Ollama taking so long to respond

matsyui · November 4, 2025, 8:18am

So i have tried the 2 method trying to get way to generate/chat with my local hosted ollama

via HTTPS: it keeps failing (timeout of 300000ms exceeded)
Ollama > Message A mode: I got it work but took 28mins, yes… “Minutes” to respond and after that re-testing it goes error similar to HTTPs

This only happens in n8n, upon testing Ollama API to other projects it works fine. (different prompt)

So to make sure or verify that… the problem is in n8n I created a test using thunderclient and Talend with the exact settings and exact prompt.

For thunder client it respond 12secs.

For Talend took 26secs.

So, for the flow of my actual testing I restarted everyting.. first tested in thunder client then talend next is n8n again both https and message a model (ollama). Issue still same….

The credential url of my ollama is same as the testing on thuderclient and talend, because localhost:11434 / ollama:11434 is not working. (also tried cloudflare tunnel for localhost:11434 it works but output still same for n8n working on the others but not in n8n similar to 192.168.1.26:11434)

Piotr_Sikora · November 4, 2025, 11:34am

Hey @matsyui

Please try ngrok for tunneling.
I’ve been using it for small AI Agents.

matsyui · November 4, 2025, 11:49am

(post deleted by author)

matsyui · November 4, 2025, 12:17pm

Don’t mind the tunnel because it is also slow using the local network ip.

also the tunnel works fine, no problem with it.

Parintele_Damaskin · November 4, 2025, 12:20pm

Curios… I use the HTTP node with http://localhost:11434/v1/completions

and it works OK.

No delay (and I use gemma3.1B as model).

matsyui · November 4, 2025, 12:24pm

This is the model I’ve tried:

The phi and gemma works great got response 10-30secs (to other projects)..

Only request from n8n is failing and i don’t understand why.

Parintele_Damaskin · November 4, 2025, 12:39pm

Pfff… you could set timeout, retry pooling to the HTTP node, but since the problem is why is taking too long only from n8n, sry can’t help further .

matsyui · November 4, 2025, 12:42pm

yeah timeout retry can be the solution but

28mins is the only request that worked and it only happen once hahaha…

compare it to my other projects that got seconds of response rate

other solution that i am thinking of is create another project and make api and another api will fetch it