"PyMuPDF Plugin for Dify" (0.0.2) executes ".bat" files without reading them.

[CTD-Networks-CO-LTD]

Description of the bug

• PyMuPDF Plugin for Dify 0.0.2
• Dify 1.1.3

PyMuPDF.Plugin.for.Dify.0.0.2.executes.bat.files.without.reading.them.mp4

"PyMuPDF Plugin for Dify" (0.0.2) executes ".bat" files without reading them.
I want to extract the file without executing the ".bat" file.

How to reproduce the bug

1. I confirmed that "test_ping.bat", which uploads files, can be executed on Windows.

@echo off
cd %~dp0
SET IPADDR=192.168.1.254
ping %IPADDR%
pause

2. I entered "Explain this action"(この動作を説明して) in the prompt and executed the Define chatflow with "test_ping.bat".

3. "test_ping.bat" was sent to "PyMuPDF Tool".

4. "PyMuPDF Tool" sent the results of executing "test_ping.bat" to LLM, so LLM responded with the error content.

Conversation Log(会話ログ)
Input(入力)
{ "LLM": "GPT", "files": { "dify_model_identity": "__dify__file__", "id": null, "tenant_id": "7f018048-900d-48c3-988a-cfb96ddfd6d3", "type": "custom", "transfer_method": "local_file", "remote_url": "", "related_id": "3c5d1fd0-be1b-4d27-a56a-6427a3f609f9", "filename": "test_ping.bat", "extension": ".bat", "mime_type": "application/octet-stream", "size": 79, "url": "/files/3c5d1fd0-be1b-4d27-a56a-6427a3f609f9/file-preview?timestamp=1745283359&nonce=77bd5826be88a1e03d80890c021b2850&sign=qg0M3ZQsofNOM9aw_g7IAsbNMAWBXvY4W7gBvnkcdeM=" }, "sys.query": "この動作を説明して", "sys.files": [], "sys.conversation_id": "f8f0d5cf-9173-4595-949a-8eb873e9593e", "sys.user_id": "d8738074-f72c-4396-a659-9f98707e714f", "sys.dialogue_count": 0, "sys.app_id": "67097e3a-2cad-4f36-9d2b-3868676c8178", "sys.workflow_id": "a22684e9-2b73-46c9-9e98-df07e9a3e8a2", "sys.workflow_run_id": "d1e7643a-24b5-4145-9fa0-02e8f0820c90" }
Output(出力)
{ "answer": "Error processing test_ping.bat: Request URL is missing an 'http://' or 'https://' protocol." }

↓
(LLM Input)
↓
(LLM Output)

この資料は、「test_ping.bat」というバッチファイルを実行した際に発生したエラー内容を示しています。

【動作説明】
- 「test_ping.bat」は何らかのネットワーク通信やWebリクエスト（例：API通信やサーバー疎通確認）を自動実行するためのバッチファイルです。
- バッチファイル内でWebリクエストを行う際に、指定したURLの先頭に "http://" または "https://" といったプロトコル指定が抜けていたため、エラーが発生しています。

【エラー内容】
- "Request URL is missing an 'http://' or 'https://' protocol."
  - これは、例えば curl や wget などのコマンドでURLを指定する際、"localhost:8080/api" のようにプロトコル部分が省略されている場合に表示されるエラーメッセージです。

【まとめ】
- 「test_ping.bat」はネットワークテストや通信確認を目的としたバッチファイルですが、URLの記述ミスにより正しく動作しませんでした。
- 解決するには、バッチファイル中のURLを "http://" もしくは "https://" で始まるように修正する必要があります。

PyMuPDF version

Built from source ( yevanchen-pymupdf_0.0.2.difypkg )

Operating system

Linux (Dify on Docker)

Python version

3.12

[JorjMcKie]

The error message "Request URL is missing an 'http://' or 'https://' protocol." clearly is not a PyMuPDF error.
PyMuPDF can only read files from local disks or from memory - not from internet resources or the like.

[CTD-Networks-CO-LTD]

Hi, @JorjMcKie

Thank you for your reply.
and, Thank you for your reply to #4452.

Background is that we cannot to reproduce your environment without installing additional software ("dify" etc.). This is beyond the scope we can cover here.

-> I understand that very well, but even if this is a bug in Dify, I would like to ask for support from PyMuPDF!
So, for example, by reporting it to Dify's issues from the PyMuPDF camp.

By the way, who is the owner of "PyMuPDF Plugin for Dify"?

BTW: When trying to understand the provided JSON snippets it seems that you are trying to directly open a remote file. This is not supported by PyMuPDF. PyMuPDF can only access files on local disks or as file images resident in memory.
Therefore, non-local files must first be downloaded to either a local disk or provided as memory.

-> If you watch this video recording on Dify, you will see that,

PyMuPDF.Plugin.for.Dify.0.0.2.executes.bat.files.without.reading.them.mp4

You are not trying to open a remote file directly, but uploading a file on your local disk to Dify.

The other day, I was able to solve the problem with the standard "Document Extractor" block of Dify by using "PyMuPDF Plugin for Dify"! ... and I was just happy...
The "validation error" that occurred when selecting "Other file types" for the upload file has been resolved.(Dify 1.1.3 + Document Extractor and PyMuPDF Plugin) #17967

Please help us find a solution!

[JorjMcKie]

As a first action, I am going to remove your posts from the issues tab. It is a Discussions item.

[CTD-Networks-CO-LTD]

If "PyMuPDF Plugin for Dify" doesn't solve the problem,
I thought the only solution would be to add an extension to Dify's standard "Document Extractor"...so I tried that.

• Dify Discussions #18591

In the process, I deleted document_extractor/node.py in the Dify API container,
and got the same error output as the one reported here.
"answer": "Error processing xxx.pdf: Request URL is missing an 'http://' or 'https://' protocol."

In other words, it seems that "PyMuPDF Plugin for Dify" is not installed correctly in Dify.

-> I also tried to install it from a local package after downloading "yevanchen-pymupdf_0.0.2.difypkg". .
At that time, no installation errors were displayed...

By the way, where in Dify is "PyMuPDF Plugin for Dify" installed?
I searched for the API container but couldn't find it...

$ sudo docker exec -it docker-api-1 bash

root@:/app/api# find / -iname pymupdf
root@:/app/api#

root@:/app/api# ls
Dockerfile commands.py core factories mypy.ini schedule tests
README.md configs dify_app.py fields poetry.lock services
__pycache__ constants docker libs poetry.toml storage
app.py contexts events migrations pyproject.toml tasks
app_factory.py controllers extensions models pytest.ini templates

root@:/app/api# ls core/
__init__.py errors indexing_runner.py ops variables
__pycache__ extension llm_generator plugin workflow
agent external_data_tool memory prompt
app file model_manager.py provider_manager.py
callback_handler helper model_runtime rag
entities hosting_configuration.py moderation tools