Self-Host Perplexica Private AI Search Engine on a VPS (2026): SearXNG, Docker, Ollama/OpenAI and HTTPS Deployment Guide

Perplexica is not an ordinary search engine, nor is it a pure chatbot.

It fits scenarios like these:

I understand it as: SearXNG is responsible for finding material, and the large model is responsible for reading it.

If you just want to search web pages, SearXNG is enough. If you want to get answers with sources directly after searching, then Perplexica makes sense.

Here too, there are two cases.

The officially recommended simple way is Docker. Perplexica's full image packs SearXNG inside, so you don't need to deploy the search backend separately.

First log in to the VPS:

ssh root@your_server_ip

Update the system:

apt update && apt upgrade -y

Install Docker:

apt install -y ca-certificates curl gnupg ufw
install -m 0755 -d /etc/apt/keyrings
curl -fsSL https://download.docker.com/linux/ubuntu/gpg | gpg --dearmor -o /etc/apt/keyrings/docker.gpg
chmod a+r /etc/apt/keyrings/docker.gpg
echo "deb [arch=$(dpkg --print-architecture) signed-by=/etc/apt/keyrings/docker.gpg] https://download.docker.com/linux/ubuntu $(. /etc/os-release && echo $VERSION_CODENAME) stable" > /etc/apt/sources.list.d/docker.list
apt update
apt install -y docker-ce docker-ce-cli containerd.io docker-buildx-plugin docker-compose-plugin

Confirm Docker:

docker --version
docker compose version

Run Perplexica:

docker run -d \
  --name perplexica \
  --restart unless-stopped \
  -p 127.0.0.1:3000:3000 \
  -v perplexica-data:/home/perplexica/data \
  itzcrazykns1337/perplexica:latest

Here I deliberately used:

127.0.0.1:3000:3000

This means only local access is allowed, and it is not directly exposed to the public internet. Later, use Caddy to reverse proxy to an HTTPS domain.

Check container status:

docker ps --filter "name=perplexica"

Check logs:

docker logs perplexica --tail=100

Test locally:

curl -I http://127.0.0.1:3000

If it returns 200, 302, or a similar response, the service is basically up.

If you prefer a more maintainable approach, use Compose.

Create the directory:

mkdir -p /opt/perplexica
cd /opt/perplexica

Write docker-compose.yml:

services:
  perplexica:
    image: itzcrazykns1337/perplexica:latest
    container_name: perplexica
    restart: unless-stopped
    ports:
      - "127.0.0.1:3000:3000"
    volumes:
      - perplexica-data:/home/perplexica/data

volumes:
  perplexica-data:

Start:

docker compose up -d

Check:

docker compose ps
docker logs perplexica --tail=100

This approach suits most people: one container, built-in SearXNG, and data persisted with a volume.

Assume you plan to use:

search.example.com

Add an A record in DNS:

search.example.com -> your_server_ip

Install Caddy:

apt install -y debian-keyring debian-archive-keyring apt-transport-https curl
curl -1sLf 'https://dl.cloudsmith.io/public/caddy/stable/gpg.key' | gpg --dearmor -o /usr/share/keyrings/caddy-stable-archive-keyring.gpg
curl -1sLf 'https://dl.cloudsmith.io/public/caddy/stable/debian.deb.txt' > /etc/apt/sources.list.d/caddy-stable.list
apt update
apt install -y caddy

Edit the configuration:

nano /etc/caddy/Caddyfile

Write:

search.example.com {
    reverse_proxy 127.0.0.1:3000
}

Check and reload:

caddy validate --config /etc/caddy/Caddyfile
systemctl reload caddy

Now visit:

https://search.example.com

Caddy will automatically apply for and renew HTTPS certificates.

If you use Caddy reverse proxy, you only need to open SSH, HTTP, and HTTPS:

ufw allow 22/tcp
ufw allow 80/tcp
ufw allow 443/tcp
ufw enable
ufw status

It is not recommended to expose port 3000 directly to the public internet:

ufw deny 3000/tcp

If Docker above has already bound to 127.0.0.1:3000:3000, port 3000 is not accessible from outside anyway, which is more secure.

After opening the web page, first configure the AI Provider.

Common choices:

If you're just using it personally, I suggest first getting the flow working with an API Provider. Once you're sure you really need a local model, then tinker with Ollama.

After configuration is complete, ask a simple question to test:

最近 Docker Compose 部署自托管服务有哪些常见坑？

Focus on three things:

• Whether it can return an answer
• Whether there are cited sources
• Whether the search results are relevant to the question

If there are no search sources and it just looks like an ordinary chatbot, then check whether SearXNG is working properly.

If you want to connect Ollama, first install it on the host machine:

curl -fsSL https://ollama.com/install.sh | sh

Pull a model:

ollama pull qwen2.5:7b

Test Ollama:

curl http://127.0.0.1:11434/api/tags

If Perplexica is in a Docker container, it cannot use localhost to access the host machine's Ollama.

