Unable to connect to Ollama

[rmd1433]

Hello!

I'm experiencing an issue connecting to the local Ollama server while using your page-assist extension. The application displays "Unable to connect to Ollama 🦙" message and a red icon.

I've tried various recommendations from the documentation and discussions, but the problem persists.

Could you provide instructions for running the local Ollama server or suggest other solutions to this issue?

Thank you for your assistance!

[n4ze3m]

Hey, sorry about that. Is your Ollama server running locally? If not, you need to run it. You can download Ollama from here: https://ollama.com/download

[oatmealm]

I'm jumping in to ask about using a remote installation of ollama. When trying something like http://10.0.0.10:11434 as the url_base it can't find the server. The server is accessible remotely otherwise from other client, like ios or cli.

[oatmealm]

Just to say that the project is really impressive (I've tried with a local server as well). thanks!

[n4ze3m]

I'm jumping in to ask about using a remote installation of ollama. When trying something like http://10.0.0.10:11434 as the url_base it can't find the server. The server is accessible remotely otherwise from other client, like ios or cli.

Usually, page assist works with the remote server. I will try to debug it. Maybe it's because of this issue: ollama/ollama#2335

Just to say that the project is really impressive (I've tried with a local server as well). thanks!

thanks :)

[oatmealm]

Great!
I've been looking around the console and I do see 403 forbidden... which is strange. OLLAMA_ORIGINS is set with a wildcard... I'm seeing same problem with Lumos

[oatmealm]

some more info... I'm running ollama manually on the server to trace:

OLLAMA_HOST=0.0.0.0 OLLAMA_ORIGINS=chrome-extension://* ollama serve

[n4ze3m]

🤔 can you try this command

OLLAMA_HOST=0.0.0.0 OLLAMA_ORIGINS=* ollama serve

[oatmealm]

It did.
Again, amazing work. Really enjoy using the app.
Would love to help in testing if you need help or feedback.

[n4ze3m]

Hey guys, a new fix has been released for this issue. All you need to do is go to Settings > Ollama setting and enable custom origin URL, which will resolve the issue.

for more information: https://github.com/n4ze3m/page-assist/blob/main/docs/connection-issue.md

[MicPec]

I am having trouble getting Ollama to work on my Vivaldi browser. Ollama is up and running, everything works fine on Firefox, but I can't seem to establish a connection on Vivaldi.

[n4ze3m]

Hey @MicPec

I just tested it on the Vivaldi browser (windows) , and it seems to be working fine. Can you try turning on this setting from the web UI settings > Ollama setting? It may resolve the issue.

for more details: https://www.youtube.com/watch?v=fydtRnxjfJU or https://github.com/n4ze3m/page-assist/blob/main/docs/connection-issue.md#solutions

[MicPec]

Thanks for quick response. I did try this before, no luck. BTW this is the Linux machine.
Open WebUI and Lobe Chat are working fine on Vivaldi so connection is blocked from extension. Maybe some vivaldi hidden settings, but I could't find.

[n4ze3m]

Hey @MicPec , I will test on my Linux machine with Vivaldi. For OpenWebUI and LobreChat, they have a server, but for PageAssist, everything happens in the browser, which is why this issue exists

Another fix is by changing Custom origin url in advance settings to chrome-extension://jfgfiigpkhlkbnfnbobbkinehhfdhndo, then you need to restart Ollama with the following environment variable, which may work:

OLLAMA_ORIGINS=chrome-extension://* ollama serve

source: ollama/ollama#2335 (comment)

[MicPec]

I've managed to run it, the problem was completely different. The connection was blocked by uBlock 🤦🏻‍♂️, so I had to add jfgfiigpkhlkbnfnbobbkinehhfdhndo to the exceptions lists. Now everything works fine.
Thanks for Your help anyway

[oatmealm]

Talking of which, and it's not specific to page-assist, I had to change OLLAMA_ORIGINS=chrome-extension://* to a promiscuous * , since nothing was able to connect any more remotely, or locally like from this extension (I'm using Firefox).

[n4ze3m]

Hey @oatmealm

In the latest version of Page Assist, there is an advanced URL configuration. All you need to do is enable Enable or Disable Custom Origin URL.

[oatmealm]

Is it needed when OLLAMA_ORIGINS is set correctly?

