AIPersonality: The provided path does not exist.

[jagbarcelo]

Expected Behavior

Executing webui.bat should download dependencies and end listening for commands on http://localhost:9600

Current Behavior

Executing webui.bat ends with the following error:

...
Checking models...
gpt4all-lora-quantized-ggml.bin
Model already installed
Virtual environment created and packages installed successfully.
Launching application...
No local configuration file found. Building from scratch
Traceback (most recent call last):
  File "D:\gpt4all-ui\GPT4All\app.py", line 736, in <module>
    personality = AIPersonality(f"personalities/{config['personality_language']}/{config['personality_category']}/{config['personality']}")
  File "D:\gpt4all-ui\GPT4All\env\lib\site-packages\pyaipersonality\__init__.py", line 82, in __init__
    raise ValueError("The provided path does not exist.")
ValueError: The provided path does not exist.

Steps to Reproduce

1. Download webui.bat and place it into a new empty directory, for instance D:\gpt4all-ui\
2. Execute webui.bat
3. Wait for the error to happen

Possible Solution

Everything was working just fine 2 hours ago. There must be something wrong with the some of the recent commits.

[ParisNeo]

Yes I lnow whjat's wrong

the default gpt4all category has changed from general to generic.

Just change that in your configs/config_local.yaml

[andzejsp]

removing config_local.yaml usually helps to reset and update default values.. it regenerates after launch

[jagbarcelo]

Both my /configs/default.yaml and /configs/local_default.yaml contain:

personality_language: english
personality_category: generic

The problem also occurs from a brand-new-empty reinstall through webui.bat

Also tried removing local_default.yaml (which is regenerated with the next run), but the problem persists.

[andzejsp]

but did the editing the value manually for the personality fixed the issue? the config part that gets generated needs some love

[jagbarcelo]

Editing local_default.yaml and setting personality_category: generic does not fix anything. It does not work setting personality_category: general either.

I have even removed the whole GPT4All directory and let the autoinstall download everything again. The error is still there.

[jagbarcelo]

I think I figured out the problem. It seems the webui.bat installer, as it is configured right now, does not install any personalities (the folder /personalities did not exist after running webui.bat for the first time).

I have manually run add_personality.bat selecting the options:

Enter the number of the desired language: 1 (English)
Enter the number of the desired category: 12 (generic)
Enter the number of the desired personality: 1 (gpt4all)

After that, webui.bat runs fine. Maybe you need to update it and check that at least the default personality exists.

[andzejsp]

"someone" made a big oopsy.. gonna check what i can do to fix this

[ParisNeo]

Problem fixed. Please just do a pull

[ParisNeo]

now there are no personalities by default The personalities folder is empty and by default a gpt4all hardcoded in the app is used When people mess up their personalities, this will garantee that the app will continue working. you can install new personalities using add_personality.bat file

[loup73]

The issue persists on Ubuntu even after running add_personality.sh. I pulled tand installed the repo today, run.sh returns the error. It looks for personalities/english/general/gpt4all and that path does not exist. Even if I create it in the personalities folder, it says it doesn't exist, I guess it expects an absolute path.

[ParisNeo]

Ok. Add_personality script installs the personality but doesn't set it to be used. You have to change the configs/config_local.yaml file to the right personality category and name

[loup73]

I try to run this on a properly protected public page that proxies to localhost:9600 of a local machine. Any hints on how to overcome random errors related to dialog loading, timeouts when returning etc? So far what I've done was adding upgrade-insecure-requests to the index.html page, to be able to route from https to http, which made the basic functionality of asking a question, available. Can it serve multiple parallel requests, as an example?

[andzejsp]

I try to run this on a properly protected public page that proxies to localhost:9600 of a local machine. Any hints on how to overcome random errors related to dialog loading, timeouts when returning etc? So far what I've done was adding upgrade-insecure-requests to the index.html page, to be able to route from https to http, which made the basic functionality of asking a question, available. Can it serve multiple parallel requests, as an example?

not sure what if its possible on old UI, but new UI works well behind a proxy, i run it daily to test the new UI. check out the local_defaults.yaml in the configs for ui flags.

As for concurrent replies, i dont think its implemented yet. It is planned, not by me but by main dev. Not sure when tho.

[andzejsp]

Is this issue resolved?

[Rejoov]

Im still having this issue and none of the steps above resolve it.