brianpetro / obsidian-smart-connections

Chat with your notes & see links to related content with AI embeddings. Use local models or 100+ via APIs like Claude, Gemini, ChatGPT & Llama 3
https://smartconnections.app
GNU General Public License v3.0
2.5k stars 172 forks source link

After the upgrade, it has been stuck here, is there a problem? #497

Open liuling73 opened 6 months ago

liuling73 commented 6 months ago

image

brianpetro commented 6 months ago

Is your OpenAI API key working? You can click the save button in the settings to test it

liuling73 commented 6 months ago

Computer directly jammed, only forced to turn off Obsidian

brianpetro commented 6 months ago

Can you check the developer console for errors? https://github.com/brianpetro/obsidian-smart-connections/issues/488#issuecomment-1971189339

liuling73 commented 6 months ago
image

After clicking, the system is always in a dead state, should not consider that it can not link to GPT, but also can be performed quickly

liuling73 commented 6 months ago

image

brianpetro commented 6 months ago

@liuling73 those errors are OK to ignore. Are there any other errors? Are you using OpenAI or Local embeddings?

liuling73 commented 6 months ago

I use the api of OPENAI, and it seems that I can't ignore it now, so I can only turn off the plug-in temporarily, otherwise, Obsidian will always be in a state of crash, and I need you to upgrade to a lower level, thank you!

brianpetro commented 6 months ago

@liuling73 I'm not sure I understand what you're saying

marketfailuree commented 6 months ago

i have the same issue

JasonBates commented 6 months ago

same for me... 0 tokens per second.

Interestingly when saving the Open API key... a message pops up that the key is invalid. So I created a new API key at Open AI, carefully copied and pasted that into the Open API key box, hit save, and it is still shown as invalid.

These are the errors shown: image

Is there something about how the new app is working that is exceeding typical API limits? I'm on Tier 2 usage??

brianpetro commented 6 months ago

@JasonBates this is a common problem that people solve by adding a credit card to their OpenAI billing account. Sometimes prepaying for API credits is also necessary.

I can't comment on anything like OpenAI billing tiers. But I can say that the 429 error you're seeing is because OpenAI is restricting your account usage.

The local embeddings are also available to use and do not require an OpenAI API key.

🌴

adrianomarqui commented 5 months ago

After the upgrade, when I open obsidian, always start re-embedding my 1500+ notes... it is slow. I am putting my config here image

brianpetro commented 5 months ago

@adrianomarqui thanks for letting me know.

Jina-4k is being removed in the next version because issues, try using BGE-micro for both models and let me know if that changes anything.

🌴

adrianomarqui commented 5 months ago

@adrianomarqui thanks for letting me know.

Jina-4k is being removed in the next version because issues, try using BGE-micro for both models and let me know if that changes anything.

🌴

Now, with BGE, its working... thanks a lot brian...

LilSizzles commented 5 months ago

Hello @brianpetro,

thank you for you work, dedication and help in problem shooting.

I upgraded to the new version and now it won't generate embeddings for me. I have an OpenAI API account with a credit card and account credit.

It just hangs on 0/374 smart notes.

I may of made a rookie move. Any guidance would be great. I am running a MacBook Air M1, with Sanoma.

Thank you.

The process attempted thus far:

  1. Delete and re-install Smart Connections - 3 times.
  2. Refresh and paste in a new API code - 3 times. Each time the code is confirmed as working.

Please see screen shots below:

Screenshot 2024-03-26 at 16 45 39@2x

Screenshot 2024-03-26 at 16 46 10@2x

Screenshot 2024-03-26 at 16 47 37@2x

brianpetro commented 5 months ago

@LilSizzles, thanks for letting me know about this issue.

And thanks for providing the screenshots. It makes things a lot easier to debug.

Can you see if the BGE-micro local model works? If it does, this will rule out a lot of possible issues.

If it doesn't work, check to ensure the "Smart View" is opened in the sidebar. Currently, keeping the Smart View open is required for many of the features to work.

Lastly, if you could toggle on the "debug at startup time" option in the "community plugins" section, that would help improve the outputs in the developer console. This will help me be able to better figure out what is wrong based on the screenshots of the logs, like the ones you provided above.

Thanks for your help in patience while we get this figured out!

🌴

LilSizzles commented 5 months ago

Thank you for your response. Detailed, clear and generous. I followed your instructions and include some of my own thoughts. My messages are short because I have a damage arm and this is all dictation.

Can you see if the BGE-micro local model works? If it does, this will rule out a lot of possible issues. I am running it and it still doesn't work. Previously I tried this and the AI models and it would get to about 20 or 30 notes and then just sit and hang there. Screenshot 2024-03-27 at 11 31 47@2x

If it doesn't work, check to ensure the "Smart View" is opened in the sidebar. Currently, keeping the Smart View open is required for many of the features to work.

I've done as you suggested and you can see so by the screenshot above.

Lastly, if you could toggle on the "debug at startup time" Please find the screenshot for that just here, this is with the options as enabled in the original screenshot: Screenshot 2024-03-27 at 11 35 04@2x Screenshot 2024-03-27 at 11 35 12@2x Screenshot 2024-03-27 at 11 35 21@2x

If there's a more effective way to share the debugging to you beyond just screenshots, please let me know.

Thanks for your time.

