: (*Filter parameter*), `string`  
  The ID of the document to retrieve.
- `create_time_from`: (*Filter parameter*), `integer`
+- `create_time_from`: (*Filter parameter*), `integer`  
  Unix timestamp for filtering documents created after this time. 0 means no filter. Defaults to `0`.
- `create_time_to`: (*Filter parameter*), `integer`
+- `create_time_to`: (*Filter parameter*), `integer`  
  Unix timestamp for filtering documents created before this time. 0 means no filter. Defaults to `0`.
+- `suffix`: (*Filter parameter*), `array[string]`  
+  Filter by file suffix. Supports multiple values, e.g., `pdf`, `txt`, and `docx`. Defaults to all suffixes.
+- `run`: (*Filter parameter*), `array[string]`  
+  Filter by document processing status. Supports numeric, text, and mixed formats:  
+  - Numeric format: `["0", "1", "2", "3", "4"]`
+  - Text format: `[UNSTART, RUNNING, CANCEL, DONE, FAIL]`
+  - Mixed format: `[UNSTART, 1, DONE]` (mixing numeric and text formats)
+  - Status mapping:
+    - `0` / `UNSTART`: Document not yet processed
+    - `1` / `RUNNING`: Document is currently being processed
+    - `2` / `CANCEL`: Document processing was cancelled
+    - `3` / `DONE`: Document processing completed successfully
+    - `4` / `FAIL`: Document processing failed  
+  Defaults to all statuses.
+
+##### Usage examples
+
+**A request with multiple filtering parameters**
+
+```bash
+curl --request GET \
+     --url 'http://{address}/api/v1/datasets/{dataset_id}/documents?suffix=pdf&run=DONE&page=1&page_size=10' \
+     --header 'Authorization: Bearer <YOUR_API_KEY>'
+```

 #### Response

@ -1270,7 +1295,7 @@ Success:
                "process_duration": 0.0,
                "progress": 0.0,
                "progress_msg": "",
-                "run": "0",
+                "run": "UNSTART",
                "size": 7,
                "source_type": "local",
                "status": "1",
--- a/docs/release_notes.md
+++ b/docs/release_notes.md
@ -9,8 +9,8 @@ Key features, improvements and bug fixes in the latest releases.

 :::info
 Each RAGFlow release is available in two editions:
- **Slim edition**: excludes built-in embedding models and is identified by a **-slim** suffix added to the version name. Example: `infiniflow/ragflow:v0.21.0-slim`
- **Full edition**: includes built-in embedding models and has no suffix added to the version name. Example: `infiniflow/ragflow:v0.21.0`
+- **Slim edition**: excludes built-in embedding models and is identified by a **-slim** suffix added to the version name. Example: `infiniflow/ragflow:v0.21.1-slim`
+- **Full edition**: includes built-in embedding models and has no suffix added to the version name. Example: `infiniflow/ragflow:v0.21.1`
 :::

 :::danger IMPORTANT
@ -22,6 +22,23 @@ The embedding models included in a full edition are:
 These two embedding models are optimized specifically for English and Chinese, so performance may be compromised if you use them to embed documents in other languages.
 :::

+## v0.21.1
+
+Released on October 23, 2025.
+
+### New features
+
+- Experimental: Adds support for PDF document parsing using MinerU. See [here](./faq.mdx#how-to-use-mineru-to-parse-pdf-documents).
+
+### Improvements
+
+- Enhances UI/UX for the dataset and personal center pages.
+- Upgrades RAGFlow's document engine, [Infinity](https://github.com/infiniflow/infinity), to v0.6.1.
+
+### Fixed issues
+
+- An issue with video parsing.
+
 ## v0.21.0

 Released on October 15, 2025.
@ -30,7 +47,7 @@ Released on October 15, 2025.

 - Orchestratable ingestion pipeline: Supports customized data ingestion and cleansing workflows, enabling users to flexibly design their data flows or directly apply the official data flow templates on the canvas.
 - GraphRAG & RAPTOR write process optimized: Replaces the automatic incremental build process with manual batch building, significantly reducing construction overhead.
- Long-context RAG: Automatically generates document-level table of contents (TOC) structures to mitigate context loss caused by inaccurate or excessive chunking, substantially improving retrieval quality. This feature is now available via a TOC extraction template.
+- Long-context RAG: Automatically generates document-level table of contents (TOC) structures to mitigate context loss caused by inaccurate or excessive chunking, substantially improving retrieval quality. This feature is now available via a TOC extraction template. See [here](./guides/dataset/extract_table_of_contents.md).
 - Video file parsing: Expands the system's multimodal data processing capabilities by supporting video file parsing.
 - Admin CLI: Introduces a new command-line tool for system administration, allowing users to manage and monitor RAGFlow's service status via command line.

--- a/graphrag/general/extractor.py
+++ b/graphrag/general/extractor.py
@ -105,16 +105,36 @@ class Extractor:

        async def extract_all(doc_id, chunks, max_concurrency=MAX_CONCURRENT_PROCESS_AND_EXTRACT_CHUNK):
            out_results = []
+            error_count = 0
+            max_errors = 3
+
            limiter = trio.Semaphore(max_concurrency)

            async def worker(chunk_key_dp: tuple[str, str], idx: int, total: int):
+                nonlocal error_count
                async with limiter:
-                    await self._process_single_content(chunk_key_dp, idx, total, out_results)
+                    try:
+                        await self._process_single_content(chunk_key_dp, idx, total, out_results)
+                    except Exception as e:
+                        error_count += 1
+                        error_msg = f"Error processing chunk {idx+1}/{total}: {str(e)}"
+                        logging.warning(error_msg)
+                        if self.callback:
+                            self.callback(msg=error_msg)
+
+                        if error_count > max_errors:
+                            raise Exception(f"Maximum error count ({max_errors}) reached. Last errors: {str(e)}")

            async with trio.open_nursery() as nursery:
                for i, ck in enumerate(chunks):
                    nursery.start_soon(worker, (doc_id, ck), i, len(chunks))

+            if error_count > 0:
+                warning_msg = f"Completed with {error_count} errors (out of {len(chunks)} chunks processed)"
+                logging.warning(warning_msg)
+                if self.callback:
+                    self.callback(msg=warning_msg)
+
            return out_results

        out_results = await extract_all(doc_id, chunks, max_concurrency=MAX_CONCURRENT_PROCESS_AND_EXTRACT_CHUNK)
@ -129,8 +149,8 @@ class Extractor:
                maybe_edges[tuple(sorted(k))].extend(v)
            sum_token_count += token_count
        now = trio.current_time()
-        if callback:
-            callback(msg=f"Entities and relationships extraction done, {len(maybe_nodes)} nodes, {len(maybe_edges)} edges, {sum_token_count} tokens, {now - start_ts:.2f}s.")
+        if self.callback:
+            self.callback(msg=f"Entities and relationships extraction done, {len(maybe_nodes)} nodes, {len(maybe_edges)} edges, {sum_token_count} tokens, {now - start_ts:.2f}s.")
        start_ts = now
        logging.info("Entities merging...")
        all_entities_data = []
@ -138,8 +158,8 @@ class Extractor:
            for en_nm, ents in maybe_nodes.items():
                nursery.start_soon(self._merge_nodes, en_nm, ents, all_entities_data)
        now = trio.current_time()
-        if callback:
-            callback(msg=f"Entities merging done, {now - start_ts:.2f}s.")
+        if self.callback:
+            self.callback(msg=f"Entities merging done, {now - start_ts:.2f}s.")

        start_ts = now
        logging.info("Relationships merging...")
@ -148,8 +168,8 @@ class Extractor:
            for (src, tgt), rels in maybe_edges.items():
                nursery.start_soon(self._merge_edges, src, tgt, rels, all_relationships_data)
        now = trio.current_time()
-        if callback:
-            callback(msg=f"Relationships merging done, {now - start_ts:.2f}s.")
+        if self.callback:
+            self.callback(msg=f"Relationships merging done, {now - start_ts:.2f}s.")

        if not len(all_entities_data) and not len(all_relationships_data):
            logging.warning("Didn't extract any entities and relationships, maybe your LLM is not working")
--- a/helm/values.yaml
+++ b/helm/values.yaml
@ -56,7 +56,7 @@ env:
 ragflow:
  image:
    repository: infiniflow/ragflow
-    tag: v0.21.0-slim
+    tag: v0.21.1-slim
    pullPolicy: IfNotPresent
    pullSecrets: []
  # Optional service configuration overrides
@ -96,7 +96,7 @@ ragflow:
 infinity:
  image:
    repository: infiniflow/infinity
-    tag: v0.6.0
+    tag: v0.6.1
    pullPolicy: IfNotPresent
    pullSecrets: []
  storage:
--- a/pyproject.toml
+++ b/pyproject.toml
@ -1,6 +1,6 @@
 [project]
 name = "ragflow"
-version = "0.21.0"
+version = "0.21.1"
 description = "[RAGFlow](https://ragflow.io/) is an open-source RAG (Retrieval-Augmented Generation) engine based on deep document understanding. It offers a streamlined RAG workflow for businesses of any scale, combining LLM (Large Language Models) to provide truthful question-answering capabilities, backed by well-founded citations from various complex formatted data."
 authors = [{ name = "Zhichang Yu", email = "yuzhichang@gmail.com" }]
 license-files = ["LICENSE"]
@ -46,7 +46,7 @@ dependencies = [
    "html-text==0.6.2",
    "httpx[socks]>=0.28.1,<0.29.0",
    "huggingface-hub>=0.25.0,<0.26.0",
-    "infinity-sdk==0.6.0",
+    "infinity-sdk==0.6.1",
    "infinity-emb>=0.0.66,<0.0.67",
    "itsdangerous==2.1.2",
    "json-repair==0.35.0",
--- a/rag/app/book.py
+++ b/rag/app/book.py
@ -20,11 +20,14 @@ import re
 from io import BytesIO

 from deepdoc.parser.utils import get_text
+from rag.app import naive
 from rag.nlp import bullets_category, is_english,remove_contents_table, \
    hierarchical_merge, make_colon_as_title, naive_merge, random_choices, tokenize_table, \
    tokenize_chunks
 from rag.nlp import rag_tokenizer
-from deepdoc.parser import PdfParser, DocxParser, PlainParser, HtmlParser
+from deepdoc.parser import PdfParser, PlainParser, HtmlParser
+from deepdoc.parser.figure_parser import vision_figure_parser_pdf_wrapper,vision_figure_parser_docx_wrapper
+from PIL import Image


 class Pdf(PdfParser):
@ -81,13 +84,15 @@ def chunk(filename, binary=None, from_page=0, to_page=100000,
    sections, tbls = [], []
    if re.search(r"\.docx$", filename, re.IGNORECASE):
        callback(0.1, "Start to parse.")
-        doc_parser = DocxParser()
+        doc_parser = naive.Docx()
        # TODO: table of contents need to be removed
        sections, tbls = doc_parser(
-            binary if binary else filename, from_page=from_page, to_page=to_page)
+            filename, binary=binary, from_page=from_page, to_page=to_page)
        remove_contents_table(sections, eng=is_english(
            random_choices([t for t, _ in sections], k=200)))
-        tbls = [((None, lns), None) for lns in tbls]
+        tbls=vision_figure_parser_docx_wrapper(sections=sections,tbls=tbls,callback=callback,**kwargs)
+        # tbls = [((None, lns), None) for lns in tbls]
+        sections=[(item[0],item[1] if item[1] is not None else "") for item in sections if not isinstance(item[1], Image.Image)]
        callback(0.8, "Finish parsing.")

    elif re.search(r"\.pdf$", filename, re.IGNORECASE):
@ -96,6 +101,7 @@ def chunk(filename, binary=None, from_page=0, to_page=100000,
            pdf_parser = PlainParser()
        sections, tbls = pdf_parser(filename if not binary else binary,
                                    from_page=from_page, to_page=to_page, callback=callback)
+        tbls=vision_figure_parser_pdf_wrapper(tbls=tbls,callback=callback,**kwargs)

    elif re.search(r"\.txt$", filename, re.IGNORECASE):
        callback(0.1, "Start to parse.")
--- a/rag/app/manual.py
+++ b/rag/app/manual.py
@ -23,6 +23,7 @@ from io import BytesIO
 from rag.nlp import rag_tokenizer, tokenize, tokenize_table, bullets_category, title_frequency, tokenize_chunks, docx_question_level
 from rag.utils import num_tokens_from_string
 from deepdoc.parser import PdfParser, PlainParser, DocxParser
+from deepdoc.parser.figure_parser import vision_figure_parser_pdf_wrapper,vision_figure_parser_docx_wrapper
 from docx import Document
 from PIL import Image

@ -252,7 +253,7 @@ def chunk(filename, binary=None, from_page=0, to_page=100000,
            tk_cnt = num_tokens_from_string(txt)
            if sec_id > -1:
                last_sid = sec_id
-
+        tbls=vision_figure_parser_pdf_wrapper(tbls=tbls,callback=callback,**kwargs)
        res = tokenize_table(tbls, doc, eng)
        res.extend(tokenize_chunks(chunks, doc, eng, pdf_parser))
        return res
@ -261,6 +262,7 @@ def chunk(filename, binary=None, from_page=0, to_page=100000,
        docx_parser = Docx()
        ti_list, tbls = docx_parser(filename, binary,
                                    from_page=0, to_page=10000, callback=callback)
+        tbls=vision_figure_parser_docx_wrapper(sections=ti_list,tbls=tbls,callback=callback,**kwargs)
        res = tokenize_table(tbls, doc, eng)
        for text, image in ti_list:
            d = copy.deepcopy(doc)
--- a/rag/app/naive.py
+++ b/rag/app/naive.py
@ -16,10 +16,10 @@

 import logging
 import re
+import os
 from functools import reduce
 from io import BytesIO
 from timeit import default_timer as timer
-
 from docx import Document
 from docx.image.exceptions import InvalidImageStreamError, UnexpectedEndOfFileError, UnrecognizedImageError
 from docx.opc.pkgreader import _SerializedRelationships, _SerializedRelationship
@ -30,9 +30,11 @@ from tika import parser

 from api.db import LLMType
 from api.db.services.llm_service import LLMBundle
+from api.utils.file_utils import extract_embed_file
 from deepdoc.parser import DocxParser, ExcelParser, HtmlParser, JsonParser, MarkdownElementExtractor, MarkdownParser, PdfParser, TxtParser
-from deepdoc.parser.figure_parser import VisionFigureParser, vision_figure_parser_figure_data_wrapper
+from deepdoc.parser.figure_parser import VisionFigureParser,vision_figure_parser_docx_wrapper,vision_figure_parser_pdf_wrapper
 from deepdoc.parser.pdf_parser import PlainParser, VisionParser
+from deepdoc.parser.mineru_parser import MinerUParser
 from rag.nlp import concat_img, find_codec, naive_merge, naive_merge_with_images, naive_merge_docx, rag_tokenizer, tokenize_chunks, tokenize_chunks_with_images, tokenize_table


@ -435,6 +437,7 @@ def chunk(filename, binary=None, from_page=0, to_page=100000,
        Successive text will be sliced into pieces using 'delimiter'.
        Next, these successive pieces are merge into chunks whose token number is no more than 'Max token number'.
    """
+    

    is_english = lang.lower() == "english"  # is_english(cks)
    parser_config = kwargs.get(
@ -448,27 +451,37 @@ def chunk(filename, binary=None, from_page=0, to_page=100000,
    res = []
    pdf_parser = None
    section_images = None
+
+    is_root = kwargs.get("is_root", True)
+    embed_res = []
+    if is_root:
+        # Only extract embedded files at the root call
+        embeds = []
+        if binary is not None:
+            embeds = extract_embed_file(binary)
+        else:
+            raise Exception("Embedding extraction from file path is not supported.")
+        
+        # Recursively chunk each embedded file and collect results
+        for embed_filename, embed_bytes in embeds:
+            try:
+                sub_res = chunk(embed_filename, binary=embed_bytes, lang=lang, callback=callback, is_root=False, **kwargs) or []
+                embed_res.extend(sub_res)
+            except Exception as e:
+                if callback:
+                    callback(0.05, f"Failed to chunk embed {embed_filename}: {e}")
+                continue
+
    if re.search(r"\.docx$", filename, re.IGNORECASE):
        callback(0.1, "Start to parse.")

