Docker likely not running API request failed: Internal Server Error

[qapaj]

[ntegrals]

@qapaj You have to download Docker engine and start it before running the application or click retry.

[Myrinia]

Running the complete Stack on Windows, i can confirm, that the Communication with Docker API is not Working.

Running Docker Desktop, Service is running, and in terminal: docker -v outputs

docker -v
Docker version 28.3.0, build 38b7060

// Working Solution is

• backend/src/services/docker.ts

line 8 to:

const docker = new Docker({
  host: '127.0.0.1',
  port: 2375
});

[MOHAMAD-ZUBI]

+1

this didn't solve the issue as well const docker = new Docker({
host: '127.0.0.1',
port: 2375
});

[qapaj]

I did everything but the same error. Do you have Discord to see for yourself? @ntegrals

[Myrinia]

In Docker, you have to expose the Daemon.

[qapaj]

@Myrinia Thank you brother, it worked, but a new problem appeared.

[Myrinia]

Hi,

Well, seems like the docker Container can't connect to your LLM.

So, in case u run ollama locally, your config should look like this:

// Make sure to replace the values with your actual API key and model

// USING ANTHROPIC CLAUDE SONNET 4 is strongly recommended for best results

export const config = {
  aiSdk: {
    // The base URL for the AI SDK, leave blank for e.g. openai
    baseUrl: "http://host.docker.internal:11434/v1",

    // Your API key for provider, if using Ollama enter "ollama" here
    apiKey: "ollama",

    // The model to use, e.g., "gpt-4", "gpt-3.5-turbo", or "ollama/llama2"
    model: "codellama",
  },
} as const;

i think lots of people, just get the base Url wrong. If you think about it, the docker container, tries to access localhost, which has no service running in its virtual environment, so you need to take the hosts network, so ensure that the baseurl is NOT looking like this

baseUrl: "http://localhost:11434/v1",

it should be like that, for ollama running on the host mashine

 baseUrl: "http://host.docker.internal:11434/v1",

[qapaj]

@Myrinia
Thanks again for your help.

Unfortunately, the issue is still not resolved — I’ve updated the baseUrl to http://host.docker.internal:11434/v1 exactly as suggested, but the Docker container still can’t connect to Ollama running on my host machine.

Would be great if we can troubleshoot this more directly.
Here’s my Discord: _alitix — feel free to add me.

Appreciate your time!

[Myrinia]

Hi,

I think the Issue may help others aswell, so we should keep this conversation here.

Ensure, that your Ollama is reachable, so rightclick it => Settings, and ensure that it is accessable using the network.

Then u can just test this yourself using curl or your browser:

// Please use the correct IP here

curl http://192.168.1.100:11434/api/tags  

If it still does not work, stop the ollama service, and run this script to ensure, that ollama allows connections on all interfaces

Yet again: ensure to modify this script, to your correct ollama.exe path.

@echo off
set OLLAMA_HOST=0.0.0.0
start "" "C:\Program Files\Ollama\ollama.exe" serve

[qapaj]

Hi,

Thanks for the help.

I updated the script and made sure ollama.exe is running using this command:

bat
Copy
Edit

@echo off
set OLLAMA_HOST=0.0.0.0
start "" "C:\Users\alidi\AppData\Local\Programs\Ollama\ollama.exe" serve

Ollama runs without errors.

However, the issue still exists — there’s no response from the AI in the website chat. It just hangs and doesn’t reply to any prompt.

I also tried accessing http://localhost:11434/api/tags in the browser and with curl, and it works. So the Ollama API is reachable, but the frontend AI chat is not responding.

Any ideas?

Thanks again. @Myrinia

[Myrinia]

Hi,

Seems like your Docker is not able to communicate with ollama.

You can test two things:

1.) Connect into your Docker and try the curl command

Open a cmd and enter

docker exec -it <mycontainer> sh

so in your example this would be:

docker exec -it 9d9fc850 sh

In this Shell, you can try to get the path to the API right, so you can try:

curl http://host.docker.internal:11434/api/tags  

curl http://localhost:11434/api/tags  

curl http://172.17.0.1:11434/api/tags  

2.) 172.17.0.1
Sometimes there is a DNS issue within Docker, so please Try to change the baseUrl to 172.17.0.1 instead of the host.docker

// Make sure to replace the values with your actual API key and model

// USING ANTHROPIC CLAUDE SONNET 4 is strongly recommended for best results

export const config = {
  aiSdk: {
    // The base URL for the AI SDK, leave blank for e.g. openai
    baseUrl: "http://172.17.0.1:11434/v1",

    // Your API key for provider, if using Ollama enter "ollama" here
    apiKey: "ollama",

    // The model to use, e.g., "gpt-4", "gpt-3.5-turbo", or "ollama/llama2"
    model: "codellama",
  },
} as const;

3.) Windows Firewall:

Disable your windows Firewall for this test. Windows Firewall / Defender sometimes block such requests.
Also Check if the ollama.exe is not blocked by the firewall

[qapaj]

Hi,

Thanks again for the suggestions. I tried all of them, but unfortunately none of them worked. Here's what I did and what happened:

1. Docker exec + curl attempts
   I ran:
   docker exec -it 9d9fc850 sh
   Then tried:

   • curl http://host.docker.internal:11434/api/tags → No response, it hangs indefinitely.
   • curl http://localhost:11434/api/tags → Same issue, no response.
   • curl http://172.17.0.1:11434/api/tags → Also hangs, no output.

2. Windows Firewall
   I disabled the Windows Firewall entirely and confirmed that ollama.exe is allowed, but still the same issue — no response from the curl command inside Docker.

3. Port Listening
   I confirmed that Ollama is listening on port 11434:

   netstat -ano | findstr :11434

   Shows:

   TCP    0.0.0.0:11434          0.0.0.0:0              LISTENING       32132
   TCP    [::]:11434             [::]:0                 LISTENING       32132
   

   So the port is open and Ollama is running.

Despite all that, Docker still can't communicate with Ollama through any of the tested addresses. Do you have any other ideas? Could this be a network bridge or Docker Desktop config issue?

Thanks again for your help.
@Myrinia

[bsplu]

I was running december on my development machine and accessing it remotely from my laptop, and I encountered the same error. I am confident that Docker is correctly configured on the development machine. After checking the browser console, I saw the following error:

I traced the issue to the file:
december/frontend/src/lib/backend/api.ts
where the backend URL is hard-coded as:

const API_BASE_URL = "http://localhost:4000";

Here, localhost refers to my laptop, not the development machine where the Docker containers are actually running. After replacing localhost with the IP address of my development machine, the problem was resolved and the frontend was able to connect to the Docker backend correctly.

I also found that localhost appears in the following files. These hard-coded addresses prevent the project page from connecting to Docker to render the project after it is created. Replacing them with the actual IP address of my development machine fixes the issue.

backend/src/routes/containers.ts
backend/src/services/docker.ts
frontend/src/app/create/components/ChatMessage.tsx
frontend/src/app/editor/CodeEditor.tsx
frontend/src/app/projects/components/TopNavigation.tsx
frontend/src/app/projects/components/WorkspaceDashboard.tsx
frontend/src/lib/backend/api.ts

Hard-coding the IP address is obviously not an ideal solution, but it does resolve the problem for now. It would be much better if API_BASE_URL could be made configurable via a proper config mechanism instead of being scattered as hard-coded values throughout the codebase.