I sometimes got 403 error when i ask question from internet.

[zsw12abc]

My Ollama is hosted on my NAS via docker, (192.168.31.254)
and the Ollama port is same as the default one http://192.168.31.254:11434/

but still got 403 errors
the environment variable already added in the Ollama.

[n4ze3m]

Hey, sorry about the issue. Can you enable the custom origin header under Advanced Settings in Settings > Ollama Settings? This should hopefully resolve the issue, as Ollama seems to have a CORS issue.

[continue-revolution]

It seems that whenever we chat with deepseek for the second time in a chat history, we always get 403 no matter what, even if enabling custom origin header.

[n4ze3m]

The 403 issue in Ollama occurs due to a CORS problem, as described in this ollama/ollama#2335. Page Assist attempts to resolve it by rewriting headers, but unfortunately, this doesn't always work in certain browsers. #45 #308

Browser APIs often change with different releases, and since using web search is causing issues, I believe there is a bug in the origin header that is not being overridden. I will look into it.

The only current reliable solution is to add the following environment variable:

OLLAMA_ORIGINS=*

Steps to Add the Environment Variable

On Windows

1. Press Win + S and search for Environment Variables.
2. Open Edit the system environment variables.
3. In the System Properties window, click on Environment Variables.
4. Under System variables, click New.
   • Variable name: OLLAMA_ORIGINS
   • Variable value: *
5. Click OK to save the changes.
6. Restart Ollama.

On Mac

1. Open the Terminal.
2. Edit your shell configuration file:
   • For Bash: nano ~/.bash_profile
   • For Zsh (default on newer macOS): nano ~/.zshrc
3. Add the following line at the end of the file:

   export OLLAMA_ORIGINS=*
   

4. Save the file (Ctrl + O, then Enter, and Ctrl + X).
5. Run source ~/.bash_profile or source ~/.zshrc to apply the changes.
6. Restart Ollama.

On Linux

1. Open the Terminal.
2. Edit your shell configuration file:
   • For Bash: nano ~/.bashrc
   • For Zsh: nano ~/.zshrc
3. Add the following line at the end of the file:

   export OLLAMA_ORIGINS=*
   

4. Save the file (Ctrl + O, then Enter, and Ctrl + X).
5. Run source ~/.bashrc or source ~/.zshrc to apply the changes.
6. Restart Ollama.

This should resolve the issue. Hopefully, it works!

[continue-revolution]

I am on manjaro (zsh) with chrome. I did not permanently change zshrc. instead, I type export OLLAMA_ORIGINS=* then run ollama run deepseek-r1:32b , the 403 problem still exists. Then I tried running OLLAMA_ORIGINS=chrome-extension://* ollama run deepseek-r1:32b, the 403 problem still exists.

I guess it's not @n4ze3m 's bug. Instead, there could be some issue with my browser or ollama.

[n4ze3m]

Oh, that’s bad. Did you restart the Ollama?

try

OLLAMA_ORIGINS=* ollama serve

[continue-revolution]

It turned out that by running OLLAMA_ORIGINS=* ollama serve, it works properly. Thank you very much for your help

[odysseyalive]

There's another thing to remember, too. Google's search API is free to use. Half the time, Google doesn't return subsequent results--most likely due to resource monitoring per IP. So Page Assit might need to catch those failed API calls and retry.

[yCoffeeGames]

The keywords in Settings > RAG Settings > Prompt > Web does not fit the deepseek-r1 models well. After the first search on internet, large portion of the chat_history will be included in the search query sent to search engines. This causes the search engine to return errors (depending on search engine) and break the chat.

I tried many new keywords, but they are not working very well unless I delete the {chat_history}

[n4ze3m]

Yeah, the current follow-up query generation isn't really great using R1. I think I'll add an option to choose another model for generating follow-up questions or add a setting to turn off follow-up question generation.