-        try:
-            vision_model = LLMBundle(kwargs["tenant_id"], LLMType.IMAGE2TEXT)
-            callback(0.15, "Visual model detected. Attempting to enhance figure extraction...")
-        except Exception:
-            vision_model = None
+        

        # fix "There is no item named 'word/NULL' in the archive", referring to https://github.com/python-openxml/python-docx/issues/1105#issuecomment-1298075246
        _SerializedRelationships.load_from_xml = load_from_xml_v2
        sections, tables = Docx()(filename, binary)

-        if vision_model:
-            figures_data = vision_figure_parser_figure_data_wrapper(sections)
-            try:
-                docx_vision_parser = VisionFigureParser(vision_model=vision_model, figures_data=figures_data, **kwargs)
-                boosted_figures = docx_vision_parser(callback=callback)
-                tables.extend(boosted_figures)
-            except Exception as e:
-                callback(0.6, f"Visual model error: {e}. Skipping figure parsing enhancement.")
+        tables=vision_figure_parser_docx_wrapper(sections=sections,tbls=tables,callback=callback,**kwargs)

        res = tokenize_table(tables, doc, is_english)
        callback(0.8, "Finish parsing.")
@ -481,10 +494,12 @@ def chunk(filename, binary=None, from_page=0, to_page=100000,
                "delimiter", "\n!?。；！？"))

        if kwargs.get("section_only", False):
+            chunks.extend(embed_res)
            return chunks

        res.extend(tokenize_chunks_with_images(chunks, doc, is_english, images))
        logging.info("naive_merge({}): {}".format(filename, timer() - st))
+        res.extend(embed_res)
        return res

    elif re.search(r"\.pdf$", filename, re.IGNORECASE):
@ -495,29 +510,28 @@ def chunk(filename, binary=None, from_page=0, to_page=100000,

        if layout_recognizer == "DeepDOC":
            pdf_parser = Pdf()
-
-            try:
-                vision_model = LLMBundle(kwargs["tenant_id"], LLMType.IMAGE2TEXT)
-                callback(0.15, "Visual model detected. Attempting to enhance figure extraction...")
-            except Exception:
-                vision_model = None
-
-            if vision_model:
-                sections, tables, figures = pdf_parser(filename if not binary else binary, from_page=from_page, to_page=to_page, callback=callback, separate_tables_figures=True)
-                callback(0.5, "Basic parsing complete. Proceeding with figure enhancement...")
-                try:
-                    pdf_vision_parser = VisionFigureParser(vision_model=vision_model, figures_data=figures, **kwargs)
-                    boosted_figures = pdf_vision_parser(callback=callback)
-                    tables.extend(boosted_figures)
-                except Exception as e:
-                    callback(0.6, f"Visual model error: {e}. Skipping figure parsing enhancement.")
-                    tables.extend(figures)
-            else:
-                sections, tables = pdf_parser(filename if not binary else binary, from_page=from_page, to_page=to_page, callback=callback)
+            sections, tables = pdf_parser(filename if not binary else binary, from_page=from_page, to_page=to_page, callback=callback)
+            tables=vision_figure_parser_pdf_wrapper(tbls=tables,callback=callback,**kwargs)

            res = tokenize_table(tables, doc, is_english)
            callback(0.8, "Finish parsing.")

+        elif layout_recognizer == "MinerU":
+            mineru_executable = os.environ.get("MINERU_EXECUTABLE", "mineru")
+            pdf_parser = MinerUParser(mineru_path=mineru_executable)
+            if not pdf_parser.check_installation():
+                callback(-1, "MinerU not found.")
+                return res
+
+            sections, tables = pdf_parser.parse_pdf(
+                filepath=filename,
+                binary=binary,
+                callback=callback,
+                output_dir=os.environ.get("MINERU_OUTPUT_DIR", ""),
+                delete_output=bool(int(os.environ.get("MINERU_DELETE_OUTPUT", 1))),
+            )
+            parser_config["chunk_token_num"] = 0
+            callback(0.8, "Finish parsing.")
        else:
            if layout_recognizer == "Plain Text":
                pdf_parser = PlainParser()
@ -604,7 +618,6 @@ def chunk(filename, binary=None, from_page=0, to_page=100000,
            callback(0.8, f"tika.parser got empty content from {filename}.")
            logging.warning(f"tika.parser got empty content from {filename}.")
            return []
-
    else:
        raise NotImplementedError(
            "file type not supported yet(pdf, xlsx, doc, docx, txt supported)")
@ -621,6 +634,7 @@ def chunk(filename, binary=None, from_page=0, to_page=100000,
                                            "chunk_token_num", 128)), parser_config.get(
                                            "delimiter", "\n!?。；！？"))
        if kwargs.get("section_only", False):
+            chunks.extend(embed_res)
            return chunks

        res.extend(tokenize_chunks_with_images(chunks, doc, is_english, images))
@ -630,11 +644,14 @@ def chunk(filename, binary=None, from_page=0, to_page=100000,
                "chunk_token_num", 128)), parser_config.get(
                "delimiter", "\n!?。；！？"))
        if kwargs.get("section_only", False):
+            chunks.extend(embed_res)
            return chunks

        res.extend(tokenize_chunks(chunks, doc, is_english, pdf_parser))

    logging.info("naive_merge({}): {}".format(filename, timer() - st))
+    if embed_res:
+        res.extend(embed_res)
    return res


--- a/rag/app/one.py
+++ b/rag/app/one.py
@ -23,6 +23,7 @@ from deepdoc.parser.utils import get_text
 from rag.app import naive
 from rag.nlp import rag_tokenizer, tokenize
 from deepdoc.parser import PdfParser, ExcelParser, PlainParser, HtmlParser
+from deepdoc.parser.figure_parser import vision_figure_parser_pdf_wrapper,vision_figure_parser_docx_wrapper


 class Pdf(PdfParser):
@ -57,13 +58,8 @@ class Pdf(PdfParser):

        sections = [(b["text"], self.get_position(b, zoomin))
                    for i, b in enumerate(self.boxes)]
-        for (img, rows), poss in tbls:
-            if not rows:
-                continue
-            sections.append((rows if isinstance(rows, str) else rows[0],
-                             [(p[0] + 1 - from_page, p[1], p[2], p[3], p[4]) for p in poss]))
        return [(txt, "") for txt, _ in sorted(sections, key=lambda x: (
-            x[-1][0][0], x[-1][0][3], x[-1][0][1]))], None
+            x[-1][0][0], x[-1][0][3], x[-1][0][1]))], tbls


 def chunk(filename, binary=None, from_page=0, to_page=100000,
@ -80,6 +76,7 @@ def chunk(filename, binary=None, from_page=0, to_page=100000,
    if re.search(r"\.docx$", filename, re.IGNORECASE):
        callback(0.1, "Start to parse.")
        sections, tbls = naive.Docx()(filename, binary)
+        tbls=vision_figure_parser_docx_wrapper(sections=sections,tbls=tbls,callback=callback,**kwargs)
        sections = [s for s, _ in sections if s]
        for (_, html), _ in tbls:
            sections.append(html)
@ -89,8 +86,14 @@ def chunk(filename, binary=None, from_page=0, to_page=100000,
        pdf_parser = Pdf()
        if parser_config.get("layout_recognize", "DeepDOC") == "Plain Text":
            pdf_parser = PlainParser()
-        sections, _ = pdf_parser(
+        sections, tbls = pdf_parser(
            filename if not binary else binary, to_page=to_page, callback=callback)
+        tbls=vision_figure_parser_pdf_wrapper(tbls=tbls,callback=callback,**kwargs)
+        for (img, rows), poss in tbls:
+            if not rows:
+                continue
+            sections.append((rows if isinstance(rows, str) else rows[0],
+                             [(p[0] + 1 - from_page, p[1], p[2], p[3], p[4]) for p in poss]))
        sections = [s for s, _ in sections if s]

    elif re.search(r"\.xlsx?$", filename, re.IGNORECASE):
--- a/rag/app/paper.py
+++ b/rag/app/paper.py
@ -18,12 +18,12 @@ import logging
 import copy
 import re

+from deepdoc.parser.figure_parser import vision_figure_parser_pdf_wrapper
 from api.db import ParserType
 from rag.nlp import rag_tokenizer, tokenize, tokenize_table, add_positions, bullets_category, title_frequency, tokenize_chunks
 from deepdoc.parser import PdfParser, PlainParser
 import numpy as np

-
 class Pdf(PdfParser):
    def __init__(self):
        self.model_speciess = ParserType.PAPER.value
@ -160,6 +160,9 @@ def chunk(filename, binary=None, from_page=0, to_page=100000,
            pdf_parser = Pdf()
            paper = pdf_parser(filename if not binary else binary,
                               from_page=from_page, to_page=to_page, callback=callback)
+        tbls=paper["tables"]
+        tbls=vision_figure_parser_pdf_wrapper(tbls=tbls,callback=callback,**kwargs)
+        paper["tables"] = tbls
    else:
        raise NotImplementedError("file type not supported yet(pdf supported)")

--- a/rag/app/picture.py
+++ b/rag/app/picture.py
@ -23,44 +23,62 @@ from PIL import Image
 from api.db import LLMType
 from api.db.services.llm_service import LLMBundle
 from deepdoc.vision import OCR
-from rag.nlp import tokenize
+from rag.nlp import rag_tokenizer, tokenize
 from rag.utils import clean_markdown_block
-from rag.nlp import rag_tokenizer
-

 ocr = OCR()

+# Gemini supported MIME types
+VIDEO_EXTS = [".mp4", ".mov", ".avi", ".flv", ".mpeg", ".mpg", ".webm", ".wmv", ".3gp", ".3gpp", ".mkv"]
+

 def chunk(filename, binary, tenant_id, lang, callback=None, **kwargs):
-    img = Image.open(io.BytesIO(binary)).convert('RGB')
    doc = {
        "docnm_kwd": filename,
        "title_tks": rag_tokenizer.tokenize(re.sub(r"\.[a-zA-Z]+$", "", filename)),
-        "image": img,
-        "doc_type_kwd": "image"
    }
-    bxs = ocr(np.array(img))
-    txt = "\n".join([t[0] for _, t in bxs if t[0]])
    eng = lang.lower() == "english"
-    callback(0.4, "Finish OCR: (%s ...)" % txt[:12])
-    if (eng and len(txt.split()) > 32) or len(txt) > 32:
-        tokenize(doc, txt, eng)
-        callback(0.8, "OCR results is too long to use CV LLM.")
-        return [doc]

-    try:
-        callback(0.4, "Use CV LLM to describe the picture.")
-        cv_mdl = LLMBundle(tenant_id, LLMType.IMAGE2TEXT, lang=lang)
-        img_binary = io.BytesIO()
-        img.save(img_binary, format='JPEG')
-        img_binary.seek(0)
-        ans = cv_mdl.describe(img_binary.read())
-        callback(0.8, "CV LLM respond: %s ..." % ans[:32])
-        txt += "\n" + ans
-        tokenize(doc, txt, eng)
-        return [doc]
-    except Exception as e:
-        callback(prog=-1, msg=str(e))
+    if any(filename.lower().endswith(ext) for ext in VIDEO_EXTS):
+        try:
+            doc.update({"doc_type_kwd": "video"})
+            cv_mdl = LLMBundle(tenant_id, llm_type=LLMType.IMAGE2TEXT, lang=lang)
+            ans = cv_mdl.chat(system="", history=[], gen_conf={}, video_bytes=binary, filename=filename)
+            callback(0.8, "CV LLM respond: %s ..." % ans[:32])
+            ans += "\n" + ans
+            tokenize(doc, ans, eng)
+            return [doc]
+        except Exception as e:
+            callback(prog=-1, msg=str(e))
+    else:
+        img = Image.open(io.BytesIO(binary)).convert("RGB")
+        doc.update(
+            {
+                "image": img,
+                "doc_type_kwd": "image",
+            }
+        )
+        bxs = ocr(np.array(img))
+        txt = "\n".join([t[0] for _, t in bxs if t[0]])
+        callback(0.4, "Finish OCR: (%s ...)" % txt[:12])
+        if (eng and len(txt.split()) > 32) or len(txt) > 32:
+            tokenize(doc, txt, eng)
+            callback(0.8, "OCR results is too long to use CV LLM.")
+            return [doc]
+
+        try:
+            callback(0.4, "Use CV LLM to describe the picture.")
+            cv_mdl = LLMBundle(tenant_id, LLMType.IMAGE2TEXT, lang=lang)
+            img_binary = io.BytesIO()
+            img.save(img_binary, format="JPEG")
+            img_binary.seek(0)
+            ans = cv_mdl.describe(img_binary.read())
+            callback(0.8, "CV LLM respond: %s ..." % ans[:32])
+            txt += "\n" + ans
+            tokenize(doc, txt, eng)
+            return [doc]
+        except Exception as e:
+            callback(prog=-1, msg=str(e))

    return []

@ -79,7 +97,7 @@ def vision_llm_chunk(binary, vision_model, prompt=None, callback=None):

    try:
        with io.BytesIO() as img_binary:
-            img.save(img_binary, format='JPEG')
+            img.save(img_binary, format="JPEG")
            img_binary.seek(0)
            ans = clean_markdown_block(vision_model.describe_with_prompt(img_binary.read(), prompt))
            txt += "\n" + ans
--- a/rag/flow/parser/parser.py
+++ b/rag/flow/parser/parser.py
@ -29,6 +29,7 @@ from api.db.services.llm_service import LLMBundle
 from api.utils import get_uuid
 from api.utils.base64_image import image2id
 from deepdoc.parser import ExcelParser
+from deepdoc.parser.mineru_parser import MinerUParser
 from deepdoc.parser.pdf_parser import PlainParser, RAGFlowPdfParser, VisionParser
 from rag.app.naive import Docx
 from rag.flow.base import ProcessBase, ProcessParamBase
@ -138,9 +139,16 @@ class ParserParam(ProcessParamBase):
                    "oggvorbis",
                    "ape"
                ],
-                "output_format": "json",
+                "output_format": "text",
+            },
+            "video": {
+                "suffix":[
+                    "mp4",
+                    "avi",
+                    "mkv"
+                ],
+                "output_format": "text",
            },
-            "video": {},
        }

    def check(self):
@ -149,7 +157,7 @@ class ParserParam(ProcessParamBase):
            pdf_parse_method = pdf_config.get("parse_method", "")
            self.check_empty(pdf_parse_method, "Parse method abnormal.")

