Smart Connect App (1.3.7) stuck in 'Making Smart Connections... Embedding progress: 210 / 16803 25571 tokens/sec using TaylorAl/bge-micro-v2'
Opened this issue · 8 comments
Obsidian and all plugins up-to-date
Smart Connect App cannot finish the embedding.
Reloading, re-installing Smart Connect app and Smart-Connections plugin -> no result
I have attached the errors from the console log of Smart Connect app.
Here are the settings:
{
"smart_blocks": {
"embed_model": {
"model_key": "TaylorAI/bge-micro-v2"
},
"embed_blocks": true
},
"smart_sources": {
"embed_model": {
"model_key": "TaylorAI/bge-micro-v2",
"legacy_transformers": true,
"TaylorAI/bge-micro-v2": {
"gpu_batch_size": 10,
"min_chars": 300
}
},
"pdf_adapter": {
"chat_model": {
"openai": {}
}
},
"image_adapter": {
"chat_model": {
"openai": {}
}
},
"smart_change": {
"active": true
}
},
"file_exclusions": "Untitled",
"folder_exclusions": "smart-chats, readwise, journals, assets, export, kindle-highlights",
"is_obsidian_vault": true,
"excluded_headings": ""
}
Thanks for the detailed bug report and screenshots 😊
I tried a few different ways to reproduce the issue, but I haven't had any luck.
Can you create a "test vault" that reliably reproduces the issue?
This way I can dig in to what might be happening 🌴
I may have found the problem. It seems that Smart-Connect is unable to process filenames containing the '#' sign.
All my files get embedded, except those meeting this criterion.
I have attached a one-file test-vault that reproduces the error reliably I think.
Oh yeah, that's got to be it. # in file names is currently not compatible.
It should be better handled, though. As in, providing a warning and not breaking things.
I'll have to see what I can do.
Thanks for the test vault
🌴
@sjhogerzeil the error should be prevented in the Obsidian Smart Connections plugin as of v2.2.85/2.3.26
The patch will also be included with the next Smart Connect app update (>=1.3.10) 🌴
@sjhogerzeil Hi , I'd like to know how to get the plugin's setting's file? I incorrectly change the AI model and now I met the same problem as you had met, moreover, my obsidian doesn't work when the Embedding process started and stuck at 10/1421 tokens, now I want to change my AI model to see if I can solve the problem but before the embedding process done, I can't change the AI model through the plugin's settings from obsidian
you see, I change the default AI model, and I tried to delete the plugin and reinstall it ,but the model is pointed to the one I changed before.... so I thought get the setting file might be a solution (cuz I can change the content of the file directly)
That helps! Thanks a lot 🥰 now the embedding process is ongoing although remains a long time xd