Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Lookup Tool Stalls #917

Open
jaickler opened this issue Dec 7, 2024 · 8 comments
Open

Lookup Tool Stalls #917

jaickler opened this issue Dec 7, 2024 · 8 comments

Comments

@jaickler
Copy link

jaickler commented Dec 7, 2024

Bug

When using smart chat the lookup tool will show hypotheticals and then the loading wheel will display indefinitely.

Expected behavior

Recurve message from LLM after lookup finishes.

How to replicate

Use gpt-4o-mini with tools results being sent in user message or not.
Trigger the lookup tool.
Wait indefinitely for message to return.

Notes

I am using Linux on my laptop and do not have mobile enabled. Embedding have fully loaded already.

Repository owner deleted a comment Dec 7, 2024
@brianpetro
Copy link
Owner

Hi @jaickler

Are there any errors in the developer console logs when this happens?

Thanks for the report
🌴

@jaickler
Copy link
Author

jaickler commented Dec 8, 2024

Hello! Yes, I was able to check the developer logs and saw the vector sizes didn't match. I deleted the multi folder and then relaunched Obsidian and that fixed half the issue. The other half is that when using OpenAI's 4o-mini model I have to send the tool results in the user message or I also receive errors.
Sending the results in the user message does allow chat to work normally though.

I suppose the nature of the current bug is that, at least for 4o-mini the tool results must be sent in the user prompt.

@brianpetro
Copy link
Owner

Thanks for following up.

4o-mini seems to be working with/without sending tools response in the user message. Does toggling that setting on/off consistently produce the same success/fail results for you?

Please do let me know if you notice any other patterns or anything that might help narrow down what might be causing the issue 🌴

@Keffus
Copy link

Keffus commented Dec 8, 2024

I'm having the same problem with gemini

@brianpetro
Copy link
Owner

@jaickler @Keffus any screenshots that might help me better see what's going on would be helpful 🌴

@Keffus
Copy link

Keffus commented Dec 8, 2024

yeah sure, on the first screenshot i asked gemini if it can summarize the wiki description of cat and it did it but only in lookup, and it was loading the answer indefinitely. But when i asked the next question it responded normally(second screenshot). I think the lookup thing is related to asking about the vault because when i dont do it it answers normally. I have the review context option turned off.
Capture1
Capture2

@Keffus
Copy link

Keffus commented Dec 8, 2024

I check one more think i opened a new chat and just asked if it can tell me somethink(didn't ask to use vault) and it did answer in lookup only so I don't think it is realed to vault

@Keffus
Copy link

Keffus commented Dec 8, 2024

I don't know how much this will help but i have figure out that if gemini can't answer the question it will anser normally(without lookup
Capture1
Capture2
)

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

3 participants