Nginx 502 Bad Gateway with NPM Build LibreChat

[markcichy]

CONTEXT

LIBRECHAT SETUP
I ran with the instructions located under the Install LibreChat Locally Using npm section of the docs. No fancy changes here:

• Defined HOST in .env (used local IP with port 3080, but also attempted custom URL)
• Defined DOMAIN_CLIENT && DOMAIN_SERVER vars with my custom URL
• Confused as to how to interpret the TRUSTED_PROXY var; I have set to default 1 however not sure this is correct as both my Nginx server and LibreChat server VM's are on same network they have separate subnets - Nginx is on 10.0.1.0/24 && LiberChat is on 10.1.16.0/22 (both machines can ping with 0% oacket loss)
• UPDATE* Attempted added a second VNIC to LibreChat server and pointing to that IP 10.0.1.50 which is on the same 10.0.1.0/24 subnet as the Nginx host, no success
• I've modified/changed the .env file are refelct modified secrets as suggested in docs
• Lastly, I am using SendGrid for registrations and password resets
• Redis has not been configured
• No modifications have been made to the librechat.yaml conf

I have created a systemd service to autostart LibreChat on boot.

[Unit]
Description=LibreChat Service
After=network.target

[Service]
Type=simple
WorkingDirectory=/opt/librechat
ExecStart=/usr/bin/npm run backend
Restart=on-failure
User=<librechatusr>
Group=<librechatgrp>
Environment=PATH=/usr/bin:/usr/local/bin
Environment=NODE_ENV=production

[Install]
WantedBy=multi-user.target

The service is up and running without any issue.

● librechat.service - LibreChat Service
     Loaded: loaded (/etc/systemd/system/librechat.service; enabled; vendor preset: enabled)
     Active: active (running) since Tue 2025-05-06 15:12:41 UTC; 1s ago
   Main PID: 44201 (npm run backend)
      Tasks: 26 (limit: 19054)
     Memory: 143.0M
        CPU: 2.439s
     CGroup: /system.slice/librechat.service
             ├─44201 "npm run backend" "" "" "" "" "" "" "" "" "" "" "" "" "" ""
             ├─44212 sh -c "cross-env NODE_ENV=production node api/server/index.js"
             ├─44213 node /opt/librechat/node_modules/.bin/cross-env NODE_ENV=production node api/server/index.js
             └─44220 node api/server/index.js

May 06 15:12:41 tor-990librechat systemd[1]: Started LibreChat Service.
May 06 15:12:41 tor-990librechat npm[44201]: > [email protected] backend
May 06 15:12:41 tor-990librechat npm[44201]: > cross-env NODE_ENV=production node api/server/index.js

NGINX SETUP
I followed the Secure Deployment with Nginx docs to configure my server block (pulling the LetsEncrypt cert via certbot which I had running - I validated the cert also). Nothing "fancy" going on here other than I prefer to autoroute http requests to https via my own Nginx server blocks and so upon reading this I fear my config may be going against the grain?

Here is my server block:

server {
        # ssl
        listen  443 ssl  http2;
        listen [::]:443 ssl http2;
        # listen 80;

        server_name my.customdomain.com;
        
        location / {
                proxy_pass http://<local_server_ip>:3080;
                proxy_http_version 1.1;
                proxy_set_header Upgrade $http_upgrade;
                proxy_set_header Connection 'upgrade';
                proxy_set_header Host $host;
                ## ADD BELOW PER EXPRESS.JS TRUST PROXY DOCS
                ## https://expressjs.com/en/guide/behind-proxies.html
                proxy_set_header X-Forwarded-Host $host;
                proxy_set_header X-Forwarded-For $proxy_add_X_forwarded_for;
                proxy_set_header X-Forwarded-Proto $scheme;
                ## ADD ABOVE PER EXPRESS.JS TRUST PROXY DOCS
                proxy_cache_bypass $http_upgrade;
        }

        ssl_certificate /etc/letsencrypt/live/my.customdomain.com/fullchain.pem;
        ssl_certificate_key /etc/letsencrypt/live/my.customdomain.com/privkey.pem;

}

server {
    if ($host = my.customdomain.com) {
        return 301 https://$host$request_uri;
    }

    listen 80;
    listen [::]:80;

    server_name my.customdomain.com;
    return 404;

}

Nginx does not complain upon sudo nginx -t and subsequent checks/balances point to "all systems go." However, I am receiving the dreaded Bad Gateway error when attempting to hit my.customdomain.com.

nginx: the configuration file /etc/nginx/nginx.conf syntax is ok
nginx: configuration file /etc/nginx/nginx.conf test is successful

I've reviewed every article I could find here and none have helped, other than the link shared above which suggests that there is some form of internal routing happening with the configs. Tho, this suggests that client/nginx.conf only applies to Docker based installs?

Your help is much appreciated, please let me know if more context is required.

[markcichy]

SOLVED!!! By doing the following ..

Set .env files to point to the localhost and let Nginx handle EVERYTHING.

#==================================================#
#               Server Configuration               #
#==================================================#

HOST=10.0.1.50
PORT=3080

MONGO_URI=mongodb://127.0.0.1:27017/LibreChat

DOMAIN_CLIENT=http://localhost:3080
DOMAIN_SERVER=http://localhost:3080
# DOMAIN_CLIENT=https://my.customdomain.com # comment this out to force localhost
# DOMAIN_SERVER=https://my.customdomain.com # comment this out to force localhost

Doing this, while maintaing all of the above settings resolved the issue. I feel a bit silly for not attemping this to begin with as it is the most obvious solution. 🤦‍♂️

Sorry in advance if I've wasted anyone else's time other than my own.

[danny-avila]

Thanks for reporting the solution!

[markcichy]

No problem, @danny-avila ~ I will also add, that I finally managed to get redirects to work properly as well! Both local as well as third-party provider (tested on Google Workspace domains). Note that both DOMAIN_CLIENT && DOMAIN_SERVER must reflect the publicly accessible FQDN in order for redirect to work, duh. So, the above would be revised as ..

#==================================================#
#               Server Configuration               #
#==================================================#

HOST=<internal_ip_address>
PORT=3080

MONGO_URI=mongodb://127.0.0.1:27017/LibreChat

DOMAIN_CLIENT=https://my.customdomain.com
DOMAIN_SERVER=https://my.customdomain.com

If using a systemd service (as shown above) you'll need to reboot the service for changes to take effect. I've set this up on three production instances and can report that it is working 💯