Fast, free, self-hosted Artificial Intelligence Server for any platform, any language

CodeProject.AI Server is a locally installed, self-hosted, fast, free and Open Source Artificial Intelligence server for any platform, any language. No off-device or out of network data transfer, no messing around with dependencies, and able to be used from any platform, any language. Runs as a Windows Service or a Docker container.

 

Docker Image: https://hub.docker.com/r/codeproject/ai-server

I'm looking to get this set up on my Unraid box. What GPU did you use for this? Are you also using Blue Iris? If so, is Blue Iris installed on a VM in Unraid or on a separate Windows box? Thanks.

Hey, I think you need to add the paths and add a gpu template too. That would be really helpful. 

 

Like include the - - gpus all 

I updated the docker to latest version and when you look at the modules loaded it shows a couple modules that have updates available, when you click update, it will not allow you to update the modules.  Reviewing the logs and it shows the following error below. 

 

Module description for 'FaceProcessing' is invalid. A 'pre-installed' module can't be downloaded

On 8/15/2023 at 12:32 PM, JM2005 said:

I updated the docker to latest version and when you look at the modules loaded it shows a couple modules that have updates available, when you click update, it will not allow you to update the modules.  Reviewing the logs and it shows the following error below. 

 

Module description for 'FaceProcessing' is invalid. A 'pre-installed' module can't be downloaded

 

@JM2005,  it looks like there is an active issue listed on github for this https://github.com/codeproject/CodeProject.AI-Server/issues/59#issue-1857894572.

If you want to use a Coral USB TPU module, I got it to work by setting the docker to privileged....adding the device to the template (see screenshot below)....and then downloading the module in the webui. I got the idea from looking at how the Frigate template in CA handles it then doing it the same way.

 

Edited August 20, 20232 yr by lzrdking71

Looking for any additional help with CodeProject AI Unraid Docker and USB Coral.

 

Per previous post, I have added a device to the Docker by using value "/dev/bus/usb" and set the Docker to privileged mode.

 

When running "lsusb" in Unraid I was showing the Coral as "ID 1a6e:089a Global Unichip Corp. "

 

Per official instructions, I have installed and reinstalled the ObjectDetection (Coral) module several times, until it finally found the Coral which caused it to change how it appears in lsusb command to "ID 18d1:9302 Google Inc. "

 

I also see "objectdetection_coral_adapter.py: Edge TPU detected" in the CodeProjectg AI server log. 

 

However I am unable to run any images. When I try the first object detection I get the following error message: 

 

"RuntimeError: Encountered an unresolved custom op. Did you miss a custom op or delegate?Node number 11 (EdgeTpuDelegateForCustomOp) failed to invoke."

 

All subsequent attempts return:

 

"The interpreter is in use. Please try again later"

 

If I restart the Docker, or stop/start the Coral module, it goes back to that first error message. Any ideas?

Hi all,

 

Just updated the codeprojectai container today and it is not working with my gtx970 gpu anymore - haven't figured it out yet. Been working fine for 2 years. Rebooted etc. Nothing.

Changed back to CodeProject.AI_ServerGPU container app and this works fine.

Something in the new image that is not compatible with older gtx gpu's perhaps?

Please let me know if anyone runs CodeProject.AI_Server with a gtx 900 series to see if it detects it.
Mine defaults to CPU and can't seem to change it. It always picks up the gpu automatically.

 

Many thanks

Hi All,

 

The WebUI keeps reminding me of a later version being available but I cannot see anyway to update the solution, I'm assuming the docker hub needs to be updated to 2.5.x?

 

When does this usually happen, looks like the latest version has been out about a month?

 

Cheers

 

Michael

I was able to update the Docker image last night via the unraid/docker page, by clicking 'update' in the Version column, but I can't update modules (through the CodeProject GUI).  This github link from an ealier post describes my issue exactly but claims it was fixed in a release last October.  Anyone else having trouble when they try to update a module?

Edited February 16, 20242 yr by curtis-r

Now I'm (more) perplexed.  I uninstalled 2 modules I wasn't using, uninstalled the CodeProject Docker app entirely.  Manually deleted through MC the residual CodeProject Cache folder. Restarted unraid. 

 

Now when I reinstall the CodeProject Docker, it not only shows installed the Module I can't seem to update, but the other 2 modules I had uninstalled. This is on a fresh install! Where is it getting this ghost data from?