-            if pdf_parse_method.lower() not in ["deepdoc", "plain_text"]:
+            if pdf_parse_method.lower() not in ["deepdoc", "plain_text", "mineru"]:
                self.check_empty(pdf_config.get("lang", ""), "PDF VLM language")

            pdf_output_format = pdf_config.get("output_format", "")
@ -185,6 +193,10 @@ class ParserParam(ProcessParamBase):
        if audio_config:
            self.check_empty(audio_config.get("llm_id"), "Audio VLM")

+        video_config = self.setups.get("video", "")
+        if video_config:
+            self.check_empty(video_config.get("llm_id"), "Video VLM")
+
        email_config = self.setups.get("email", "")
        if email_config:
            email_output_format = email_config.get("output_format", "")
@ -207,13 +219,34 @@ class Parser(ProcessBase):
        elif conf.get("parse_method").lower() == "plain_text":
            lines, _ = PlainParser()(blob)
            bboxes = [{"text": t} for t, _ in lines]
+        elif conf.get("parse_method").lower() == "mineru":
+            mineru_executable = os.environ.get("MINERU_EXECUTABLE", "mineru")
+            pdf_parser = MinerUParser(mineru_path=mineru_executable)
+            if not pdf_parser.check_installation():
+                raise RuntimeError("MinerU not found. Please install it via: pip install -U 'mineru[core]'.")
+
+            lines, _ = pdf_parser.parse_pdf(
+                filepath=name,
+                binary=blob,
+                callback=self.callback,
+                output_dir=os.environ.get("MINERU_OUTPUT_DIR", ""),
+                delete_output=bool(int(os.environ.get("MINERU_DELETE_OUTPUT", 1))),
+            )
+            bboxes = []
+            for t, poss in lines:
+                box = {
+                    "image": pdf_parser.crop(poss, 1),
+                    "positions": [[pos[0][-1], *pos[1:]] for pos in pdf_parser.extract_positions(poss)],
+                    "text": t,
+                }
+                bboxes.append(box)
        else:
            vision_model = LLMBundle(self._canvas._tenant_id, LLMType.IMAGE2TEXT, llm_name=conf.get("parse_method"), lang=self._param.setups["pdf"].get("lang"))
            lines, _ = VisionParser(vision_model=vision_model)(blob, callback=self.callback)
            bboxes = []
            for t, poss in lines:
-                pn, x0, x1, top, bott = poss.split(" ")
-                bboxes.append({"page_number": int(pn), "x0": float(x0), "x1": float(x1), "top": float(top), "bottom": float(bott), "text": t})
+                for pn, x0, x1, top, bott in RAGFlowPdfParser.extract_positions(poss):
+                    bboxes.append({"page_number": int(pn[0]), "x0": float(x0), "x1": float(x1), "top": float(top), "bottom": float(bott), "text": t})

        if conf.get("output_format") == "json":
            self.set_output("json", bboxes)
@ -357,6 +390,17 @@ class Parser(ProcessBase):

            self.set_output("text", txt)

+    def _video(self, name, blob):
+        self.callback(random.randint(1, 5) / 100.0, "Start to work on an video.")
+
+        conf = self._param.setups["video"]
+        self.set_output("output_format", conf["output_format"])
+
+        cv_mdl = LLMBundle(self._canvas.get_tenant_id(), LLMType.IMAGE2TEXT)
+        txt = cv_mdl.chat(system="", history=[], gen_conf={}, video_bytes=blob, filename=name)
+
+        self.set_output("text", txt)
+
    def _email(self, name, blob):
        self.callback(random.randint(1, 5) / 100.0, "Start to work on an email.")

@ -483,6 +527,7 @@ class Parser(ProcessBase):
            "word": self._word,
            "image": self._image,
            "audio": self._audio,
+            "video": self._video,
            "email": self._email,
        }
        try:
--- a/rag/llm/chat_model.py
+++ b/rag/llm/chat_model.py
@ -167,7 +167,7 @@ class Base(ABC):
        ans = response.choices[0].message.content.strip()
        if response.choices[0].finish_reason == "length":
            ans = self._length_stop(ans)
-        return ans, self.total_token_count(response)
+        return ans, total_token_count_from_response(response)

    def _chat_streamly(self, history, gen_conf, **kwargs):
        logging.info("[HISTORY STREAMLY]" + json.dumps(history, ensure_ascii=False, indent=4))
@ -193,7 +193,7 @@ class Base(ABC):
                reasoning_start = False
                ans = resp.choices[0].delta.content

-            tol = self.total_token_count(resp)
+            tol = total_token_count_from_response(resp)
            if not tol:
                tol = num_tokens_from_string(resp.choices[0].delta.content)

@ -283,7 +283,7 @@ class Base(ABC):
                for _ in range(self.max_rounds + 1):
                    logging.info(f"{self.tools=}")
                    response = self.client.chat.completions.create(model=self.model_name, messages=history, tools=self.tools, tool_choice="auto", **gen_conf)
-                    tk_count += self.total_token_count(response)
+                    tk_count += total_token_count_from_response(response)
                    if any([not response.choices, not response.choices[0].message]):
                        raise Exception(f"500 response structure error. Response: {response}")

@ -401,7 +401,7 @@ class Base(ABC):
                            answer += resp.choices[0].delta.content
                            yield resp.choices[0].delta.content

-                        tol = self.total_token_count(resp)
+                        tol = total_token_count_from_response(resp)
                        if not tol:
                            total_tokens += num_tokens_from_string(resp.choices[0].delta.content)
                        else:
@ -437,7 +437,7 @@ class Base(ABC):
                    if not resp.choices[0].delta.content:
                        resp.choices[0].delta.content = ""
                        continue
-                    tol = self.total_token_count(resp)
+                    tol = total_token_count_from_response(resp)
                    if not tol:
                        total_tokens += num_tokens_from_string(resp.choices[0].delta.content)
                    else:
@ -472,9 +472,6 @@ class Base(ABC):

        yield total_tokens

-    def total_token_count(self, resp):
-        return total_token_count_from_response(resp)
-
    def _calculate_dynamic_ctx(self, history):
        """Calculate dynamic context window size"""

@ -604,7 +601,7 @@ class BaiChuanChat(Base):
                ans += LENGTH_NOTIFICATION_CN
            else:
                ans += LENGTH_NOTIFICATION_EN
-        return ans, self.total_token_count(response)
+        return ans, total_token_count_from_response(response)

    def chat_streamly(self, system, history, gen_conf={}, **kwargs):
        if system and history and history[0].get("role") != "system":
@ -627,7 +624,7 @@ class BaiChuanChat(Base):
                if not resp.choices[0].delta.content:
                    resp.choices[0].delta.content = ""
                ans = resp.choices[0].delta.content
-                tol = self.total_token_count(resp)
+                tol = total_token_count_from_response(resp)
                if not tol:
                    total_tokens += num_tokens_from_string(resp.choices[0].delta.content)
                else:
@ -691,9 +688,9 @@ class ZhipuChat(Base):
                        ans += LENGTH_NOTIFICATION_CN
                    else:
                        ans += LENGTH_NOTIFICATION_EN
-                    tk_count = self.total_token_count(resp)
+                    tk_count = total_token_count_from_response(resp)
                if resp.choices[0].finish_reason == "stop":
-                    tk_count = self.total_token_count(resp)
+                    tk_count = total_token_count_from_response(resp)
                yield ans
        except Exception as e:
            yield ans + "\n**ERROR**: " + str(e)
@ -812,7 +809,7 @@ class MiniMaxChat(Base):
                ans += LENGTH_NOTIFICATION_CN
            else:
                ans += LENGTH_NOTIFICATION_EN
-        return ans, self.total_token_count(response)
+        return ans, total_token_count_from_response(response)

    def chat_streamly(self, system, history, gen_conf):
        if system and history and history[0].get("role") != "system":
@ -847,7 +844,7 @@ class MiniMaxChat(Base):
                if "choices" in resp and "delta" in resp["choices"][0]:
                    text = resp["choices"][0]["delta"]["content"]
                ans = text
-                tol = self.total_token_count(resp)
+                tol = total_token_count_from_response(resp)
                if not tol:
                    total_tokens += num_tokens_from_string(text)
                else:
@ -886,7 +883,7 @@ class MistralChat(Base):
                ans += LENGTH_NOTIFICATION_CN
            else:
                ans += LENGTH_NOTIFICATION_EN
-        return ans, self.total_token_count(response)
+        return ans, total_token_count_from_response(response)

    def chat_streamly(self, system, history, gen_conf={}, **kwargs):
        if system and history and history[0].get("role") != "system":
@ -1110,7 +1107,7 @@ class BaiduYiyanChat(Base):
        system = history[0]["content"] if history and history[0]["role"] == "system" else ""
        response = self.client.do(model=self.model_name, messages=[h for h in history if h["role"] != "system"], system=system, **gen_conf).body
        ans = response["result"]
-        return ans, self.total_token_count(response)
+        return ans, total_token_count_from_response(response)

    def chat_streamly(self, system, history, gen_conf={}, **kwargs):
        gen_conf["penalty_score"] = ((gen_conf.get("presence_penalty", 0) + gen_conf.get("frequency_penalty", 0)) / 2) + 1
@ -1124,7 +1121,7 @@ class BaiduYiyanChat(Base):
            for resp in response:
                resp = resp.body
                ans = resp["result"]
-                total_tokens = self.total_token_count(resp)
+                total_tokens = total_token_count_from_response(resp)

                yield ans

@ -1478,7 +1475,7 @@ class LiteLLMBase(ABC):
        if response.choices[0].finish_reason == "length":
            ans = self._length_stop(ans)

-        return ans, self.total_token_count(response)
+        return ans, total_token_count_from_response(response)

    def _chat_streamly(self, history, gen_conf, **kwargs):
        logging.info("[HISTORY STREAMLY]" + json.dumps(history, ensure_ascii=False, indent=4))
@ -1512,7 +1509,7 @@ class LiteLLMBase(ABC):
                reasoning_start = False
                ans = delta.content

-            tol = self.total_token_count(resp)
+            tol = total_token_count_from_response(resp)
            if not tol:
                tol = num_tokens_from_string(delta.content)

@ -1665,7 +1662,7 @@ class LiteLLMBase(ABC):
                        timeout=self.timeout,
                    )

-                    tk_count += self.total_token_count(response)
+                    tk_count += total_token_count_from_response(response)

                    if not hasattr(response, "choices") or not response.choices or not response.choices[0].message:
                        raise Exception(f"500 response structure error. Response: {response}")
@ -1797,7 +1794,7 @@ class LiteLLMBase(ABC):
                            answer += delta.content
                            yield delta.content

-                        tol = self.total_token_count(resp)
+                        tol = total_token_count_from_response(resp)
                        if not tol:
                            total_tokens += num_tokens_from_string(delta.content)
                        else:
@ -1846,7 +1843,7 @@ class LiteLLMBase(ABC):
                    delta = resp.choices[0].delta
                    if not hasattr(delta, "content") or delta.content is None:
                        continue
-                    tol = self.total_token_count(resp)
+                    tol = total_token_count_from_response(resp)
                    if not tol:
                        total_tokens += num_tokens_from_string(delta.content)
                    else:
@ -1880,17 +1877,6 @@ class LiteLLMBase(ABC):

        yield total_tokens

-    def total_token_count(self, resp):
-        try:
-            return resp.usage.total_tokens
-        except Exception:
-            pass
-        try:
-            return resp["usage"]["total_tokens"]
-        except Exception:
-            pass
-        return 0
-
    def _calculate_dynamic_ctx(self, history):
        """Calculate dynamic context window size"""

--- a/rag/llm/cv_model.py
+++ b/rag/llm/cv_model.py
@ -13,12 +13,16 @@
 #  See the License for the specific language governing permissions and
 #  limitations under the License.
 #
+
 import base64
 import json
 import os
+import tempfile
+import logging
 from abc import ABC
 from copy import deepcopy
 from io import BytesIO
+from pathlib import Path
 from urllib.parse import urljoin
 import requests
 from openai import OpenAI
@ -46,7 +50,7 @@ class Base(ABC):
    def describe_with_prompt(self, image, prompt=None):
        raise NotImplementedError("Please implement encode method!")

