You can retry with another model. However, in my test case, I used OpenChat with a larger file (2.43MB) and it still works.

OpenChat on Tinyllama paper:

Openhermes on Tinyllama paper:

from jan.

hi @chenshaoju, can you give us the settings.json in jan\settings? thank you

from jan.

@chenshaoju thank you, can you give us the settings in the Tools section of your thread?

from jan.

@chenshaoju, nothing seems wrong so far, can you give us the entire conversation? jan\threads\
Thank you for your time 🙏

from jan.

@Van-QA I'm very happy to help solve this issue.🤗

here is the conversation:
jan_1709104282.zip

from jan.

@chenshaoju, can you also try our latest nightly build? and see if the issue resolved?
https://github.com/janhq/jan?tab=readme-ov-file#download

from jan.

@louis-jan Sure thing, here is the thread with the issue, with 0.4.7-304 version of Jan: jan_1709880132.zip

Here are the logs: app.zip

The PDF files in the thread are from: https://www.linkedin.com/posts/activity-7167774533339299841-M5xc , it's a public paper, with no sensitive information. 👏

from jan.

@chenshaoju I found the root cause of the issue: you deleted a model marked as default from the previous thread. As a result, new threads have a model of "*" in threads.json, thus, no files were ingested afterward. In the meantime, you can work around this by deleting older threads. Create a new thread and select a model from the drop-down list, then upload files. We will fix this issue as soon as possible in upcoming releases. cc @Van-QA

from jan.

@Van-QA Thanks for the reply, version 0.4.9-343 doesn't seem to work properly, It's stuck on "Generation Response..." and has no GPU usage.

here are the logs folder and threads folder:
logs_threads.zip

hi @chenshaoju, another issue came up, I will notify you again once it's resolved, thank you for your time 🙏

from jan.

hi @chenshaoju, the build Jan v0.4.9-345 was able to function properly with document analysis, can you try?

from jan.

Thanks, The version is 0.4.9-348 now, but it's working perfectly.

Thanks you!
🎉

from jan.

Thanks for the reply.

I try it again:

I also changed the language of the question (in simplified Chinese) and the model:

After rollback to 0.4.6, everything looks fine now:

Simplified Chinese with the same model before:

If you need any more information, please feel free to ask.👋

from jan.

@Van-QA Here you are, thanks for the reply.

Version 0.4.6:
settings_0.4.6.json

Version 0.4.7:
settings_0.4.7.json

Here is screenshot of all my settings in UI:
| | |
| | |
| |

from jan.

@Van-QA here you are:
| | |
| | |

from jan.

I had a similar problem, I wanted to chat with books, I sent a 13MB Chinese book, my graphics card is an RTX3060 12G The graphics card is highly utilized, but it keeps getting stuck in the progress bar living without an answer, and at this point clicking on stop the program won't stop it. I am using full default settings Openchat-3.5 7B Q4.
Jan v0.4.7 win11

from jan.

@Van-QA Thanks for the reply, But unfortunately, it's still not working.

Here are screenshots, threads, and logs:

jan_1709185945.zip

app.zip

from jan.

The refresh button may lead to the same issue.

AI can recognize documents by:

- Sending the document along with the question directly.
or
- Clicking the edit button, and then clicking submit.

Image:

AI cannot recognize documents by:

- Clicking the refresh button to make the AI regenerate.

Image:

from jan.

@chenshaoju Could you please share the compressed version of the broken thread so we can reproduce, ensuring there is no sensitive personal information?

from jan.

@chenshaoju we resolved the issue as of Jan v0.4.9-343, would you be able to retry and see if the issue persist? 🙏

from jan.

@Van-QA Thanks for the reply, version 0.4.9-343 doesn't seem to work properly, It's stuck on "Generation Response..." and has no GPU usage.

here are the logs folder and threads folder:
logs_threads.zip

from jan.