For those running CP.AI using the gpu docker container. It seems like CP hasn't been building it for about 5 months. What you can do is change it to a cuda container and have the latest version.

 

Simply update `codeproject/ai-server:gpu` to one of the following depending on your GPUs capabilities.

 

- `codeproject/ai-server:cuda11_7`

- `codeproject/ai-server:cuda12_2`

 

On 2/23/2024 at 2:02 PM, 5252525111 said:

On 2/23/2024 at 2:02 PM, 5252525111 said:

For those running CP.AI using the gpu docker container. It seems like CP hasn't been building it for about 5 months. What you can do is change it to a cuda container and have the latest version.

 

Simply update `codeproject/ai-server:gpu` to one of the following depending on your GPUs capabilities.

 

- `codeproject/ai-server:cuda11_7`

- `codeproject/ai-server:cuda12_2`

 

This worked thanks but still no mesh option on this one.

Regarding the mesh I have the standard codeproject/ai-server running with its mesh function. Ran up several other CPAI nodes (a few on lxc proxmox debian and alpine containers and one windows installation on a VM.

The unraid mesh does not communicate with any - added them to hostnames - they are visible but cannot ping them. UDP port enabled also.

The lxc containers read the windows cpai but not the other way around.

Would really like to get the mesh working on the unraid cpai container app.

What other security/firewall settings do I need to change to get them to communicate with each other?

Thanks

 

Edited March 22, 20242 yr by only-university6482

I am running codeproject/ai-server with the cuda12_2 tag, and it works well.

 

However, I've noticed that the container size is 11.1GB in size, when looking at the docker container size section. Is this normal?

 

 

 

Thanks

Can something assist in helping me set up a 2nd instance of codeproject please?
I have tried to add the 2nd instance with different ports/name/appdata name etc. All unique values.
But still the data and modules folders are empty on codeproject_ai-2 using 32168 for udp/tcp etc.
Even checked in portainer and tried to recreate from there with all 2nd instance unique values.
Says up and running but seems to be using the 1st port 32168 and not opening in 32170 for example.

Anyone had luck in setting up a 2nd instance?

Thanks

On 4/11/2024 at 5:22 AM, kyoumei said:

I am running codeproject/ai-server with the cuda12_2 tag, and it works well.

 

However, I've noticed that the container size is 11.1GB in size, when looking at the docker container size section. Is this normal?

 

 

 

Thanks

It's a large container. it is normal. 

 

On 4/23/2024 at 9:06 AM, only-university6482 said:

Can something assist in helping me set up a 2nd instance of codeproject please?
I have tried to add the 2nd instance with different ports/name/appdata name etc. All unique values.
But still the data and modules folders are empty on codeproject_ai-2 using 32168 for udp/tcp etc.
Even checked in portainer and tried to recreate from there with all 2nd instance unique values.
Says up and running but seems to be using the 1st port 32168 and not opening in 32170 for example.

Anyone had luck in setting up a 2nd instance?

Thanks

code project will report it is communicating on port 32168, but if you changed the port in the docker edit tab that is the port your unraid machine will be communicating to point to that container.  

 

 

On 3/22/2024 at 11:05 AM, only-university6482 said:

 

For the mesh you'll likely need to change from bridge to a different network. the issue is that the container is on it's own network I havn't tried but host may work. What i did tog et it to work for me was switch to custom network BR0 and gave it a static ip, and had my router assign it the same ip then the mesh worked.

I'm currently running this container with Repository: codeproject/ai-server:cuda12_2  I have a Tesla P4 installed. Object Detection (YOLOv5 6.2) 1.9.1 is running great with the GPU.  Does anybody know how to or if the License Plate Reader module can use the GPU?  Every time I try enabling GPU it reverts back to CPU.

 

Also if I update YOLOv5 6.2 to 1.9.2 it seems to break GPU functionality.

 

 

Edited July 29, 20241 yr by kimocal

On 7/28/2024 at 9:22 PM, kimocal said:

I'm currently running this container with Repository: codeproject/ai-server:cuda12_2  I have a Tesla P4 installed. Object Detection (YOLOv5 6.2) 1.9.1 is running great with the GPU.  Does anybody know how to or if the License Plate Reader module can use the GPU?  Every time I try enabling GPU it reverts back to CPU.