-    def _form_history(self, system, history, images=[]):
+    def _form_history(self, system, history, images=None):
        hist = []
        if system:
            hist.append({"role": "system", "content": system})
@ -74,7 +78,7 @@ class Base(ABC):
            })
        return pmpt

-    def chat(self, system, history, gen_conf, images=[], **kwargs):
+    def chat(self, system, history, gen_conf, images=None, **kwargs):
        try:
            response = self.client.chat.completions.create(
                model=self.model_name,
@ -85,7 +89,7 @@ class Base(ABC):
        except Exception as e:
            return "**ERROR**: " + str(e), 0

-    def chat_streamly(self, system, history, gen_conf, images=[], **kwargs):
+    def chat_streamly(self, system, history, gen_conf, images=None, **kwargs):
        ans = ""
        tk_count = 0
        try:
@ -170,6 +174,7 @@ class GptV4(Base):
    def __init__(self, key, model_name="gpt-4-vision-preview", lang="Chinese", base_url="https://api.openai.com/v1", **kwargs):
        if not base_url:
            base_url = "https://api.openai.com/v1"
+        self.api_key = key
        self.client = OpenAI(api_key=key, base_url=base_url)
        self.model_name = model_name
        self.lang = lang
@ -223,6 +228,61 @@ class QWenCV(GptV4):
            base_url = "https://dashscope.aliyuncs.com/compatible-mode/v1"
        super().__init__(key, model_name, lang=lang, base_url=base_url, **kwargs)

+    def chat(self, system, history, gen_conf, images=None, video_bytes=None, filename=""):
+        if video_bytes:
+            try:
+                summary, summary_num_tokens = self._process_video(video_bytes, filename)
+                return summary, summary_num_tokens
+            except Exception as e:
+                return "**ERROR**: " + str(e), 0
+
+        return "**ERROR**: Method chat not supported yet.", 0
+
+    def _process_video(self, video_bytes, filename):
+        from dashscope import MultiModalConversation
+
+        video_suffix = Path(filename).suffix or ".mp4"
+        with tempfile.NamedTemporaryFile(delete=False, suffix=video_suffix) as tmp:
+            tmp.write(video_bytes)
+            tmp_path = tmp.name
+
+        video_path = f"file://{tmp_path}"
+        messages = [
+            {
+                "role": "user",
+                "content": [
+                    {
+                        "video": video_path,
+                        "fps": 2,
+                    },
+                    {
+                        "text": "Please summarize this video in proper sentences.",
+                    },
+                ],
+            }
+        ]
+
+        def call_api():
+            response = MultiModalConversation.call(
+                api_key=self.api_key,
+                model=self.model_name,
+                messages=messages,
+            )
+            summary = response["output"]["choices"][0]["message"].content[0]["text"]
+            return summary, num_tokens_from_string(summary)
+
+        try:
+            return call_api()
+        except Exception as e1:
+            import dashscope
+
+            dashscope.base_http_api_url = "https://dashscope-intl.aliyuncs.com/api/v1"
+            try:
+                return call_api()
+            except Exception as e2:
+                raise RuntimeError(f"Both default and intl endpoint failed.\nFirst error: {e1}\nSecond error: {e2}")
+
+

 class HunyuanCV(GptV4):
    _FACTORY_NAME = "Tencent Hunyuan"
@ -254,6 +314,17 @@ class StepFunCV(GptV4):
        self.lang = lang
        Base.__init__(self, **kwargs)

+class VolcEngineCV(GptV4):
+    _FACTORY_NAME = "VolcEngine"
+
+    def __init__(self, key, model_name, lang="Chinese", base_url="https://ark.cn-beijing.volces.com/api/v3", **kwargs):
+        if not base_url:
+            base_url = "https://ark.cn-beijing.volces.com/api/v3"
+        ark_api_key = json.loads(key).get("ark_api_key", "")
+        self.client = OpenAI(api_key=ark_api_key, base_url=base_url)
+        self.model_name = json.loads(key).get("ep_id", "") + json.loads(key).get("endpoint_id", "")
+        self.lang = lang
+        Base.__init__(self, **kwargs)

 class LmStudioCV(GptV4):
    _FACTORY_NAME = "LM-Studio"
@ -435,7 +506,7 @@ class OllamaCV(Base):
            options["frequency_penalty"] = gen_conf["frequency_penalty"]
        return options

-    def _form_history(self, system, history, images=[]):
+    def _form_history(self, system, history, images=None):
        hist = deepcopy(history)
        if system and hist[0]["role"] == "user":
            hist.insert(0, {"role": "system", "content": system})
@ -476,7 +547,7 @@ class OllamaCV(Base):
        except Exception as e:
            return "**ERROR**: " + str(e), 0

-    def chat(self, system, history, gen_conf, images=[]):
+    def chat(self, system, history, gen_conf, images=None):
        try:
            response = self.client.chat(
                model=self.model_name,
@ -490,7 +561,7 @@ class OllamaCV(Base):
        except Exception as e:
            return "**ERROR**: " + str(e), 0

-    def chat_streamly(self, system, history, gen_conf, images=[]):
+    def chat_streamly(self, system, history, gen_conf, images=None):
        ans = ""
        try:
            response = self.client.chat(
@ -518,13 +589,14 @@ class GeminiCV(Base):

        client.configure(api_key=key)
        _client = client.get_default_generative_client()
+        self.api_key=key
        self.model_name = model_name
        self.model = GenerativeModel(model_name=self.model_name)
        self.model._client = _client
        self.lang = lang
        Base.__init__(self, **kwargs)

-    def _form_history(self, system, history, images=[]):
+    def _form_history(self, system, history, images=None):
        hist = []
        if system:
            hist.append({"role": "user", "parts": [system, history[0]["content"]]})
@ -560,7 +632,15 @@ class GeminiCV(Base):
                res = self.model.generate_content(input)
                return res.text, total_token_count_from_response(res)

-    def chat(self, system, history, gen_conf, images=[]):
+
+    def chat(self, system, history, gen_conf, images=None, video_bytes=None, filename=""):
+        if video_bytes:
+            try:
+                summary, summary_num_tokens = self._process_video(video_bytes, filename)
+                return summary, summary_num_tokens
+            except Exception as e:
+                return "**ERROR**: " + str(e), 0
+
        generation_config = dict(temperature=gen_conf.get("temperature", 0.3), top_p=gen_conf.get("top_p", 0.7))
        try:
            response = self.model.generate_content(
@ -571,7 +651,7 @@ class GeminiCV(Base):
        except Exception as e:
            return "**ERROR**: " + str(e), 0

-    def chat_streamly(self, system, history, gen_conf, images=[]):
+    def chat_streamly(self, system, history, gen_conf, images=None):
        ans = ""
        response = None
        try:
@ -592,6 +672,46 @@ class GeminiCV(Base):

        yield total_token_count_from_response(response)

+    def _process_video(self, video_bytes, filename):
+        from google import genai
+        from google.genai import types
+
+        video_size_mb = len(video_bytes) / (1024 * 1024)
+        client = genai.Client(api_key=self.api_key)
+
+        tmp_path = None
+        try:
+            if video_size_mb <= 20:
+                response = client.models.generate_content(
+                    model="models/gemini-2.5-flash",
+                    contents=types.Content(parts=[
+                        types.Part(inline_data=types.Blob(data=video_bytes, mime_type="video/mp4")),
+                        types.Part(text="Please summarize the video in proper sentences.")
+                    ])
+                )
+            else:
+                logging.info(f"Video size {video_size_mb:.2f}MB exceeds 20MB. Using Files API...")
+                video_suffix = Path(filename).suffix or ".mp4"
+                with tempfile.NamedTemporaryFile(delete=False, suffix=video_suffix) as tmp:
+                    tmp.write(video_bytes)
+                    tmp_path = Path(tmp.name)
+                uploaded_file = client.files.upload(file=tmp_path)
+
+                response = client.models.generate_content(
+                    model="gemini-2.5-flash",
+                    contents=[uploaded_file, "Please summarize this video in proper sentences."]
+                )
+
+            summary = response.text or ""
+            logging.info(f"Video summarized: {summary[:32]}...")
+            return summary, num_tokens_from_string(summary)
+        except Exception as e:
+            logging.error(f"Video processing failed: {e}")
+            raise
+        finally:
+            if tmp_path and tmp_path.exists():
+                tmp_path.unlink()
+

 class NvidiaCV(Base):
    _FACTORY_NAME = "NVIDIA"
@ -662,7 +782,7 @@ class NvidiaCV(Base):
            total_token_count_from_response(response)
        )

-    def chat(self, system, history, gen_conf, images=[], **kwargs):
+    def chat(self, system, history, gen_conf, images=None, **kwargs):
        try:
            response = self._request(self._form_history(system, history, images), gen_conf)
            return (
@ -672,7 +792,7 @@ class NvidiaCV(Base):
        except Exception as e:
            return "**ERROR**: " + str(e), 0

-    def chat_streamly(self, system, history, gen_conf, images=[], **kwargs):
+    def chat_streamly(self, system, history, gen_conf, images=None, **kwargs):
        total_tokens = 0
        try:
            response = self._request(self._form_history(system, history, images), gen_conf)
@ -738,7 +858,7 @@ class AnthropicCV(Base):
            gen_conf["max_tokens"] = self.max_tokens
        return gen_conf

-    def chat(self, system, history, gen_conf, images=[]):
+    def chat(self, system, history, gen_conf, images=None):
        gen_conf = self._clean_conf(gen_conf)
        ans = ""
        try:
@ -759,7 +879,7 @@ class AnthropicCV(Base):
        except Exception as e:
            return ans + "\n**ERROR**: " + str(e), 0

-    def chat_streamly(self, system, history, gen_conf, images=[]):
+    def chat_streamly(self, system, history, gen_conf, images=None):
        gen_conf = self._clean_conf(gen_conf)
        total_tokens = 0
        try:
@ -843,13 +963,13 @@ class GoogleCV(AnthropicCV, GeminiCV):
        else:
            return GeminiCV.describe_with_prompt(self, image, prompt)

-    def chat(self, system, history, gen_conf, images=[]):
+    def chat(self, system, history, gen_conf, images=None):
        if "claude" in self.model_name:
            return AnthropicCV.chat(self, system, history, gen_conf, images)
        else:
            return GeminiCV.chat(self, system, history, gen_conf, images)

-    def chat_streamly(self, system, history, gen_conf, images=[]):
+    def chat_streamly(self, system, history, gen_conf, images=None):
        if "claude" in self.model_name:
            for ans in AnthropicCV.chat_streamly(self, system, history, gen_conf, images):
                yield ans
--- a/rag/nlp/init.py
+++ b/rag/nlp/init.py
@ -459,12 +459,10 @@ def tree_merge(bull, sections, depth):
                return len(BULLET_PATTERN[bull])+1, text
            else:
                return len(BULLET_PATTERN[bull])+2, text
-    
    level_set = set()
    lines = []
    for section in sections:
        level, text = get_level(bull, section)
-
        if not text.strip("\n"):
            continue
            
@ -797,8 +795,8 @@ class Node:
    def __init__(self, level, depth=-1, texts=None):
        self.level = level
        self.depth = depth
-        self.texts = texts if texts is not None else []  # 存放内容
-        self.children = []  # 子节点
+        self.texts = texts or []
+        self.children = [] 

    def add_child(self, child_node):
        self.children.append(child_node)
@ -825,35 +823,51 @@ class Node:
        return f"Node(level={self.level}, texts={self.texts}, children={len(self.children)})"

    def build_tree(self, lines):
-        stack = [self]  
-        for line in lines:
-            level, text = line
-            node = Node(level=level, texts=[text])
-
-            if level <= self.depth or self.depth == -1:
-                while stack and level <= stack[-1].get_level():
-                    stack.pop()
-
-                stack[-1].add_child(node)
-                stack.append(node)
-            else:
+        stack = [self]
+        for level, text in lines:
+            if self.depth != -1 and level > self.depth:
+                # Beyond target depth: merge content into the current leaf instead of creating deeper nodes
                stack[-1].add_text(text)
-        return self  
+                continue
+
+            # Move up until we find the proper parent whose level is strictly smaller than current
+            while len(stack) > 1 and level <= stack[-1].get_level():
+                stack.pop()
+
+            node = Node(level=level, texts=[text])
+            # Attach as child of current parent and descend
+            stack[-1].add_child(node)
+            stack.append(node)
+
+        return self

    def get_tree(self):
        tree_list = []  
-        self._dfs(self, tree_list, 0, [])
+        self._dfs(self, tree_list, [])
        return tree_list

-    def _dfs(self, node, tree_list, current_depth, titles):
+    def _dfs(self, node, tree_list, titles):
+        level = node.get_level()
+        texts = node.get_texts()
+        child = node.get_children()

-        if node.get_texts():
-            if 0 < node.get_level() < self.depth:
-                titles.extend(node.get_texts())
-            else:
-                combined_text = ["\n".join(titles + node.get_texts())]
-                tree_list.append(combined_text)
+        if level == 0 and texts:
+            tree_list.append("\n".join(titles+texts))

+        # Titles within configured depth are accumulated into the current path
+        if 1 <= level <= self.depth:
+            path_titles = titles + texts
+        else:
+            path_titles = titles

-        for child in node.get_children():
-            self._dfs(child, tree_list, current_depth + 1, titles.copy())
+        # Body outside the depth limit becomes its own chunk under the current title path
+        if level > self.depth and texts:
+            tree_list.append("\n".join(path_titles + texts))
+
+        # A leaf title within depth emits its title path as a chunk (header-only section)
+        elif not child and (1 <= level <= self.depth):
+            tree_list.append("\n".join(path_titles))
+        
+        # Recurse into children with the updated title path
+        for c in child:
+            self._dfs(c, tree_list, path_titles)
--- a/rag/nlp/search.py
+++ b/rag/nlp/search.py
@ -388,6 +388,7 @@ class Dealer:
            else:
                # Don't need rerank here since Infinity normalizes each way score before fusion.
                sim = [sres.field[id].get("_score", 0.0) for id in sres.ids]
+                sim = [s if s is not None else 0. for s in sim]
                tsim = sim
                vsim = sim
        # Already paginated in search function
--- a/rag/raptor.py
+++ b/rag/raptor.py
@ -114,7 +114,7 @@ class RecursiveAbstractiveProcessing4TreeOrganizedRetrieval:
                            ),
                        }
                    ],
-                    {"max_tokens": self._max_token},
+                    {"max_tokens": max(self._max_token, 512)}, # fix issue:  #10235
                )
                cnt = re.sub(
                    "(······\n由于长度的原因，回答被截断了，要继续吗？|For the content length reason, it stopped, continue?)",
--- a/rag/svr/task_executor.py
+++ b/rag/svr/task_executor.py
@ -228,9 +228,10 @@ async def collect():
    canceled = False
    if msg.get("doc_id", "") in [GRAPH_RAPTOR_FAKE_DOC_ID, CANVAS_DEBUG_DOC_ID]:
        task = msg
-        if task["task_type"] in ["graphrag", "raptor", "mindmap"] and msg.get("doc_ids", []):
+        if task["task_type"] in ["graphrag", "raptor", "mindmap"]:
            task = TaskService.get_task(msg["id"], msg["doc_ids"])
-            task["doc_ids"] = msg["doc_ids"]
+            task["doc_id"] = msg["doc_id"]
+            task["doc_ids"] = msg.get("doc_ids", []) or []
    else:
        task = TaskService.get_task(msg["id"])

