n4ze3m / page-assist

Use your locally running AI models to assist you in your web browsing
https://chromewebstore.google.com/detail/page-assist-a-web-ui-for/jfgfiigpkhlkbnfnbobbkinehhfdhndo
MIT License
1.56k stars 148 forks source link

It can't find info in the Page Assist Github page #217

Open gerroon opened 1 month ago

gerroon commented 1 month ago

Hi

I just tested to see if it can find the default shortcut for Page Assist in the Github repo. It wasn't able to achieve the task in the web interaction mode. I tested it with Llama3.1 and Gemma latest using Ollama

open https://github.com/n4ze3m/page-assist enable "chat with current page"

what is the default keyboard shortcut

llama3-chatqa:latest
Sorry. I cannot find the answer based on the context you provided.
gerroon commented 1 month ago

Interesting enough, the Chrome addon finds the answer but the Firefox one can't. I use the default settings in both versions. Chrome

what is the default keyboard shortcut page assist

llama3.1:latest
According to the provided context, in Chrome:
Default Keyboard Shortcut: Ctrl+Shift+P
This opens the sidebar of Page Assist.

Firefox

what is the default keyboard shortcut page assist

llama3.1:latest
I don't know. The provided context doesn't mention a default keyboard shortcut for Page Assist. If you're looking for information on how to use or customize Page Assist, I can try to provide more general guidance based on the available documentation.
n4ze3m commented 1 month ago

Yes, I noticed the issue. It's because of semantic search, which may not provide an accurate answer, but you can turn it off.

https://github.com/user-attachments/assets/6868854e-8092-49cc-a7e3-600edc2c8328

gerroon commented 1 month ago

Thanks for the reply. What do we lose by turning it off? Isn't vector embeddings supposed to be a powerful tool?

n4ze3m commented 1 month ago

Yes, it works great most of the time. Page Assist is currently using simple semantic search, which I will improve in the upcoming version :)

zono50 commented 3 weeks ago

I can't get this to work with pdfs and when i search a simple term on the website, it can't find it image