[n4ze3m]

Is it needed when OLLAMA_ORIGINS is set correctly?

Not needed :) An alternative method only

[ben-xD]

Talking of which, and it's not specific to page-assist, I had to change OLLAMA_ORIGINS=chrome-extension://* to a promiscuous * , since nothing was able to connect any more remotely, or locally like from this extension (I'm using Firefox).

Firefox uses moz-extension://, but that didn't work (I tried Environment="OLLAMA_ORIGINS=https://ollama-ui.github.io,app://obsidian.md*,chrome-extension://*,moz-extension://*")

My settings that work (the last lines is just so i use a HDD with lots of space for all the models):

[Service]
Environment="OLLAMA_HOST=0.0.0.0"
Environment="OLLAMA_ORIGINS=*"
Environment="OLLAMA_MODELS=/home/ben/ollama/models"

[uygur55]

Hi,
I have two internet browser, Edge and Chrome. I did put page-assist extension to Chrome.
page-assist can not connect to my local ollama running

.
i did try "Page Assist, there is an advanced URL configuration. All you need to do is enable Enable or Disable Custom Origin URL."
still not working
if you can help me
Thanks

[n4ze3m]

Hey, sorry about the issue, @uygur55 . Is Ollama and the extension you're trying to connect on the same computer, or are they on two different ones?

[uygur55]

on the same computer.
thanks

[n4ze3m]

Hey @uygur55, that's weird. It usually connects automatically. If possible, can you add this to your environment variable: OLLAMA_ORIGINS=chrome-extension://* [0] and restart Ollama? Usually, this resolves the issue. Please let me know if it works.

[0] ollama/ollama#2335 (comment)

[uygur55]

Hi,
i followed the way . still not working.

[n4ze3m]

Hey @uygur55, that’s unfortunate. Since you added the env, could you enable the advanced URL option? Please add chrome-extension://jfgfiigpkhlkbnfnbobbkinehhfdhndo as the URL. Make sure to save and refresh the page afterward in the tab.

Also, if you use uBlock, make sure to add the extension ID 'jfgfiigpkhlkbnfnbobbkinehhfdhndo,' as MicPec pointed out in the comment above

[uygur55]

still not working

[n4ze3m]

Hey @uygur55, You don't need to add ollama serve in the environment; only chrome-extension://* is needed. Make sure to restart Ollama after saving. :) I will investigate further and let you know if there is a fix.

[wuchewuche]

I currently have multiple computers, with two computers installed with Ollama. One of the computers has a model, and I want the computer without the model to be able to access the model on the computer that has it. My Ollama environment variable is also set to 0.0.0.0
But it still cannot connect properly

[n4ze3m]

Hi @wuchewuche, you don't need to use the Chrome extension URL as the origin URL. Instead, change it to http://127.0.0.1:11434, and it should work.

Alternatively, you can set OLLAMA_ORIGINS="*" as an environment variable on the computer running Ollama and then restart Ollama.

Hopefully, it works fine.

[wuchewuche]

hi @n4ze3m It is working, but it cannot find the model on the other computer that I want to access

[wuchewuche]

emmm @n4ze3m Other PC is http://192.168.1.83:11434 . what can i do?

[n4ze3m]

Hey @wuchewuche, can you check if this endpoint is working: http://192.168.1.83:11434/api/tags?

[wuchewuche]

hey @n4ze3m

What’s the next step?

[n4ze3m]

Hey @wuchewuche, that's why Page Assist can't access models from another computer.
Make sure you have added OLLAMA_HOST=0.0.0.0 as an environment variable on the computer hosting Ollama

[wuchewuche]

hey @n4ze3m
windows is this

The other computer is a Mac M2. How should I set the OLLAMA_HOST=0.0.0.0?

[n4ze3m]

Hey @wuchewuche , sorry for asking so many questions. Are you running Ollama on Windows and trying to access it from a Mac, or running it on a Mac and trying to access it from Windows?

[wuchewuche]

hey @n4ze3m I have a model on my Mac, and it has been successfully started locally. I plan to allow other Windows machines to directly use the model on my Mac without needing the model files locally.

[n4ze3m]

Hey @wuchewuche , You need to set OLLAMA_HOST and OLLAMA_ORIGIN on the Mac and make sure to restart the Ollama daemon. It should work.