Another note as I consider what else may be affecting this, I do run LuLu as a firewall and I allow all traffic from obsidian. Screenshot 2024-03-27 at 11 37 32@2x

With that said the original embedding in a earlier version, before I updated to 2.0, worked no problem and very swiftly.

liuling73 commented 5 months ago

This problem has always existed on my side, resulting in the plugin can not be used normally, and no effective solution has been found.

liuling73 commented 5 months ago

Except for an 8 in the title, there is basically no similarity in the content, as shown in the figure below:

image

The Settings are as shown below:

image

liuling73 commented 5 months ago

There's still a big gap here, and it hasn't been solved.

image

brianpetro commented 5 months ago

@liuling73 v2.1 has a multi-lingual BGE model. That might improve the Smart View relevance.

And I haven't been able to replicate the gap at the top of the Smart View. It might be worth toggling some of your other plugins to see which might be inserting the CSS that's impacting this. If you can narrow down the plugin then we can see what needs to be done to override it.

Thanks for following up 🌴

liuling73 commented 5 months ago

@brianpetro Thanks for your reply. Could you please tell me the operation steps of upgrading to V2.1

liuling73 commented 5 months ago

image

I returned the plug-in to V2.0.106 version, which can run normally, but when I upgraded to the latest version, the Settings as shown in the figure cannot work normally. I don't know where the problem is, please kindly take time to check the updated code of the plug-in, thank you for your hard work.

liuling73 commented 5 months ago

image

Set as shown in the picture, the quality of the associated card is also good, but need to run on the lower version of the plug-in, I am currently using VV2.0.106 version, it seems that the new version is more difficult to use the same, the white area on the right of the low version is no longer, but also hard to study the author again, thank you again!

liuling73 commented 5 months ago

The current configuration of the Notes Embedding Model is experiencing issues, potentially due to internal code problems. We need @brianpetro to review it. A big thank you to @brianpetro for their dedication in developing this useful tool.

liuling73 commented 5 months ago

v2.1

Is there an additional charge for v2.1?

brianpetro commented 5 months ago

@liuling73 v2.1 was just moved from supporter-only early release to publicly available general release today.

You should now be able to upgrade via the Obsidian community plugin index 🌴

liuling73 commented 5 months ago

Could you please tell me the specific upgrade steps? Thank you

christianngarcia commented 5 months ago

How to Update an Obsidian Plugin

Follow these detailed steps to ensure a smooth update process for your Obsidian plugin:

  1. Open Obsidian - Start by opening your Obsidian application.

  2. Access Settings - Click on the settings icon, typically located in the lower left corner of the app (the icon looks like a gear).

  3. Navigate to Community Plugins - In the settings menu, find the section labeled "Community Plugins." Click on it to open the community plugins area.

  4. Turn off Safe Mode - If Safe Mode is enabled, you'll need to disable it to manage your installed plugins. There is a toggle for Safe Mode at the top of the Community Plugins page—turn it off.

  5. Update Plugins - After disabling Safe Mode, you will see a list of your installed plugins. There should be an "Update" button next to each plugin that has an available update. Click this button for the plugin you wish to update. If you want to check for updates manually, there is usually a "Check for updates" button at the top of the plugin list.

  6. Review Changes - Some plugins might show release notes or a changelog during the update process. Review these notes to understand what changes or new features the update includes.

  7. Restart Obsidian - After updating the plugin(s), restart Obsidian to ensure that all changes take effect properly.

  8. Verify Plugin Functionality - Once Obsidian restarts, check the functionality of the updated plugin to make sure it works as expected. You might need to re-enable the plugin or adjust its settings depending on the update.

  9. Turn Safe Mode Back On (Optional) - For security purposes, you may choose to re-enable Safe Mode if you do not plan to install more plugins or updates immediately.

  10. Backup Your Notes (Recommended) - It's a good idea to periodically back up your Obsidian vault, especially after installing updates or new plugins, to protect your data.

liuling73 commented 5 months ago

I have updated, thank you, but using the same KEY, one display is valid, one display is not, somehow. As shown in the picture below:

image

liuling73 commented 5 months ago

If the following Settings are adopted, the different cards are also very similar, as shown in the figure below:

image

image
brianpetro commented 5 months ago

The issue with the "test" for the OpenAI API Key for embeddings has been fixed in v2.1.46. Thanks for bringing this to my attention.

And I'm unsure what you're showing me in the last two images.

🌴

liuling73 commented 5 months ago

Adopt a model『BGE-micro-v2 (Local, 512 tokens, 384 dim)』Associated result

liuling73 commented 5 months ago

@liuling73 v2.1 has a multi-lingual BGE model. That might improve the Smart View relevance. v2.1 具有多语言 BGE 模型。这可能会提高 Smart View 的相关性。

And I haven't been able to replicate the gap at the top of the Smart View. It might be worth toggling some of your other plugins to see which might be inserting the CSS that's impacting this. If you can narrow down the plugin then we can see what needs to be done to override it.而且我无法复制 Smart View 顶部的差距。可能值得切换您的其他一些插件,看看哪些插件可能插入了影响这一点的 CSS。如果您可以缩小插件的范围,那么我们可以看到需要做些什么来覆盖它。

Thanks for following up 感谢您的跟进 🌴

I have found the problem of the blank area in the upper right corner, I restored to version 2.0.112, I found that the title of the current card is on the top and bottom, as shown below, but the latest version of the title and the icon are displayed side by side, but when the width changes, the content of the title will be lengthened naturally, and the blank area in the upper right corner will appear.

image