@ -1052,13 +1053,14 @@ async def task_manager():

 async def main():
    logging.info(r"""
-  ______           __      ______                     __
- /_  __/___ ______/ /__   / ____/  _____  _______  __/ /_____  _____
-  / / / __ `/ ___/ //_/  / __/ | |/_/ _ \/ ___/ / / / __/ __ \/ ___/
- / / / /_/ (__  ) ,<    / /____>  </  __/ /__/ /_/ / /_/ /_/ / /
-/_/  \__,_/____/_/|_|  /_____/_/|_|\___/\___/\__,_/\__/\____/_/
+    ____                      __  _
+   /  _/___  ____ ____  _____/ /_(_)___  ____     ________  ______   _____  _____
+   / // __ \/ __ `/ _ \/ ___/ __/ / __ \/ __ \   / ___/ _ \/ ___/ | / / _ \/ ___/
+ _/ // / / / /_/ /  __(__  ) /_/ / /_/ / / / /  (__  )  __/ /   | |/ /  __/ /
+/___/_/ /_/\__, /\___/____/\__/_/\____/_/ /_/  /____/\___/_/    |___/\___/_/
+          /____/
    """)
-    logging.info(f'TaskExecutor: RAGFlow version: {get_ragflow_version()}')
+    logging.info(f'RAGFlow version: {get_ragflow_version()}')
    settings.init_settings()
    print_rag_settings()
    if sys.platform != "win32":
--- a/rag/utils/oss_conn.py
+++ b/rag/utils/oss_conn.py
@ -106,7 +106,7 @@ class RAGFlowOSS:

    @use_prefix_path
    @use_default_bucket
-    def put(self, bucket, fnm, binary):
+    def put(self, bucket, fnm, binary, tenant_id=None):
        logging.debug(f"bucket name {bucket}; filename :{fnm}:")
        for _ in range(1):
            try:
@ -123,7 +123,7 @@ class RAGFlowOSS:

    @use_prefix_path
    @use_default_bucket
-    def rm(self, bucket, fnm):
+    def rm(self, bucket, fnm, tenant_id=None):
        try:
            self.conn.delete_object(Bucket=bucket, Key=fnm)
        except Exception:
@ -131,7 +131,7 @@ class RAGFlowOSS:

    @use_prefix_path
    @use_default_bucket
-    def get(self, bucket, fnm):
+    def get(self, bucket, fnm, tenant_id=None):
        for _ in range(1):
            try:
                r = self.conn.get_object(Bucket=bucket, Key=fnm)
@ -145,7 +145,7 @@ class RAGFlowOSS:

    @use_prefix_path
    @use_default_bucket
-    def obj_exist(self, bucket, fnm):
+    def obj_exist(self, bucket, fnm, tenant_id=None):
        try:
            if self.conn.head_object(Bucket=bucket, Key=fnm):
                return True
@ -157,7 +157,7 @@ class RAGFlowOSS:

    @use_prefix_path
    @use_default_bucket
-    def get_presigned_url(self, bucket, fnm, expires):
+    def get_presigned_url(self, bucket, fnm, expires, tenant_id=None):
        for _ in range(10):
            try:
                r = self.conn.generate_presigned_url('get_object',
--- a/sdk/python/pyproject.toml
+++ b/sdk/python/pyproject.toml
@ -1,6 +1,6 @@
 [project]
 name = "ragflow-sdk"
-version = "0.21.0"
+version = "0.21.1"
 description = "Python client sdk of [RAGFlow](https://github.com/infiniflow/ragflow). RAGFlow is an open-source RAG (Retrieval-Augmented Generation) engine based on deep document understanding."
 authors = [{ name = "Zhichang Yu", email = "yuzhichang@gmail.com" }]
 license = { text = "Apache License, Version 2.0" }
--- a/sdk/python/uv.lock
+++ b/sdk/python/uv.lock
@ -342,7 +342,7 @@ wheels = [

 [[package]]
 name = "ragflow-sdk"
-version = "0.21.0"
+version = "0.21.1"
 source = { virtual = "." }
 dependencies = [
    { name = "beartype" },
--- a/test/testcases/test_http_api/test_chunk_management_within_dataset/test_retrieval_chunks.py
+++ b/test/testcases/test_http_api/test_chunk_management_within_dataset/test_retrieval_chunks.py
@ -83,7 +83,7 @@ class TestChunksRetrieval:
                "ValueError('Search does not support negative slicing.')",
                marks=pytest.mark.skip,
            ),
-            pytest.param({"page": 2, "page_size": 2}, 0, 2, "", marks=pytest.mark.skip(reason="issues/6646")),
+            ({"page": 2, "page_size": 2}, 0, 2, ""),
            ({"page": 3, "page_size": 2}, 0, 0, ""),
            ({"page": "3", "page_size": 2}, 0, 0, ""),
            pytest.param(
@ -124,9 +124,9 @@ class TestChunksRetrieval:
                marks=pytest.mark.skip,
            ),
            # ({"page_size": 0}, 0, 0, ""),
-            ({"page_size": 1}, 0, 1, ""),
+            pytest.param({"page_size": 1}, 0, 1, "", marks=pytest.mark.skip(reason="issues/10692")),
            ({"page_size": 5}, 0, 4, ""),
-            ({"page_size": "1"}, 0, 1, ""),
+            pytest.param({"page_size": "1"}, 0, 1, "", marks=pytest.mark.skip(reason="issues/10692")),
            # ({"page_size": -1}, 0, 0, ""),
            pytest.param(
                {"page_size": "a"},
--- a/uv.lock
+++ b/uv.lock
@ -291,9 +291,9 @@ dependencies = [
    { name = "python-dateutil" },
    { name = "types-python-dateutil" },
 ]
-sdist = { url = "https://pypi.tuna.tsinghua.edu.cn/packages/2e/00/0f6e8fcdb23ea632c866620cc872729ff43ed91d284c866b515c6342b173/arrow-1.3.0.tar.gz", hash = "sha256:d4540617648cb5f895730f1ad8c82a65f2dad0166f57b75f3ca54759c4d67a85" }
+sdist = { url = "https://pypi.tuna.tsinghua.edu.cn/packages/2e/00/0f6e8fcdb23ea632c866620cc872729ff43ed91d284c866b515c6342b173/arrow-1.3.0.tar.gz", hash = "sha256:d4540617648cb5f895730f1ad8c82a65f2dad0166f57b75f3ca54759c4d67a85", size = 131960, upload-time = "2023-09-30T22:11:18.25Z" }
 wheels = [
-    { url = "https://pypi.tuna.tsinghua.edu.cn/packages/f8/ed/e97229a566617f2ae958a6b13e7cc0f585470eac730a73e9e82c32a3cdd2/arrow-1.3.0-py3-none-any.whl", hash = "sha256:c728b120ebc00eb84e01882a6f5e7927a53960aa990ce7dd2b10f39005a67f80" },
+    { url = "https://pypi.tuna.tsinghua.edu.cn/packages/f8/ed/e97229a566617f2ae958a6b13e7cc0f585470eac730a73e9e82c32a3cdd2/arrow-1.3.0-py3-none-any.whl", hash = "sha256:c728b120ebc00eb84e01882a6f5e7927a53960aa990ce7dd2b10f39005a67f80", size = 66419, upload-time = "2023-09-30T22:11:16.072Z" },
 ]

 [[package]]
@ -2709,7 +2709,7 @@ wheels = [

 [[package]]
 name = "infinity-sdk"
-version = "0.6.0"
+version = "0.6.1"
 source = { registry = "https://pypi.tuna.tsinghua.edu.cn/simple" }
 dependencies = [
    { name = "numpy" },
@ -2726,7 +2726,7 @@ dependencies = [
    { name = "thrift" },
 ]
 wheels = [
-    { url = "https://pypi.tuna.tsinghua.edu.cn/packages/f4/12/1ce243cbede6da5fc28e5462d90d96b13995446b3a90889287d31255b36e/infinity_sdk-0.6.0-py3-none-any.whl", hash = "sha256:e379853ffc44acba428572d633032e6c9bb842d1f08e9cad690916f52a8c6ba8", size = 75256, upload-time = "2025-10-14T12:05:13.918Z" },
+    { url = "https://pypi.tuna.tsinghua.edu.cn/packages/44/0e/7a596a41a79d15bb6c87e76862aa287bb98243a40fa7a31096b57df01613/infinity_sdk-0.6.1-py3-none-any.whl", hash = "sha256:b9cb1f7fee28569de8b763c353aa299fa141af70e67ceadc70562c84237229e4", size = 75260, upload-time = "2025-10-21T13:11:06.265Z" },
 ]

 [[package]]
@ -5467,7 +5467,7 @@ wheels = [

 [[package]]
 name = "ragflow"
-version = "0.21.0"
+version = "0.21.1"
 source = { virtual = "." }
 dependencies = [
    { name = "akshare" },
@ -5679,7 +5679,7 @@ requires-dist = [
    { name = "httpx", extras = ["socks"], specifier = ">=0.28.1,<0.29.0" },
    { name = "huggingface-hub", specifier = ">=0.25.0,<0.26.0" },
    { name = "infinity-emb", specifier = ">=0.0.66,<0.0.67" },
-    { name = "infinity-sdk", specifier = "==0.6.0" },
+    { name = "infinity-sdk", specifier = "==0.6.1" },
    { name = "itsdangerous", specifier = "==2.1.2" },
    { name = "json-repair", specifier = "==0.35.0" },
    { name = "langfuse", specifier = ">=2.60.0" },
--- a/web/.storybook/main.ts
+++ b/web/.storybook/main.ts
@ -27,6 +27,21 @@ const config: StorybookConfig = {
              },
            ],
          },
+          {
+            test: /\.less$/,
+            use: [
+              'style-loader',
+              'css-loader',
+              {
+                loader: 'postcss-loader',
+                options: {
+                  postcssOptions: {
+                    plugins: [require('tailwindcss'), require('autoprefixer')],
+                  },
+                },
+              },
+            ],
+          },
        ],
      },
    },
--- a/web/src/components/avatar-upload.tsx
+++ b/web/src/components/avatar-upload.tsx
@ -1,5 +1,5 @@
 import { transformFile2Base64 } from '@/utils/file-util';
-import { Pencil, Upload, XIcon } from 'lucide-react';
+import { Pencil, Plus, XIcon } from 'lucide-react';
 import {
  ChangeEventHandler,
  forwardRef,
@ -12,10 +12,14 @@ import { Avatar, AvatarFallback, AvatarImage } from './ui/avatar';
 import { Button } from './ui/button';
 import { Input } from './ui/input';

-type AvatarUploadProps = { value?: string; onChange?: (value: string) => void };
+type AvatarUploadProps = {
+  value?: string;
+  onChange?: (value: string) => void;
+  tips?: string;
+};

 export const AvatarUpload = forwardRef<HTMLInputElement, AvatarUploadProps>(
-  function AvatarUpload({ value, onChange }, ref) {
+  function AvatarUpload({ value, onChange, tips }, ref) {
    const { t } = useTranslation();
    const [avatarBase64Str, setAvatarBase64Str] = useState(''); // Avatar Image base64

@ -47,9 +51,9 @@ export const AvatarUpload = forwardRef<HTMLInputElement, AvatarUploadProps>(
      <div className="flex justify-start items-end space-x-2">
        <div className="relative group">
          {!avatarBase64Str ? (
-            <div className="w-[64px] h-[64px] grid place-content-center border border-dashed	rounded-md">
+            <div className="w-[64px] h-[64px] grid place-content-center border border-dashed bg-bg-input	rounded-md">
              <div className="flex flex-col items-center">
-                <Upload />
+                <Plus />
                <p>{t('common.upload')}</p>
              </div>
            </div>
@ -86,8 +90,8 @@ export const AvatarUpload = forwardRef<HTMLInputElement, AvatarUploadProps>(
            ref={ref}
          />
        </div>
-        <div className="margin-1 text-muted-foreground">
-          {t('knowledgeConfiguration.photoTip')}
+        <div className="margin-1 text-text-secondary">
+          {tips ?? t('knowledgeConfiguration.photoTip')}
        </div>
      </div>
    );
--- a/web/src/components/card-container.tsx
+++ b/web/src/components/card-container.tsx
@ -0,0 +1,17 @@
+import { cn } from '@/lib/utils';
+import { PropsWithChildren } from 'react';
+
+type CardContainerProps = { className?: string } & PropsWithChildren;
+
+export function CardContainer({ children, className }: CardContainerProps) {
+  return (
+    <section
+      className={cn(
+        'grid gap-6 sm:grid-cols-1 md:grid-cols-2 lg:grid-cols-3 xl:grid-cols-4 2xl:grid-cols-5',
+        className,
+      )}
+    >
+      {children}
+    </section>
+  );
+}
--- a/web/src/components/collapse.tsx
+++ b/web/src/components/collapse.tsx
@ -51,8 +51,8 @@ export function Collapse({
      onOpenChange={handleOpenChange}
      disabled={disabled}
    >
-      <CollapsibleTrigger className="w-full">
-        <section className="flex justify-between items-center pb-2">
+      <CollapsibleTrigger className={'w-full'}>
+        <section className="flex justify-between items-center">
          <div className="flex items-center gap-1">
            <IconFontFill
              name={`more`}
@ -60,12 +60,18 @@ export function Collapse({
                'rotate-90': !currentOpen,
              })}
            ></IconFontFill>
-            {title}
+            <div
+              className={cn('text-text-secondary', {
+                'text-text-primary': open,
+              })}
+            >
+              {title}
+            </div>
          </div>
          <div>{rightContent}</div>
        </section>
      </CollapsibleTrigger>
-      <CollapsibleContent>{children}</CollapsibleContent>
+      <CollapsibleContent className="pt-2">{children}</CollapsibleContent>
    </Collapsible>
  );
 }
@ -94,7 +100,7 @@ export function NodeCollapsible<T extends any[]>({
    >
      {nextItems.slice(0, 3).map(children)}
      <CollapsibleContent className={nextClassName}>
-        {nextItems.slice(3).map(children)}
+        {nextItems.slice(3).map((x, idx) => children(x, idx + 3))}
      </CollapsibleContent>
      {nextItems.length > 3 && (
        <CollapsibleTrigger
--- a/web/src/components/confirm-delete-dialog.tsx
+++ b/web/src/components/confirm-delete-dialog.tsx
@ -56,13 +56,13 @@ export function ConfirmDeleteDialog({
        </AlertDialogHeader>
        <AlertDialogFooter>
          <AlertDialogCancel onClick={onCancel}>
-            {t('common.cancel')}
+            {t('common.no')}
          </AlertDialogCancel>
          <AlertDialogAction
            className="bg-state-error text-text-primary"
            onClick={onOk}
          >
-            {t('common.ok')}
+            {t('common.yes')}
          </AlertDialogAction>
        </AlertDialogFooter>
      </AlertDialogContent>
--- a/web/src/components/edit-tag/index.tsx
+++ b/web/src/components/edit-tag/index.tsx
@ -1,5 +1,4 @@
 import { PlusOutlined } from '@ant-design/icons';
-import { TweenOneGroup } from 'rc-tween-one';
 import React, { useEffect, useRef, useState } from 'react';

