A complete, production-ready Docker Compose setup for deploying your own private LLM infrastructure. Get a ChatGPT-like interface that works with any LLM provider - OpenAI, Anthropic, Google, local models, and more. Pay per token across multiple providers while managing all your API keys and usage from one centralized dashboard.
What you get:
- ChatGPT-style web interface for any LLM provider
- Unified API endpoint for all your language models
- Centralized token usage tracking and billing management
- Self-hosted, private alternative to ChatGPT Plus
- Open WebUI - Feature-rich web interface for interacting with language models
- LiteLLM - Unified API gateway for 100+ LLM providers (OpenAI, Anthropic, local models)
- PostgreSQL - Shared database for persistence
- Traefik - Reverse proxy with automatic HTTPS via Let's Encrypt
- Optional S3 Backup - Automated database backups
- Memory: Minimum 1GB RAM
- Storage: 50GB disk space (for Docker images and data)
- OS: Linux (Debian/Ubuntu recommended)
- Domain: DNS records pointing to your server
You can deploy HomeLLM in two ways:
Fork this repository and deploy automatically on every push to main.
-
Fork this repository to your GitHub account
-
Configure Repository Secrets in your forked repo: Go to Settings → Secrets and Variables → Actions, then add all the secrets from the table below.
-
Configure Repository Variables:
- Set
COMPOSE_PROFILEStobackupif you want automated backups (optional)
- Set
-
Push to main branch - deployment happens automatically!
The deployment uses DCD (Deploy Docker Compose) tool for fast, reliable deployments to your server.
Deploy directly on your server using Docker Compose.
-
Clone the repository
git clone https://github.com/g1ibby/homellm.git cd homellm -
Configure environment
cp .env.dist .env # Edit .env with your values (see table below) -
Deploy
docker-compose up -d
Both deployment methods require the same environment variables. For GitHub Actions, set these as Repository Secrets. For manual deployment, configure them in your .env file.
| Variable | Description | Example |
|---|---|---|
POSTGRES_USER |
PostgreSQL username | postgres |
POSTGRES_PASSWORD |
PostgreSQL password | your_strong_password |
LITELLM_HOSTNAME |
LiteLLM domain | litellm.yourdomain.com |
LITELLM_MASTER_KEY |
LiteLLM API key (must start with sk-) |
sk-your_master_key |
LITELLM_SALT_KEY |
LiteLLM encryption key (must start with sk-) |
sk-your_salt_key |
LITELLM_UI_USERNAME |
LiteLLM admin username | admin |
LITELLM_UI_PASSWORD |
LiteLLM admin password | your_admin_password |
OPENWEBUI_HOSTNAME |
Open WebUI domain | chat.yourdomain.com |
WEBUI_SECRET_KEY |
Open WebUI secret key | your_secret_key |
| Variable | Description | Example |
|---|---|---|
DEPLOY_TARGET |
SSH connection string | user@your-server.com:22 |
SSH_PRIVATE_KEY |
SSH private key content | -----BEGIN OPENSSH PRIVATE KEY-----... |
| Variable | Description | Example |
|---|---|---|
COMPOSE_PROFILES |
Enable backup service | backup |
S3_BUCKET |
S3 bucket name | your-backup-bucket |
S3_ID |
S3 access key ID | AKIAIOSFODNN7EXAMPLE |
S3_SECRET |
S3 secret access key | wJalrXUtnFEMI/K7MDENG/bPxRfiCYEXAMPLEKEY |
S3_REGION |
S3 region | us-east-1 |
S3_ENDPOINT |
S3 endpoint (for non-AWS) | https://s3.amazonaws.com |
Point these A records to your server's IP address:
chat.yourdomain.com(or your chosen OpenWebUI hostname)litellm.yourdomain.com(or your chosen LiteLLM hostname)
Cloudflare users: Set SSL mode to "Full" or "Full (Strict)" for optimal compatibility.
After deployment (allow 1-2 minutes for initialization):
- Chat Interface:
https://chat.yourdomain.com - LiteLLM API:
https://litellm.yourdomain.com
When COMPOSE_PROFILES=backup is set, the system automatically:
- Backs up both databases daily at midnight
- Retains backups for 7 days
- Stores backups in your configured S3 bucket
View logs for troubleshooting:
# All services
docker-compose logs
# Specific service
docker-compose logs traefik
docker-compose logs litellm
docker-compose logs open-webui- All secrets should be randomly generated
- LiteLLM keys must start with
sk-prefix - Use strong passwords for all accounts
- Keep your server and Docker images updated
- Consider setting up fail2ban for SSH protection
Note: This setup is designed for personal/small team use. For production environments with high traffic, consider additional security hardening and scaling strategies.