Deploy RAGFlow on a VPS

RAGFlow is a RAG (retrieval-augmented generation) engine with document parsing, chunking, and a web UI. It ships as a multi-container Docker stack (server, MySQL/Postgres, Elasticsearch or Infinity, Redis, MinIO), so it's memory-hungry — plan for a real VPS, not the smallest tier.

1. Prerequisites

• A RamNode VPS running Ubuntu 24.04 LTS (KVM plan recommended)
• Minimum 16 GB RAM, 4 vCPUs, 50 GB disk — RAGFlow's default stack (Elasticsearch + MySQL + MinIO + Redis + the app server) is heavy; 8 GB will swap constantly under real document loads
• A domain or subdomain pointed at the VPS's public IP (A record) — e.g. ragflow.yourdomain.com
• Root or sudo SSH access

2. Initial server setup

apt update && apt -y upgrade
apt -y install curl git ufw

# Basic firewall
ufw allow OpenSSH
ufw allow 80/tcp
ufw allow 443/tcp
ufw enable

Increase vm.max_map_count, which Elasticsearch requires:

echo "vm.max_map_count=262144" >> /etc/sysctl.conf
sysctl -p

3. Install Docker Engine + Compose plugin

curl -fsSL https://get.docker.com | sh
apt -y install docker-compose-plugin
systemctl enable --now docker

4. Pull RAGFlow and configure

git clone https://github.com/infiniflow/ragflow.git /opt/ragflow
cd /opt/ragflow/docker

Edit .env in this directory — the key settings to review:

• RAGFLOW_IMAGE — use the :v0.x.x-slim tag unless you need embedded models baked into the image (the slim image is far smaller and faster to pull)
• MYSQL_PASSWORD, MINIO_PASSWORD, ELASTIC_PASSWORD — set strong unique passwords, not the defaults
• TIMEZONE — set to your local zone
• DOC_ENGINE — elasticsearch (default) or infinity (lighter weight, fewer moving parts if RAM is tight)

If you're memory-constrained, switch DOC_ENGINE=infinity and use the corresponding docker-compose.yml variant included in the repo (docker-compose-base.yml swaps the engine).

5. Bring up the stack

docker compose -f docker-compose.yml up -d

First boot pulls several large images and can take a few minutes. Watch logs until the server reports it's listening:

docker logs -f ragflow-server

By default the web UI listens on port 80 inside the compose network, mapped to the host. Confirm with:

docker compose ps

6. Put nginx in front with TLS

Don't expose RAGFlow's container port directly to the internet long-term — front it with nginx and Let's Encrypt so you get HTTPS and can layer on rate limiting later.

apt -y install nginx certbot python3-certbot-nginx

Create /etc/nginx/sites-available/ragflow:

server {
    listen 80;
    server_name ragflow.yourdomain.com;

    client_max_body_size 100M;   # document uploads can be large

    location / {
        proxy_pass http://127.0.0.1:9380;   # RAGFlow's host-mapped port; check docker compose ps
        proxy_set_header Host $host;
        proxy_set_header X-Real-IP $remote_addr;
        proxy_set_header X-Forwarded-For $proxy_add_x_forwarded_for;
        proxy_set_header X-Forwarded-Proto $scheme;
        proxy_read_timeout 600s;   # large document parsing can be slow
    }
}

ln -s /etc/nginx/sites-available/ragflow /etc/nginx/sites-enabled/
nginx -t && systemctl reload nginx
certbot --nginx -d ragflow.yourdomain.com

Since nginx now owns 80/443, you can remove RAGFlow's own port 80 mapping in docker-compose.yml if it conflicts, or just bind it to 127.0.0.1 only.

7. First login and model setup

Visit https://ragflow.yourdomain.com, register the first admin account, then under Settings → Model Providers add your LLM/embedding provider (OpenAI-compatible endpoint, Ollama, etc.). RAGFlow doesn't ship a model itself — it orchestrates calls to whatever provider you configure.

8. Backups

The stateful pieces live in Docker volumes: MySQL/Postgres data, MinIO (uploaded documents), and the Elasticsearch/Infinity index. At minimum, cron a nightly dump:

docker exec ragflow-mysql mysqldump -u root -p"$MYSQL_PASSWORD" --all-databases > /root/backups/ragflow-mysql-$(date +%F).sql
docker run --rm -v ragflow_minio_data:/data -v /root/backups:/backup alpine \
  tar czf /backup/ragflow-minio-$(date +%F).tar.gz -C /data .

Rotate old backups off-box (RamNode object storage or another VPS) rather than keeping them on the same disk.

9. Updating

cd /opt/ragflow/docker
docker compose down
git pull
docker compose pull
docker compose up -d

Check the release notes first — RAGFlow has had breaking config/schema changes between minor versions.

Troubleshooting notes

• Elasticsearch container exits immediately → almost always the vm.max_map_count setting from step 2 wasn't applied, or RAM is too low for ES's default heap.
• Uploads fail past a certain size → check both nginx's client_max_body_size and RAGFlow's own upload limit env var.
• Slow document parsing → parsing is CPU-bound; on a 4 vCPU box, expect large PDFs to take real time. Scale vCPUs before assuming something's broken.