 import { X } from 'lucide-react';
@ -16,7 +15,7 @@ interface EditTagsProps {
 }

 const EditTag = React.forwardRef<HTMLDivElement, EditTagsProps>(
-  ({ value = [], onChange }: EditTagsProps, ref) => {
+  ({ value = [], onChange }: EditTagsProps) => {
    const [inputVisible, setInputVisible] = useState(false);
    const [inputValue, setInputValue] = useState('');
    const inputRef = useRef<HTMLInputElement>(null);
@ -57,7 +56,7 @@ const EditTag = React.forwardRef<HTMLDivElement, EditTagsProps>(
        <HoverCard key={tag}>
          <HoverCardContent side="top">{tag}</HoverCardContent>
          <HoverCardTrigger asChild>
-            <div className="w-fit flex items-center justify-center gap-2 border-dashed border px-1 rounded-sm bg-bg-card">
+            <div className="w-fit flex items-center justify-center gap-2 border-dashed border px-2 py-1 rounded-sm bg-bg-card">
              <div className="flex gap-2 items-center">
                <div className="max-w-80 overflow-hidden text-ellipsis">
                  {tag}
@ -84,11 +83,11 @@ const EditTag = React.forwardRef<HTMLDivElement, EditTagsProps>(

    return (
      <div>
-        {inputVisible ? (
+        {inputVisible && (
          <Input
            ref={inputRef}
            type="text"
-            className="h-8 bg-bg-card"
+            className="h-8 bg-bg-card mb-1"
            value={inputValue}
            onChange={handleInputChange}
            onBlur={handleInputConfirm}
@ -98,36 +97,20 @@ const EditTag = React.forwardRef<HTMLDivElement, EditTagsProps>(
              }
            }}
          />
-        ) : (
-          <Button
-            variant="dashed"
-            className="w-fit flex items-center justify-center gap-2 bg-bg-card"
-            onClick={showInput}
-            style={tagPlusStyle}
-          >
-            <PlusOutlined />
-          </Button>
-        )}
-        {Array.isArray(tagChild) && tagChild.length > 0 && (
-          <TweenOneGroup
-            className="flex gap-2 flex-wrap mt-2"
-            enter={{
-              scale: 0.8,
-              opacity: 0,
-              type: 'from',
-              duration: 100,
-            }}
-            onEnd={(e) => {
-              if (e.type === 'appear' || e.type === 'enter') {
-                (e.target as any).style = 'display: inline-block';
-              }
-            }}
-            leave={{ opacity: 0, width: 0, scale: 0, duration: 200 }}
-            appear={false}
-          >
-            {tagChild}
-          </TweenOneGroup>
        )}
+        <div className="flex gap-2 py-1">
+          {Array.isArray(tagChild) && tagChild.length > 0 && <>{tagChild}</>}
+          {!inputVisible && (
+            <Button
+              variant="dashed"
+              className="w-fit flex items-center justify-center gap-2 bg-bg-card"
+              onClick={showInput}
+              style={tagPlusStyle}
+            >
+              <PlusOutlined />
+            </Button>
+          )}
+        </div>
      </div>
    );
  },
--- a/web/src/components/floating-chat-widget-markdown.tsx
+++ b/web/src/components/floating-chat-widget-markdown.tsx
@ -1,22 +1,32 @@
 import Image from '@/components/image';
 import SvgIcon from '@/components/svg-icon';
-import { useFetchDocumentThumbnailsByIds, useGetDocumentUrl } from '@/hooks/document-hooks';
+import {
+  useFetchDocumentThumbnailsByIds,
+  useGetDocumentUrl,
+} from '@/hooks/document-hooks';
 import { IReference, IReferenceChunk } from '@/interfaces/database/chat';
-import { preprocessLaTeX, replaceThinkToSection, showImage } from '@/utils/chat';
+import {
+  preprocessLaTeX,
+  replaceThinkToSection,
+  showImage,
+} from '@/utils/chat';
 import { getExtension } from '@/utils/document-util';
 import { InfoCircleOutlined } from '@ant-design/icons';
 import { Button, Flex, Popover, Tooltip } from 'antd';
 import classNames from 'classnames';
 import DOMPurify from 'dompurify';
+import 'katex/dist/katex.min.css';
 import { omit } from 'lodash';
 import { pipe } from 'lodash/fp';
-import 'katex/dist/katex.min.css';
 import { useCallback, useEffect, useMemo } from 'react';
 import { useTranslation } from 'react-i18next';
 import Markdown from 'react-markdown';
 import reactStringReplace from 'react-string-replace';
 import SyntaxHighlighter from 'react-syntax-highlighter';
-import { oneDark, oneLight } from 'react-syntax-highlighter/dist/esm/styles/prism';
+import {
+  oneDark,
+  oneLight,
+} from 'react-syntax-highlighter/dist/esm/styles/prism';
 import rehypeKatex from 'rehype-katex';
 import rehypeRaw from 'rehype-raw';
 import remarkGfm from 'remark-gfm';
@ -39,7 +49,8 @@ const FloatingChatWidgetMarkdown = ({
  clickDocumentButton?: (documentId: string, chunk: IReferenceChunk) => void;
 }) => {
  const { t } = useTranslation();
-  const { setDocumentIds, data: fileThumbnails } = useFetchDocumentThumbnailsByIds();
+  const { setDocumentIds, data: fileThumbnails } =
+    useFetchDocumentThumbnailsByIds();
  const getDocumentUrl = useGetDocumentUrl();
  const isDarkTheme = useIsDarkTheme();

@ -51,23 +62,37 @@ const FloatingChatWidgetMarkdown = ({

  useEffect(() => {
    const docAggs = reference?.doc_aggs;
-    const docList = Array.isArray(docAggs) ? docAggs : Object.values(docAggs ?? {});
+    const docList = Array.isArray(docAggs)
+      ? docAggs
+      : Object.values(docAggs ?? {});
    setDocumentIds(docList.map((x: any) => x.doc_id).filter(Boolean));
  }, [reference, setDocumentIds]);

-  const handleDocumentButtonClick = useCallback((documentId: string, chunk: IReferenceChunk, isPdf: boolean, documentUrl?: string) => () => {
-    if (!documentId) return;
-    if (!isPdf && documentUrl) {
-      window.open(documentUrl, '_blank');
-    } else if (clickDocumentButton) {
-      clickDocumentButton(documentId, chunk);
-    }
-  }, [clickDocumentButton]);
+  const handleDocumentButtonClick = useCallback(
+    (
+      documentId: string,
+      chunk: IReferenceChunk,
+      isPdf: boolean,
+      documentUrl?: string,
+    ) =>
+      () => {
+        if (!documentId) return;
+        if (!isPdf && documentUrl) {
+          window.open(documentUrl, '_blank');
+        } else if (clickDocumentButton) {
+          clickDocumentButton(documentId, chunk);
+        }
+      },
+    [clickDocumentButton],
+  );

  const rehypeWrapReference = () => (tree: any) => {
    visitParents(tree, 'text', (node, ancestors) => {
      const latestAncestor = ancestors[ancestors.length - 1];
-      if (latestAncestor.tagName !== 'custom-typography' && latestAncestor.tagName !== 'code') {
+      if (
+        latestAncestor.tagName !== 'custom-typography' &&
+        latestAncestor.tagName !== 'code'
+      ) {
        node.type = 'element';
        node.tagName = 'custom-typography';
        node.properties = {};
@ -76,90 +101,173 @@ const FloatingChatWidgetMarkdown = ({
    });
  };

-  const getReferenceInfo = useCallback((chunkIndex: number) => {
-    const chunkItem = reference?.chunks?.[chunkIndex];
-    if (!chunkItem) return null;
-    const docAggsArray = Array.isArray(reference?.doc_aggs) ? reference.doc_aggs : Object.values(reference?.doc_aggs ?? {});
-    const document = docAggsArray.find((x: any) => x?.doc_id === chunkItem?.document_id) as any;
-    const documentId = document?.doc_id;
-    const documentUrl = document?.url ?? (documentId ? getDocumentUrl(documentId) : undefined);
-    const fileThumbnail = documentId ? fileThumbnails[documentId] : '';
-    const fileExtension = documentId ? getExtension(document?.doc_name ?? '') : '';
-    return { documentUrl, fileThumbnail, fileExtension, imageId: chunkItem.image_id, chunkItem, documentId, document };
-  }, [fileThumbnails, reference, getDocumentUrl]);
+  const getReferenceInfo = useCallback(
+    (chunkIndex: number) => {
+      const chunkItem = reference?.chunks?.[chunkIndex];
+      if (!chunkItem) return null;
+      const docAggsArray = Array.isArray(reference?.doc_aggs)
+        ? reference.doc_aggs
+        : Object.values(reference?.doc_aggs ?? {});
+      const document = docAggsArray.find(
+        (x: any) => x?.doc_id === chunkItem?.document_id,
+      ) as any;
+      const documentId = document?.doc_id;
+      const documentUrl =
+        document?.url ?? (documentId ? getDocumentUrl(documentId) : undefined);
+      const fileThumbnail = documentId ? fileThumbnails[documentId] : '';
+      const fileExtension = documentId
+        ? getExtension(document?.doc_name ?? '')
+        : '';
+      return {
+        documentUrl,
+        fileThumbnail,
+        fileExtension,
+        imageId: chunkItem.image_id,
+        chunkItem,
+        documentId,
+        document,
+      };
+    },
+    [fileThumbnails, reference, getDocumentUrl],
+  );

-  const getPopoverContent = useCallback((chunkIndex: number) => {
-    const info = getReferenceInfo(chunkIndex);
-
-    if (!info) {
-      return <div className="p-2 text-xs text-red-500">Error: Missing document information.</div>;
-    }
-
-    const { documentUrl, fileThumbnail, fileExtension, imageId, chunkItem, documentId, document } = info;
-
-    return (
-      <div key={`popover-content-${chunkItem.id}`} className="flex gap-2 widget-citation-content">
-        {imageId && (
-          <Popover placement="left" content={<Image id={imageId} className="max-w-[80vw] max-h-[60vh] rounded" />}>
-            <Image id={imageId} className="w-24 h-24 object-contain rounded m-1 cursor-pointer" />
-          </Popover>
-        )}
-        <div className="space-y-2 flex-1 min-w-0">
-          <div
-            dangerouslySetInnerHTML={{ __html: DOMPurify.sanitize(chunkItem?.content ?? '') }}
-            className="max-h-[250px] overflow-y-auto text-xs leading-relaxed p-2 bg-gray-50 dark:bg-gray-800 rounded prose-sm"
-          ></div>
-          {documentId && (
-            <Flex gap={'small'} align="center">
-              {fileThumbnail ? (
-                <img src={fileThumbnail} alt={document?.doc_name} className="w-6 h-6 rounded" />
-              ) : (
-                <SvgIcon name={`file-icon/${fileExtension}`} width={20} />
-              )}
-              <Tooltip title={!documentUrl && fileExtension !== 'pdf' ? 'Document link unavailable' : document.doc_name}>
-                <Button
-                  type="link"
-                  size="small"
-                  className="p-0 text-xs break-words h-auto text-left flex-1"
-                  onClick={handleDocumentButtonClick(documentId, chunkItem, fileExtension === 'pdf', documentUrl)}
-                  disabled={!documentUrl && fileExtension !== 'pdf'}
-                  style={{ whiteSpace: 'normal' }}
-                >
-                  <span className="truncate">{document?.doc_name ?? 'Unnamed Document'}</span>
-                </Button>
-              </Tooltip>
-            </Flex>
-          )}
-        </div>
-      </div>
-    );
-  }, [getReferenceInfo, handleDocumentButtonClick]);
-
-  const renderReference = useCallback((text: string) => {
-    return reactStringReplace(text, currentReg, (match, i) => {
-      const chunkIndex = getChunkIndex(match);
+  const getPopoverContent = useCallback(
+    (chunkIndex: number) => {
      const info = getReferenceInfo(chunkIndex);

      if (!info) {
-        return <Tooltip key={`err-tooltip-${i}`} title="Reference unavailable"><InfoCircleOutlined className={styles.referenceIcon} /></Tooltip>;
+        return (
+          <div className="p-2 text-xs text-red-500">
+            Error: Missing document information.
+          </div>
+        );
      }

-      const { imageId, chunkItem, documentId, fileExtension, documentUrl } = info;
-
-      if (showImage(chunkItem?.doc_type)) {
-        return <Image key={`img-${i}`} id={imageId} className="block object-contain max-w-full max-h-48 rounded my-2 cursor-pointer" onClick={handleDocumentButtonClick(documentId, chunkItem, fileExtension === 'pdf', documentUrl)} />;
-      }
+      const {
+        documentUrl,
+        fileThumbnail,
+        fileExtension,
+        imageId,
+        chunkItem,
+        documentId,
+        document,
+      } = info;

      return (
-        <Popover
-          content={getPopoverContent(chunkIndex)}
-          key={`popover-${i}`}
+        <div
+          key={`popover-content-${chunkItem.id}`}
+          className="flex gap-2 widget-citation-content"
        >
-          <InfoCircleOutlined className={styles.referenceIcon} />
-        </Popover>
+          {imageId && (
+            <Popover
+              placement="left"
+              content={
+                <Image
+                  id={imageId}
+                  className="max-w-[80vw] max-h-[60vh] rounded"
+                />
+              }
+            >
+              <Image
+                id={imageId}
+                className="w-24 h-24 object-contain rounded m-1 cursor-pointer"
+              />
+            </Popover>
+          )}
+          <div className="space-y-2 flex-1 min-w-0">
+            <div
+              dangerouslySetInnerHTML={{
+                __html: DOMPurify.sanitize(chunkItem?.content ?? ''),
+              }}
+              className="max-h-[250px] overflow-y-auto text-xs leading-relaxed p-2 bg-gray-50 dark:bg-gray-800 rounded prose-sm"
+            ></div>
+            {documentId && (
+              <Flex gap={'small'} align="center">
+                {fileThumbnail ? (
+                  <img
+                    src={fileThumbnail}
+                    alt={document?.doc_name}
+                    className="w-6 h-6 rounded"
+                  />
+                ) : (
+                  <SvgIcon name={`file-icon/${fileExtension}`} width={20} />
+                )}
+                <Tooltip
+                  title={
+                    !documentUrl && fileExtension !== 'pdf'
+                      ? 'Document link unavailable'
+                      : document.doc_name
+                  }
+                >
+                  <Button
+                    type="link"
+                    size="small"
+                    className="p-0 text-xs break-words h-auto text-left flex-1"
+                    onClick={handleDocumentButtonClick(
+                      documentId,
+                      chunkItem,
+                      fileExtension === 'pdf',
+                      documentUrl,
+                    )}
+                    disabled={!documentUrl && fileExtension !== 'pdf'}
+                    style={{ whiteSpace: 'normal' }}
+                  >
+                    <span className="truncate">
+                      {document?.doc_name ?? 'Unnamed Document'}
+                    </span>
+                  </Button>
+                </Tooltip>
+              </Flex>
+            )}
+          </div>
+        </div>
      );
-    });
-  }, [getPopoverContent, getReferenceInfo, handleDocumentButtonClick]);
+    },
+    [getReferenceInfo, handleDocumentButtonClick],
+  );
+
+  const renderReference = useCallback(
+    (text: string) => {
+      return reactStringReplace(text, currentReg, (match, i) => {
+        const chunkIndex = getChunkIndex(match);
+        const info = getReferenceInfo(chunkIndex);
+
+        if (!info) {
+          return (
+            <Tooltip key={`err-tooltip-${i}`} title="Reference unavailable">
+              <InfoCircleOutlined className={styles.referenceIcon} />
+            </Tooltip>
+          );
+        }
+
+        const { imageId, chunkItem, documentId, fileExtension, documentUrl } =
+          info;
+
+        if (showImage(chunkItem?.doc_type)) {
+          return (
+            <Image
+              key={`img-${i}`}
+              id={imageId}
+              className="block object-contain max-w-full max-h-48 rounded my-2 cursor-pointer"
+              onClick={handleDocumentButtonClick(
+                documentId,
+                chunkItem,
+                fileExtension === 'pdf',
+                documentUrl,
+              )}
+            />
+          );
+        }
+
+        return (
+          <Popover content={getPopoverContent(chunkIndex)} key={`popover-${i}`}>
+            <InfoCircleOutlined className={styles.referenceIcon} />
+          </Popover>
+        );
+      });
+    },
+    [getPopoverContent, getReferenceInfo, handleDocumentButtonClick],
+  );

