How to run AI models in LAMP App

[robi]

Most of the AI projects are based on python and are CLI based until someone puts a node UI on top.

If you plan to work on getting things like h2ogpt or chatbot-ui to work with gpt-llama.cpp or lit-parrot on Cloudron, you can start with the LAMP app, provided you specify a few environment variables.

1. python & npm will need the HOME env variable set, I used:
   export HOME=/app/data
2. pip will need the installed bits to be in the PATH, I used:
   export PATH=$PATH:/app/data/.local/bin

You can put both of these in the /app/data/run.sh file so they're set every time you open the terminal.

export HOME=/app/data
export PATH=$PATH:/app/data/.local/bin

Then simply git clone the repos, install deps, adjust ports & env vars, set the proxy (see below), grab a model, run the backend app, run the chatbot-ui (setup guide) and see the model spring to life.

Have fun!

[robi]

@staff does the .htaccess proxy trick not work anymore?:

RewriteEngine on
RewriteRule ^/(.*)?$ http://localhost:8000/$1 [P,L]

The node app is running on port 8000 and responds to curl, but the rule doesn't seem to pull it up from a browser, only the default lamp index or a 404 if I drop the '^/'.

Anyone else have this as a working example?

See lamp.demo.cloudron.io

[girish]

@robi Apparently, P is an alternate way of ProxyPass . I think the lamp app does not proxy module enabled.

[girish]

Works now after enabling the module.

[girish]

I pushed a new package that enables the proxy module.

[robi]

@girish Thank you!

[robi]

Hmm, @girish did you push a custom package?

After updating the LAMP app on my server, it still doesn't proxy properly via .htaccess file.

My version:

[robi]

@girish the new lamp app does have the modules enabled, but still no go via .htaccess file.

root@080866d1-7701-4faf-bc28-d6d63dbf252b:/app/data/apache# apache2ctl -M | grep proxy
 proxy_module (shared)
 proxy_http_module (shared)

Oh I see; you added the proxypass block in the apache/app.conf instead.

    <Location "/">
    ProxyPass "http://localhost:8000"
    </Location>

Not sure why .htaccess doesn't work as the rewrite_module is also loaded.

[girish]

@robi said in How to run AI models in LAMP App:

Hmm, @girish did you push a custom package?

yes, I updated the demo app instance for testing before I published the package. you can just reinstall I guess.

[girish]

@robi said in How to run AI models in LAMP App:

Not sure why .htaccess doesn't work as the rewrite_module is also loaded.

There is something wrong with that rule but I don't know htaccess configuration well enough. In fact, I wasn't aware of P before this. If you put RewriteRule "(.*)$" "http://localhost:8000/$1" [P,L] it does something (i.e a not found page from the node app atleast).

@robi said in How to run AI models in LAMP App:

RewriteRule ^/(.*)?$ http:/localhost:8000/$1 [P,L]

Also, there is a missing slash after http.

[robi]

Good catch, however I don't have the same typo in my file.

I'll try a few more things and report back.

[robi]

@girish not sure why, but what did work for the bare domain was (.*)^

RewriteRule (.*)^ http://localhost:8000/$1 [P,L]

There is also a /docs URL which then doesn't work, unless the ^ is missing.

RewriteRule (.*) http://localhost:8000/$1 [P,L]

[robi]

If you don't care to use the proxy features in LAMP via Apache2, you can launch the App-Proxy app from the App Store and configure it on a new subdomain to proxy to any of the LAMP installed apps on local ports via the internal Cloudron IP, ie. http://172.18.xx.xxx:port (make sure to use HTTP and not HTTPS).

You can get the internal IP from the web terminal by typing hostname -i