A full-stack chatbot project using the DeepSeek R1 model with Ollama, featuring:

• Angular frontend
• FastAPI backend
• Postgres database
• Nginx reverse proxy
• Ollama for model inference

nginx

• Listens on ports 80 & 443
• Loads SSL certs from /etc/letsencrypt on the host (read-only mount)
• Proxies requests to the Angular frontend, FastAPI backend, and Ollama

ollama

• Runs the deepseek-r1 model
• Exposes port 11435 for the backend to call

db (Postgres)

• Internal database store (not publicly exposed)

backend (FastAPI)

• Provides API routes for chatbot requests
• Calls Ollama for inference

frontend (Angular)

• Single-page app, served via Nginx

1. Nginx handles incoming requests on ports 80 (HTTP) or 443 (HTTPS).
2. The Angular frontend is at the root path.
3. The backend is proxied at /api/.
4. Ollama is proxied at /ollama/.
5. The backend calls Ollama to generate chat responses with the deepseek-r1:7b model.

# 1) Install Certbot (Snap-Based)
sh ./setup_certs.sh
# Installs snapd, removes old Certbot, and sets up Certbot from snap.

# 2) Stop Any Host-Level Nginx
sudo systemctl stop nginx
sudo systemctl disable nginx
# Ensures Docker can bind ports 80/443.

# 3) Obtain SSL Certificate
sudo certbot certonly --nginx -d yourdomain.com
# Or use --webroot if preferred, ensuring certs end up in /etc/letsencrypt/live/yourdomain.com

# 4) Deploy Docker Compose
sh ./deploy.sh
# Replaces "yourdomain.com" in default.conf.production, then starts all containers (nginx, backend, etc.)

# 5) Pull the DeepSeek Model
sh ./download.sh
# Enter the model name (e.g. deepseek-r1:7b). It downloads into the ollama container.

Have fun :-)