  return (
    <div className="floating-chat-widget">
@ -167,28 +275,38 @@ const FloatingChatWidgetMarkdown = ({
        rehypePlugins={[rehypeWrapReference, rehypeKatex, rehypeRaw]}
        remarkPlugins={[remarkGfm, remarkMath]}
        className="text-sm leading-relaxed space-y-2 prose-sm max-w-full"
-        components={{
-          'custom-typography': ({ children }: { children: string }) => renderReference(children),
-          code(props: any) {
-            const { children, className, node, ...rest } = props;
-            const match = /language-(\w+)/.exec(className || '');
-            return match ? (
-              <SyntaxHighlighter
-                {...omit(rest, 'inline')}
-                PreTag="div"
-                language={match[1]}
-                style={isDarkTheme ? oneDark : oneLight}
-                wrapLongLines
-              >
-                {String(children).replace(/\n$/, '')}
-              </SyntaxHighlighter>
-            ) : (
-              <code {...rest} className={classNames(className, 'text-wrap text-xs bg-gray-200 dark:bg-gray-700 px-1 py-0.5 rounded')}>
-                {children}
-              </code>
-            );
-          },
-        } as any}
+        components={
+          {
+            'custom-typography': ({ children }: { children: string }) =>
+              renderReference(children),
+            code(props: any) {
+              // eslint-disable-next-line @typescript-eslint/no-unused-vars
+              const { children, className, node, ...rest } = props;
+              const match = /language-(\w+)/.exec(className || '');
+              return match ? (
+                <SyntaxHighlighter
+                  {...omit(rest, 'inline')}
+                  PreTag="div"
+                  language={match[1]}
+                  style={isDarkTheme ? oneDark : oneLight}
+                  wrapLongLines
+                >
+                  {String(children).replace(/\n$/, '')}
+                </SyntaxHighlighter>
+              ) : (
+                <code
+                  {...rest}
+                  className={classNames(
+                    className,
+                    'text-wrap text-xs bg-gray-200 dark:bg-gray-700 px-1 py-0.5 rounded',
+                  )}
+                >
+                  {children}
+                </code>
+              );
+            },
+          } as any
+        }
      >
        {contentWithCursor}
      </Markdown>
@ -196,4 +314,4 @@ const FloatingChatWidgetMarkdown = ({
  );
 };

-export default FloatingChatWidgetMarkdown;
+export default FloatingChatWidgetMarkdown;
--- a/web/src/components/floating-chat-widget.tsx
+++ b/web/src/components/floating-chat-widget.tsx
@ -1,18 +1,10 @@
-import { MessageType, SharedFrom } from '@/constants/chat';
-import { useFetchNextConversationSSE } from '@/hooks/chat-hooks';
-import { useFetchFlowSSE } from '@/hooks/flow-hooks';
-import { useFetchExternalChatInfo } from '@/hooks/use-chat-request';
+import PdfDrawer from '@/components/pdf-drawer';
 import { useClickDrawer } from '@/components/pdf-drawer/hooks';
+import { MessageType } from '@/constants/chat';
+import { useFetchExternalChatInfo } from '@/hooks/use-chat-request';
 import i18n from '@/locales/config';
 import { MessageCircle, Minimize2, Send, X } from 'lucide-react';
-import PdfDrawer from '@/components/pdf-drawer';
-import React, {
-  useCallback,
-  useEffect,
-  useMemo,
-  useRef,
-  useState,
-} from 'react';
+import React, { useCallback, useEffect, useRef, useState } from 'react';
 import {
  useGetSharedChatSearchParams,
  useSendSharedMessage,
@ -28,12 +20,7 @@ const FloatingChatWidget = () => {
  const [isLoaded, setIsLoaded] = useState(false);
  const messagesEndRef = useRef<HTMLDivElement>(null);

-  const {
-    sharedId: conversationId,
-    from,
-    locale,
-    visibleAvatar,
-  } = useGetSharedChatSearchParams();
+  const { sharedId: conversationId, locale } = useGetSharedChatSearchParams();

  // Check if we're in button-only mode or window-only mode
  const urlParams = new URLSearchParams(window.location.search);
@ -58,14 +45,6 @@ const FloatingChatWidget = () => {

  const { data: chatInfo } = useFetchExternalChatInfo();

-  const useFetchAvatar = useMemo(() => {
-    return from === SharedFrom.Agent
-      ? useFetchFlowSSE
-      : useFetchNextConversationSSE;
-  }, [from]);
-
-  const { data: avatarData } = useFetchAvatar();
-
  const { visible, hideModal, documentId, selectedChunk, clickDocumentButton } =
    useClickDrawer();

@ -181,6 +160,40 @@ const FloatingChatWidget = () => {
    messagesEndRef.current?.scrollIntoView({ behavior: 'smooth' });
  }, [displayMessages]);

+  // Render different content based on mode
+  // Master mode - handles everything and creates second iframe dynamically
+  useEffect(() => {
+    if (mode !== 'master') return;
+    // Create the chat window iframe dynamically when needed
+    const createChatWindow = () => {
+      // Check if iframe already exists in parent document
+      window.parent.postMessage(
+        {
+          type: 'CREATE_CHAT_WINDOW',
+          src: window.location.href.replace('mode=master', 'mode=window'),
+        },
+        '*',
+      );
+    };
+
+    createChatWindow();
+
+    // Listen for our own toggle events to show/hide the dynamic iframe
+    const handleToggle = (e: MessageEvent) => {
+      if (e.source === window) return; // Ignore our own messages
+
+      const chatWindow = document.getElementById(
+        'dynamic-chat-window',
+      ) as HTMLIFrameElement;
+      if (chatWindow && e.data.type === 'TOGGLE_CHAT') {
+        chatWindow.style.display = e.data.isOpen ? 'block' : 'none';
+      }
+    };
+
+    window.addEventListener('message', handleToggle);
+    return () => window.removeEventListener('message', handleToggle);
+  }, [mode]);
+
  // Play sound only when AI response is complete (not streaming chunks)
  useEffect(() => {
    if (derivedMessages && derivedMessages.length > 0 && !sendLoading) {
@ -234,7 +247,7 @@ const FloatingChatWidget = () => {
    const syntheticEvent = {
      target: { value: inputValue },
      currentTarget: { value: inputValue },
-      preventDefault: () => { },
+      preventDefault: () => {},
    } as any;

    handleInputChange(syntheticEvent);
@ -271,46 +284,14 @@ const FloatingChatWidget = () => {

  const messageCount = displayMessages?.length || 0;

-  // Render different content based on mode
+  // Show just the button in master mode
  if (mode === 'master') {
-    // Master mode - handles everything and creates second iframe dynamically
-    useEffect(() => {
-      // Create the chat window iframe dynamically when needed
-      const createChatWindow = () => {
-        // Check if iframe already exists in parent document
-        window.parent.postMessage(
-          {
-            type: 'CREATE_CHAT_WINDOW',
-            src: window.location.href.replace('mode=master', 'mode=window'),
-          },
-          '*',
-        );
-      };
-
-      createChatWindow();
-
-      // Listen for our own toggle events to show/hide the dynamic iframe
-      const handleToggle = (e: MessageEvent) => {
-        if (e.source === window) return; // Ignore our own messages
-
-        const chatWindow = document.getElementById(
-          'dynamic-chat-window',
-        ) as HTMLIFrameElement;
-        if (chatWindow && e.data.type === 'TOGGLE_CHAT') {
-          chatWindow.style.display = e.data.isOpen ? 'block' : 'none';
-        }
-      };
-
-      window.addEventListener('message', handleToggle);
-      return () => window.removeEventListener('message', handleToggle);
-    }, []);
-
-    // Show just the button in master mode
    return (
      <div
        className={`fixed bottom-6 right-6 z-50 transition-opacity duration-300 ${isLoaded ? 'opacity-100' : 'opacity-0'}`}
      >
        <button
+          type="button"
          onClick={() => {
            const newIsOpen = !isOpen;
            setIsOpen(newIsOpen);
@ -325,8 +306,9 @@ const FloatingChatWidget = () => {
              '*',
            );
          }}
-          className={`w-14 h-14 bg-blue-600 hover:bg-blue-700 text-white rounded-full transition-all duration-300 flex items-center justify-center group ${isOpen ? 'scale-95' : 'scale-100 hover:scale-105'
-            }`}
+          className={`w-14 h-14 bg-blue-600 hover:bg-blue-700 text-white rounded-full transition-all duration-300 flex items-center justify-center group ${
+            isOpen ? 'scale-95' : 'scale-100 hover:scale-105'
+          }`}
        >
          <div
            className={`transition-transform duration-300 ${isOpen ? 'rotate-45' : 'rotate-0'}`}
@ -352,9 +334,11 @@ const FloatingChatWidget = () => {
        className={`fixed bottom-6 right-6 z-50 transition-opacity duration-300 ${isLoaded ? 'opacity-100' : 'opacity-0'}`}
      >
        <button
+          type="button"
          onClick={toggleChat}
-          className={`w-14 h-14 bg-blue-600 hover:bg-blue-700 text-white rounded-full transition-all duration-300 flex items-center justify-center group ${isOpen ? 'scale-95' : 'scale-100 hover:scale-105'
-            }`}
+          className={`w-14 h-14 bg-blue-600 hover:bg-blue-700 text-white rounded-full transition-all duration-300 flex items-center justify-center group ${
+            isOpen ? 'scale-95' : 'scale-100 hover:scale-105'
+          }`}
        >
          <div
            className={`transition-transform duration-300 ${isOpen ? 'rotate-45' : 'rotate-0'}`}
@ -431,10 +415,11 @@ const FloatingChatWidget = () => {
                  className={`flex ${message.role === MessageType.User ? 'justify-end' : 'justify-start'}`}
                >
                  <div
-                    className={`max-w-[280px] px-4 py-2 rounded-2xl ${message.role === MessageType.User
-                      ? 'bg-blue-600 text-white rounded-br-md'
-                      : 'bg-gray-100 text-gray-800 rounded-bl-md'
-                      }`}
+                    className={`max-w-[280px] px-4 py-2 rounded-2xl ${
+                      message.role === MessageType.User
+                        ? 'bg-blue-600 text-white rounded-br-md'
+                        : 'bg-gray-100 text-gray-800 rounded-bl-md'
+                    }`}
                  >
                    {message.role === MessageType.User ? (
                      <p className="text-sm leading-relaxed whitespace-pre-wrap">
@ -444,7 +429,13 @@ const FloatingChatWidget = () => {
                      <FloatingChatWidgetMarkdown
                        loading={false}
                        content={message.content}
-                        reference={message.reference || { doc_aggs: [], chunks: [], total: 0 }}
+                        reference={
+                          message.reference || {
+                            doc_aggs: [],
+                            chunks: [],
+                            total: 0,
+                          }
+                        }
                        clickDocumentButton={clickDocumentButton}
                      />
                    )}
@ -486,12 +477,13 @@ const FloatingChatWidget = () => {
                    onKeyPress={handleKeyPress}
                    placeholder="Type your message..."
                    rows={1}
-                    className="w-full resize-none border border-gray-300 rounded-2xl px-4 py-3 text-sm focus:outline-none focus:ring-2 focus:ring-blue-500 focus:border-transparent"
+                    className="w-full resize-none border border-gray-300 rounded-2xl px-4 py-3 text-sm focus:outline-none focus:ring-2 focus:ring-blue-500 focus:border-transparent text-black"
                    style={{ minHeight: '44px', maxHeight: '120px' }}
                    disabled={hasError || sendLoading}
                  />
                </div>
                <button
+                  type="button"
                  onClick={handleSendMessage}
                  disabled={!inputValue.trim() || sendLoading}
                  className="p-3 bg-blue-600 text-white rounded-full hover:bg-blue-700 disabled:opacity-50 disabled:cursor-not-allowed transition-colors"
@ -512,7 +504,7 @@ const FloatingChatWidget = () => {
        />
      </>
    );
-  }  // Full mode - render everything together (original behavior)
+  } // Full mode - render everything together (original behavior)
  return (
    <div
      className={`transition-opacity duration-300 ${isLoaded ? 'opacity-100' : 'opacity-0'}`}
@ -520,8 +512,9 @@ const FloatingChatWidget = () => {
      {/* Chat Widget Container */}
      {isOpen && (
        <div
-          className={`fixed bottom-24 right-6 z-50 bg-blue-600 rounded-2xl transition-all duration-300 ease-out ${isMinimized ? 'h-16' : 'h-[500px]'
-            } w-[380px] overflow-hidden`}
+          className={`fixed bottom-24 right-6 z-50 bg-blue-600 rounded-2xl transition-all duration-300 ease-out ${
+            isMinimized ? 'h-16' : 'h-[500px]'
+          } w-[380px] overflow-hidden`}
        >
          {/* Header */}
          <div className="flex items-center justify-between p-4 bg-gradient-to-r from-blue-600 to-blue-700 text-white rounded-t-2xl">
@ -540,12 +533,14 @@ const FloatingChatWidget = () => {
            </div>
            <div className="flex items-center space-x-1">
              <button
+                type="button"
                onClick={minimizeChat}
                className="p-1.5 hover:bg-white hover:bg-opacity-20 rounded-full transition-colors"
              >
                <Minimize2 size={16} />
              </button>
              <button
+                type="button"
                onClick={toggleChat}
                className="p-1.5 hover:bg-white hover:bg-opacity-20 rounded-full transition-colors"
              >
@ -592,10 +587,11 @@ const FloatingChatWidget = () => {
                    className={`flex ${message.role === MessageType.User ? 'justify-end' : 'justify-start'}`}
                  >
                    <div
-                      className={`max-w-[280px] px-4 py-2 rounded-2xl ${message.role === MessageType.User
-                        ? 'bg-blue-600 text-white rounded-br-md'
-                        : 'bg-gray-100 text-gray-800 rounded-bl-md'
-                        }`}
+                      className={`max-w-[280px] px-4 py-2 rounded-2xl ${
+                        message.role === MessageType.User
+                          ? 'bg-blue-600 text-white rounded-br-md'
+                          : 'bg-gray-100 text-gray-800 rounded-bl-md'
+                      }`}
                    >
                      {message.role === MessageType.User ? (
                        <p className="text-sm leading-relaxed whitespace-pre-wrap">
@ -605,7 +601,13 @@ const FloatingChatWidget = () => {
                        <FloatingChatWidgetMarkdown
                          loading={false}
                          content={message.content}
-                          reference={message.reference || { doc_aggs: [], chunks: [], total: 0 }}
+                          reference={
+                            message.reference || {
+                              doc_aggs: [],
+                              chunks: [],
+                              total: 0,
+                            }
+                          }
                          clickDocumentButton={clickDocumentButton}
                        />
                      )}
@ -650,12 +652,13 @@ const FloatingChatWidget = () => {
                      onKeyPress={handleKeyPress}
                      placeholder="Type your message..."
                      rows={1}
-                      className="w-full resize-none border border-gray-300 rounded-2xl px-4 py-3 text-sm focus:outline-none focus:ring-2 focus:ring-blue-500 focus:border-transparent"
+                      className="w-full resize-none border border-gray-300 rounded-2xl px-4 py-3 text-sm focus:outline-none focus:ring-2 focus:ring-blue-500 focus:border-transparent text-black"
                      style={{ minHeight: '44px', maxHeight: '120px' }}
                      disabled={hasError || sendLoading}
                    />
                  </div>
                  <button
+                    type="button"
                    onClick={handleSendMessage}
                    disabled={!inputValue.trim() || sendLoading}
                    className="p-3 bg-blue-600 text-white rounded-full hover:bg-blue-700 disabled:opacity-50 disabled:cursor-not-allowed transition-colors"
@ -672,9 +675,11 @@ const FloatingChatWidget = () => {
      {/* Floating Button */}
      <div className="fixed bottom-6 right-6 z-50">
        <button
+          type="button"
          onClick={toggleChat}
-          className={`w-14 h-14 bg-blue-600 hover:bg-blue-700 text-white rounded-full transition-all duration-300 flex items-center justify-center group ${isOpen ? 'scale-95' : 'scale-100 hover:scale-105'
-            }`}
+          className={`w-14 h-14 bg-blue-600 hover:bg-blue-700 text-white rounded-full transition-all duration-300 flex items-center justify-center group ${
+            isOpen ? 'scale-95' : 'scale-100 hover:scale-105'
+          }`}
        >
          <div
            className={`transition-transform duration-300 ${isOpen ? 'rotate-45' : 'rotate-0'}`}
--- a/web/src/components/home-card.tsx
+++ b/web/src/components/home-card.tsx
@ -24,7 +24,6 @@ export function HomeCard({
 }: IProps) {
  return (
    <Card
-      className="bg-bg-card  border-colors-outline-neutral-standard"
      onClick={() => {
        // navigateToSearch(data?.id);
        onClick?.();
--- a/web/src/components/layout-recognize-form-field.tsx
+++ b/web/src/components/layout-recognize-form-field.tsx
@ -17,6 +17,7 @@ import {
 export const enum ParseDocumentType {
  DeepDOC = 'DeepDOC',
  PlainText = 'Plain Text',
+  MinerU = 'MinerU',
 }