[wuchewuche]

hey @n4ze3m mac is this

But Windows still cannot connect to the model on the Mac.

[n4ze3m]

Hey @wuchewuche, sorry about that :/ Can you try accessing http://192.168.1.83:11434/api/tags from your Mac to see if it works there?

Also, let's move this issue here: #355.

[wuchewuche]

hey @n4ze3m mac can not accessing http://192.168.1.83:11434/api/tags . I am sorry .what is the next step? I'm out of ideas.

I wrote it like this using Node js and vue, and after packaging, others on Windows can access the website normally, but I still really want the version you have; yours is better. Looking forward to your further updates.

[rick-github]

@wuchewuche How did you set OLLAMA_HOST? In your .zshrc or via launchctl?

[wuchewuche]

nano ~/.bash_profile
export OLLAMA_HOST="0.0.0.0"
export OLLAMA_ORIGIN="*"

That what i do.
@rick-github

[rick-github]

Needs to be set with launchctl fi you are not running it manually (ollama serve).

[wuchewuche]

how can i do? i do not use lauchctl before. @rick-github

[rick-github]

Click on the link I provided.

[wuchewuche]

launchctl setenv OLLAMA_HOST "0.0.0.0:11434" Do the same thing in origin . the http://192.168.1.83:11434/api/tags is show the {"models":[{"name":"nomic-embed-text:latest","model":"nomic-embed-text:latest","modified_at":"2025-02-08T17:31:56.437666829+08:00","size":274302450,"digest":"0a109f422b47e3a30ba2b10eca18548e944e8a23073ee3f3e947efcf3c45e59f","details":{"parent_model":"","format":"gguf","family":"nomic-bert","families":["nomic-bert"],"parameter_size":"137M","quantization_level":"F16"}},{"name":"deepseek-r1:8b","model":"deepseek-r1:8b","modified_at":"2025-02-06T11:37:47.650769748+08:00","size":4920738407,"digest":"28f8fd6cdc677661426adab9338ce3c013d7e69a5bea9e704b364171a5d61a10","details":{"parent_model":"","format":"gguf","family":"llama","families":["llama"],"parameter_size":"8.0B","quantization_level":"Q4_K_M"}}]}.
what can i do the next step? @rick-github

[rick-github]

What next step? Ollama is working. Connect your client.

[wuchewuche]

"I mean, I want to directly access the Ollama model I deployed on my Mac from my Windows host, and both are on the same network. I also want others, who are also on the same network, to be able to access it. My Mac would act like a broadcast. I’m not sure if I understand it correctly, so what’s the next step?" @rick-github

[rick-github]

Configure your client to connect to 192.168.1.83:11434.

[wuchewuche]

The Windows machine can connect to the Mac at 192.168.1.83:11434, but it only shows "Ollama is running." The "page Assist" is unable to retrieve the model, and the "Select a model" option in the plugin displays "No data." @rick-github

[rick-github]

Have you configured "Page Assist" to connect to 192.168.1.83:11434?

What happens if you run the following on the Windows machine:

curl http://192.168.1.83:11434/api/tags

[wuchewuche]

the http://192.168.1.83:11434/api/tags is show the {"models":[{"name":"nomic-embed-text:latest","model":"nomic-embed-text:latest","modified_at":"2025-02-08T17:31:56.437666829+08:00","size":274302450,"digest":"0a109f422b47e3a30ba2b10eca18548e944e8a23073ee3f3e947efcf3c45e59f","details":{"parent_model":"","format":"gguf","family":"nomic-bert","families":["nomic-bert"],"parameter_size":"137M","quantization_level":"F16"}},{"name":"deepseek-r1:8b","model":"deepseek-r1:8b","modified_at":"2025-02-06T11:37:47.650769748+08:00","size":4920738407,"digest":"28f8fd6cdc677661426adab9338ce3c013d7e69a5bea9e704b364171a5d61a10","details":{"parent_model":"","format":"gguf","family":"llama","families":["llama"],"parameter_size":"8.0B","quantization_level":"Q4_K_M"}}]}. What other Page Assist settings need to be configured to connect to 192.168.1.83:11434? @rick-github

[rick-github]

the http://192.168.1.83:11434/api/tags is show the {"models":

So ollama is working.

What other Page Assist settings need to be configured to connect to 192.168.1.83:11434?

I don't know. Read the documentation for Page Assist.