After setup filebrain fails to run because containers are not initialized

[jozamm]

Hi,

I am trying to get filebrain to run because I need to search inteligently through lots of pdf's

During the setup phase both the tika and typesense containers are loaded and the setup sees them as healthy and the setup process finishes sucessfully.

However when I run file brain again, it says that the docker desktop is not running and in fact both containers are not loaded. There is a screenshot at the end.

How can the problem be resolved? Maybe if during startup the containers are not found, they are reloaded?

As a note, I am running docker in a virtual box VM and then pointed the docker CLI to the Virtual machine..

Thanks,

Joseph

[Hamza5]

Hi Joseph,

Thanks for reporting the problem.

I have never used the setup of running Docker in a VM, since Docker already isolates the container content from the host. On Linux, I run Docker natively. On Windows, I set it to use WSL2.

However, I have pushed an experimental version (0.1.26a2) that may solve this problem. You can upgrade to it using:

pip install -U --pre file-brain

(notice the --pre flag)

After upgrading, you can run the app again. If it still does not work, please copy the logs you see in the terminal/console and paste them here.

[jozamm]

Hi,

I have installed the latest experimental version and it is working like a charm. I am currently indexing files and searching.

My use case is search through hundreds of files related to electronics projects, so I wish to search on keywords and more complex semantics e.g. find a microcontroller project. I tried to use a RAG search + LLM but the results very disapointing.

However I had to do the following 2 modifications

1. Change the .yml file to point to 0.0.0.0 instead of 127.0.0.1
2. Change the python code so that the tika server and typsense images server use the ip address of the virtual machine instead of 127.0.0.1. Maybe this is for my particular configuation.

The software is working very well except for the following

1. After the search, when one selects a file, and clocks open or folder, these buttons dont work
2. I had to install java because there was a log error that java was missing
3. There are some errors as follows in the log

2026-03-14 19:09:34,120 [indexing_wor] [ERROR] Failed to receive startup confirmation from startServer.
2026-03-14 19:16:42,731 - file_brain - WARNING - Tika detection failed: Unable to start Tika server.
Tika server timeouts

Some features I would like to see

1. One can open the pdf file at the page where the match was found
2. The indexing would be more 'intelligent' for example recognizing table of contents

Hope this feedback works. I think I will invest in the paid editions of the software matches my needs.

Regards,

Joseph

[Hamza5]

Hi Joseph,

Regarding the changes you had to make to 127.0.0.1, I think this is because of your setup. Since Docker is not running on the same machine (because it is in the VM, which has its own "network card").

In the old versions of File Brain (before 0.1.23), it was listening for all networks (like what you did in your modification), but one developer reported that this is a security vulnerability, since anybody on the local network can perform a DoS attack by sending huge documents to Tika. That's why I restricted it to the localhost only.

Regarding the three problematic points:

1. Yes, it is a new bug! It was working before, but it seems that I broke something in the new versions. Thank you for reporting that. I will work on it, then tell you when I release a new version.
2. I think you had to install Java only because now you are running Tika locally (the python-tika library used in the code automatically downloads the Tika server when the server address is not supplied, which is actually, but given your modifications on the code, I think you changed something that caused that). If you didn't change the code, that library should use Tika from the Docker container. BTW, if you are using Tika outside the container, OCR will not work unless Tesseract is already installed on your computer and on PATH.
3. Same, I think these problems are caused by your modifications, and it might suggest that you are using a local Tika server, as I said in the previous point.

Regarding the features:

1. Acknowledged. Although there is no universal way to open a file at a specific page, I think I can add support to the most popular software that supports this via the command line (Like Adobe Acrobat and Foxit Reader). I will also change the indexing process to save the location where the match was found.
2. So, what should the indexing do when recognizing the table of contents? Skip it? I thought that File Brain is a generic search software, so the user expects it to find where the search query was mentioned, so if the match happens to be in the ToC, I consider that a relevant file too.
   BTW, File Brain is optimized to find the first match in each file, not all matches in every file. This is why it can return the search results very quickly (should not take more than a second), at the cost of lower precision and potential irrelevant results.

Anyway, thanks a lot for your detailed feedback. I will continue working on the points you mentioned.

[Hamza5]

BTW, I have pushed a version (0.1.27) that fixes the file/folder opening problem, since it is the most urgent one and affects all users. Other improvements are coming in later versions.

Again, you can upgrade by running:

pip install -U file-brain

But for your case specifically, you might need to repeat the modifications to the relevant .yml and .py files to make it work with your Docker setup.

Or, if you have cloned the repository, you can commit your changes, then pull the recent changes and merge them to have an updated code working with your setup.