 export function LayoutRecognizeFormField({
@ -38,9 +39,12 @@ export function LayoutRecognizeFormField({
  const options = useMemo(() => {
    const list = optionsWithoutLLM
      ? optionsWithoutLLM
-      : [ParseDocumentType.DeepDOC, ParseDocumentType.PlainText].map((x) => ({
-          label:
-            x === ParseDocumentType.PlainText ? t(camelCase(x)) : 'DeepDoc',
+      : [
+          ParseDocumentType.DeepDOC,
+          ParseDocumentType.PlainText,
+          ParseDocumentType.MinerU,
+        ].map((x) => ({
+          label: x === ParseDocumentType.PlainText ? t(camelCase(x)) : x,
          value: x,
        }));

--- a/web/src/components/parse-configuration/graph-rag-form-fields.tsx
+++ b/web/src/components/parse-configuration/graph-rag-form-fields.tsx
@ -70,7 +70,7 @@ export function UseGraphRagFormField({
    <FormField
      control={form.control}
      name="parser_config.graphrag.use_graphrag"
-      render={({ field }) => (
+      render={() => (
        <FormItem defaultChecked={false} className=" items-center space-y-0 ">
          <div className="flex items-center gap-1">
            <FormLabel
--- a/web/src/components/theme-toggle.tsx
+++ b/web/src/components/theme-toggle.tsx
@ -0,0 +1,48 @@
+import { ThemeEnum } from '@/constants/common';
+import { Moon, Sun } from 'lucide-react';
+import { FC, useCallback } from 'react';
+import { useIsDarkTheme, useTheme } from './theme-provider';
+import { Button } from './ui/button';
+
+const ThemeToggle: FC = () => {
+  const { setTheme } = useTheme();
+  const isDarkTheme = useIsDarkTheme();
+  const handleThemeChange = useCallback(
+    (checked: boolean) => {
+      setTheme(checked ? ThemeEnum.Dark : ThemeEnum.Light);
+    },
+    [setTheme],
+  );
+  return (
+    <Button
+      type="button"
+      onClick={() => handleThemeChange(!isDarkTheme)}
+      className="relative inline-flex h-6 w-14 items-center rounded-full  transition-colors p-0.5 border-none focus:border-none bg-bg-card hover:bg-bg-card"
+      //   aria-label={isDarkTheme ? 'Switch to light mode' : 'Switch to dark mode'}
+    >
+      <div className="inline-flex h-full w-full items-center">
+        <div
+          className={`inline-flex transform items-center justify-center rounded-full  transition-transform ${
+            isDarkTheme
+              ? '  text-text-disabled h-4 w-5'
+              : '  text-text-primary bg-bg-base h-full w-8 flex-1'
+          }`}
+        >
+          <Sun />
+        </div>
+
+        <div
+          className={`inline-flex  transform items-center justify-center rounded-full  transition-transform ${
+            isDarkTheme
+              ? ' text-text-primary bg-bg-base h-full w-8 flex-1'
+              : 'text-text-disabled h-4 w-5'
+          }`}
+        >
+          <Moon />
+        </div>
+      </div>
+    </Button>
+  );
+};
+
+export default ThemeToggle;
--- a/web/src/components/ui/card.tsx
+++ b/web/src/components/ui/card.tsx
@ -8,7 +8,10 @@ const Card = React.forwardRef<
 >(({ className, ...props }, ref) => (
  <div
    ref={ref}
-    className={cn('rounded-lg  bg-bg-card  shadow-sm', className)}
+    className={cn(
+      'rounded-lg border-border-default border shadow-sm bg-bg-input',
+      className,
+    )}
    {...props}
  />
 ));
--- a/web/src/components/ui/dialog.tsx
+++ b/web/src/components/ui/dialog.tsx
@ -73,7 +73,7 @@ const DialogFooter = ({
 }: React.HTMLAttributes<HTMLDivElement>) => (
  <div
    className={cn(
-      'flex flex-col-reverse sm:flex-row sm:justify-end sm:space-x-2',
+      'flex flex-col-reverse sm:flex-row sm:justify-end sm:space-x-4',
      className,
    )}
    {...props}
--- a/web/src/components/ui/form.tsx
+++ b/web/src/components/ui/form.tsx
@ -106,8 +106,10 @@ const FormLabel = React.forwardRef<
      htmlFor={formItemId}
      {...props}
    >
-      {required && <span className="text-destructive">*</span>}
-      {props.children}
+      <section>
+        {required && <span className="text-destructive">*</span>}
+        {props.children}
+      </section>
      {tooltip && <FormTooltip tooltip={tooltip}></FormTooltip>}
    </Label>
  );
--- a/web/src/components/ui/label.tsx
+++ b/web/src/components/ui/label.tsx
@ -7,7 +7,7 @@ import * as React from 'react';
 import { cn } from '@/lib/utils';

 const labelVariants = cva(
-  'text-sm font-medium leading-none peer-disabled:cursor-not-allowed peer-disabled:opacity-70',
+  'text-sm font-medium leading-none peer-disabled:cursor-not-allowed peer-disabled:opacity-70 text-text-secondary',
 );

 const Label = React.forwardRef<
--- a/web/src/components/ui/modal/modal.tsx
+++ b/web/src/components/ui/modal/modal.tsx
@ -116,7 +116,10 @@ const Modal: ModalType = ({
            type="button"
            disabled={confirmLoading || disabled}
            onClick={() => handleOk()}
-            className="px-2 py-1 bg-primary text-primary-foreground rounded-md hover:bg-primary/90"
+            className={cn(
+              'px-2 py-1 bg-primary text-primary-foreground rounded-md hover:bg-primary/90',
+              { 'cursor-not-allowed': disabled },
+            )}
          >
            {confirmLoading && (
              <Loader className="inline-block mr-2 h-4 w-4 animate-spin" />
@ -137,6 +140,7 @@ const Modal: ModalType = ({
      </div>
    );
  }, [
+    disabled,
    footer,
    cancelText,
    t,
@ -155,7 +159,7 @@ const Modal: ModalType = ({
          onClick={() => maskClosable && onOpenChange?.(false)}
        >
          <DialogPrimitive.Content
-            className={`relative w-[700px] ${full ? 'max-w-full' : sizeClasses[size]} ${className} bg-colors-background-neutral-standard rounded-lg shadow-lg border transition-all focus-visible:!outline-none`}
+            className={`relative w-[700px] ${full ? 'max-w-full' : sizeClasses[size]} ${className} bg-bg-base rounded-lg shadow-lg border border-border-default transition-all focus-visible:!outline-none`}
            onClick={(e) => e.stopPropagation()}
          >
            {/* title */}
--- a/web/src/components/ui/spin.tsx
+++ b/web/src/components/ui/spin.tsx
@ -14,6 +14,12 @@ const sizeClasses = {
  large: 'w-8 h-8',
 };

+const minSizeClasses = {
+  small: 'min-w-4 min-h-4',
+  default: 'min-w-6 min-h-6',
+  large: 'min-w-8 min-h-8',
+};
+
 export const Spin: React.FC<SpinProps> = ({
  spinning = true,
  size = 'default',
@ -32,7 +38,12 @@ export const Spin: React.FC<SpinProps> = ({
      )}
    >
      {spinning && (
-        <div className="absolute inset-0 z-10 flex items-center justify-center bg-text-primary/30 ">
+        <div
+          className={cn(
+            'absolute inset-0 z-10 flex items-center justify-center bg-text-primary/30',
+            minSizeClasses[size],
+          )}
+        >
          <div
            className={cn(
              'rounded-full border-muted-foreground border-2 border-t-transparent animate-spin',
--- a/web/src/constants/agent.ts
+++ b/web/src/constants/agent.ts
@ -53,6 +53,10 @@ export enum AgentCategory {
  DataflowCanvas = 'dataflow_canvas',
 }

+export enum AgentQuery {
+  Category = 'category',
+}
+
 export enum DataflowOperator {
  Begin = 'File',
  Note = 'Note',
@ -62,3 +66,55 @@ export enum DataflowOperator {
  HierarchicalMerger = 'HierarchicalMerger',
  Extractor = 'Extractor',
 }
+
+export enum Operator {
+  Begin = 'Begin',
+  Retrieval = 'Retrieval',
+  Categorize = 'Categorize',
+  Message = 'Message',
+  Relevant = 'Relevant',
+  RewriteQuestion = 'RewriteQuestion',
+  KeywordExtract = 'KeywordExtract',
+  Baidu = 'Baidu',
+  DuckDuckGo = 'DuckDuckGo',
+  Wikipedia = 'Wikipedia',
+  PubMed = 'PubMed',
+  ArXiv = 'ArXiv',
+  Google = 'Google',
+  Bing = 'Bing',
+  GoogleScholar = 'GoogleScholar',
+  DeepL = 'DeepL',
+  GitHub = 'GitHub',
+  BaiduFanyi = 'BaiduFanyi',
+  QWeather = 'QWeather',
+  ExeSQL = 'ExeSQL',
+  Switch = 'Switch',
+  WenCai = 'WenCai',
+  AkShare = 'AkShare',
+  YahooFinance = 'YahooFinance',
+  Jin10 = 'Jin10',
+  TuShare = 'TuShare',
+  Note = 'Note',
+  Crawler = 'Crawler',
+  Invoke = 'Invoke',
+  Email = 'Email',
+  Iteration = 'Iteration',
+  IterationStart = 'IterationItem',
+  Code = 'CodeExec',
+  WaitingDialogue = 'WaitingDialogue',
+  Agent = 'Agent',
+  Tool = 'Tool',
+  TavilySearch = 'TavilySearch',
+  TavilyExtract = 'TavilyExtract',
+  UserFillUp = 'UserFillUp',
+  StringTransform = 'StringTransform',
+  SearXNG = 'SearXNG',
+  Placeholder = 'Placeholder',
+  File = 'File', // pipeline
+  Parser = 'Parser',
+  Tokenizer = 'Tokenizer',
+  Splitter = 'Splitter',
+  HierarchicalMerger = 'HierarchicalMerger',
+  Extractor = 'Extractor',
+  Generate = 'Generate',
+}
--- a/web/src/hooks/common-hooks.tsx
+++ b/web/src/hooks/common-hooks.tsx
@ -94,9 +94,9 @@ export const useShowDeleteConfirm = () => {
          title: title ?? t('common.deleteModalTitle'),
          icon: <ExclamationCircleFilled />,
          content,
-          okText: t('common.ok'),
+          okText: t('common.yes'),
          okType: 'danger',
-          cancelText: t('common.cancel'),
+          cancelText: t('common.no'),
          async onOk() {
            try {
              const ret = await onOk?.();
--- a/Show More
+++ b/Show More