 

Also if I update YOLOv5 6.2 to 1.9.2 it seems to break GPU functionality.

 

 

Did you ever figure this out?

No. The inference times of the LPR on CPU are pretty quick. YOLO uses the GPU. It's working at the moment as is so I haven't tinkered with it since. Currently on 2.6.5.

Edited October 16, 20241 yr by kimocal

On 2/23/2024 at 5:02 PM, 5252525111 said:

For those running CP.AI using the gpu docker container. It seems like CP hasn't been building it for about 5 months. What you can do is change it to a cuda container and have the latest version.

 

Simply update `codeproject/ai-server:gpu` to one of the following depending on your GPUs capabilities.

 

- `codeproject/ai-server:cuda11_7`

- `codeproject/ai-server:cuda12_2`

 

Hi, I'm sorry, I'm not sure what this means. Could someone talk me through what I need to do up upgrade code project ai GPU server to a newer version within a docker in unraid. The version on my system is 2.2.4 beta, and docker says it's up to date. 

 

Thanks des 

Sorry for asking silly question guys, but I'm trying to figure this out.  The text " Simply update `codeproject/ai-server:gpu` to one of the following depending on your GPUs capabilities.

 

 

 

- `codeproject/ai-server:cuda11_7`

 

- `codeproject/ai-server:cuda12_2

"

Does it simply mean I edit the container details within docker?  At the moment I have the details attached. So change the repository and registry URL to one of the lines above?  

 

I currently have a Tesla p4 and also a Nvidia t400, so just want to pass one of these cards through to use.  Thanks 👍 

Just closing down my own question for future reference.  

 

I used the information here to get my system working.  https://www.reddit.com/r/unRAID/s/eGZ03UXtfi

 

Basically use the deep stack template and adjust it for codeProjectAI.  That allowed me to point to the cuda version and I can also install extra modules now, so have number plate reading working. Very pleased. Thanks 👍 

I'm having trouble getting the latest version of CodeProject.AI to recognize and use my GPU.

 

My current setup is Unraid 6.12.14. The CodeProject Docker container is using the codeproject/ai-server.gpu repository. Here is the System Info shown by the CodeProject WebUI:

 

Server version:   2.2.4-Beta
Operating System: Linux (Linux 6.1.118-Unraid #1 SMP PREEMPT_DYNAMIC Thu Nov 21 15:54:38 PST 2024)
CPUs:             Intel(R) Xeon(R) CPU E5-2690 v2 @ 3.00GHz (Intel)
                  1 CPU x 10 cores. 20 logical processors (x64)
GPU:              NVIDIA GeForce RTX 3060 (12 GiB) (NVIDIA) 
                  Driver: 565.57.01 CUDA: 12.7 (max supported: 12.7) Compute: 8.6
System RAM:       63 GiB
Target:           Linux
BuildConfig:      Release
Execution Env:    Docker
Runtime Env:      Production
.NET framework:   .NET 7.0.11
Video adapter info:
System GPU info:
  GPU 3D Usage       0%
  GPU RAM Usage      1.3 GiB
Global Environment variables:
  CPAI_APPROOTPATH = <root>
  CPAI_PORT        = 32168

 

The object detection module being used is YOLOv5 6.2. I'm using CodeProject as the AI for Blue Iris that is running on a VM on Unraid.

 

My current setup seems to be working ok. The CodeProject WebUI is showing that the 2.9.5 version of CodeProject is available. I've tried a number of things to update CodeProject to 2.9.5 (including changing the repository to codeproject/ai-server.cuda11_7, to codeproject/ai-server.cuda12_2 and changing the parameters in the deepstack container to the parameters used in the CodeProject container) but in each case, my GPU does not seem to be recognized and the new versions of CodeProject revert to using the CPU.

 

Can anyone help me resolve this issue?  Thank you.

How do I Add custom modules to Code Project AI? or any modules? I have the docker installed and running, and Blue Iris can communicate, but I'm missing the ALPR module and would like to add it in as well.

5 hours ago, crusemm said:

How do I Add custom modules to Code Project AI? or any modules? I have the docker installed and running, and Blue Iris can communicate, but I'm missing the ALPR module and would like to add it in as well.

Access your CP AI IP and Port

Then click the appropriate INSTALL for the module you want. Be patient as it takes time.

Edited October 5, 2025Oct 5 by kimocal