On Linux, a more reliable approach is to add this in Compose:

extra_hosts:
  - "host.docker.internal:host-gateway"

Complete example:

services:
  perplexica:
    image: itzcrazykns1337/perplexica:latest
    container_name: perplexica
    restart: unless-stopped
    ports:
      - "127.0.0.1:3000:3000"
    extra_hosts:
      - "host.docker.internal:host-gateway"
    volumes:
      - perplexica-data:/home/perplexica/data

volumes:
  perplexica-data:

Then in the web page, fill in the Ollama address as:

http://host.docker.internal:11434

Don't fill in http://localhost:11434. That's the container itself, not the host machine.

The single-container full image is suitable for quick deployment, but if you want finer control over search engines, or want multiple tools to share one SearXNG, you can deploy SearXNG separately and connect it with the slim version of Perplexica.

For example:

• Perplexica uses SearXNG
• Open WebUI also connects to SearXNG for Web Search
• You want to adjust search engines, language, and safe search separately
• You want to control SearXNG's access policy

In this case, you can use perplexica:slim-latest.

Create the directory:

mkdir -p /opt/perplexica/searxng
cd /opt/perplexica

Create the SearXNG configuration:

cat > searxng/settings.yml << 'EOF'
use_default_settings: true

server:
  secret_key: "replace-with-a-long-random-secret"

search:
  formats:
    - html
    - json
EOF

Note that json here is critical. Perplexica needs JSON results when calling SearXNG. If it is not enabled, you will often encounter the situation where "SearXNG opens fine, but Perplexica has no search results."

Write the Compose file:

services:
  searxng:
    image: searxng/searxng:latest
    container_name: perplexica-searxng
    restart: unless-stopped
    volumes:
      - ./searxng:/etc/searxng
    networks:
      - perplexica-net

  perplexica:
    image: itzcrazykns1337/perplexica:slim-latest
    container_name: perplexica
    restart: unless-stopped
    environment:
      - SEARXNG_API_URL=http://searxng:8080
    ports:
      - "127.0.0.1:3000:3000"
    volumes:
      - perplexica-data:/home/perplexica/data
    depends_on:
      - searxng
    networks:
      - perplexica-net

volumes:
  perplexica-data:

networks:
  perplexica-net:

Start:

docker compose up -d

Check:

docker compose ps
docker logs perplexica-searxng --tail=80
docker logs perplexica --tail=80

Test SearXNG JSON:

docker exec perplexica-searxng sh -c "wget -qO- 'http://127.0.0.1:8080/search?q=vps&format=json' | head"

If there is no JSON output here, fix SearXNG first, and don't rush to change Perplexica.

Self-hosting Perplexica does not equal complete anonymity.

What you can control is:

• The search entry point is under your own domain
• SearXNG does not profile users
• You can choose which search backends to use
• Query records are not by default handed to an AI search platform

But if you connect to cloud models like OpenAI, Claude, or Groq, questions and search summaries are still sent to the corresponding API provider.

So a more accurate way to put it is:

• Using cloud models: the search entry point and search aggregation are self-hosted, and model processing is at a third party
• Using Ollama: search and model are kept on your own machine as much as possible, but the VPS configuration needs to be higher

Don't understand "self-hosting" as "all data never leaves home." It depends on the model and search sources you connect to.

Perplexica's key data is in the Docker volume.

First look at the volume:

docker volume ls | grep perplexica

If you use a named volume, you can back it up like this:

mkdir -p /opt/backups
docker run --rm \
  -v perplexica-data:/data \
  -v /opt/backups:/backup \
  alpine \
  tar -czf /backup/perplexica-data-$(date +%F).tar.gz -C /data .

If you use a bind mount, such as /opt/perplexica/data, just package the directory directly:

tar -czf /opt/backups/perplexica-data-$(date +%F).tar.gz /opt/perplexica

Then sync to object storage:

rclone copy /opt/backups/perplexica-data-$(date +%F).tar.gz r2:backup-bucket/perplexica/

At minimum back up:

• Perplexica data volume
• Compose file
• SearXNG settings.yml
• Caddyfile
• The environment variables and model configuration notes you use

Single-container approach:

docker pull itzcrazykns1337/perplexica:latest
docker stop perplexica
docker rm perplexica
docker run -d \
  --name perplexica \
  --restart unless-stopped \
  -p 127.0.0.1:3000:3000 \
  -v perplexica-data:/home/perplexica/data \
  itzcrazykns1337/perplexica:latest

Compose approach:

cd /opt/perplexica
docker compose pull
docker compose up -d

Back up before upgrading. After upgrading, check:

docker compose ps
docker logs perplexica --tail=100
curl -I http://127.0.0.1:3000

Then actually search a question in the web page to confirm that search, citations, and model answers all work properly.

These tools are easy to confuse.

My suggestions:

• Just want to search web pages: SearXNG
• Want "search + AI summary + citations": Perplexica
• Want to chat with models: Open WebUI
• Want to ask your own documents: AnythingLLM
• Want to build AI applications for others to use: Dify

Don't install every tool. Start by choosing one based on your core need.

If it were for my own use, I would do this:

• VPS: start from 2 cores 4GB
• Deployment: get the Perplexica full image running first
• Domain: Caddy reverse proxy HTTPS
• Ports: only open 80/443, do not open 3000
• Model: connect OpenRouter or OpenAI first to confirm the experience
• Later: if used often, then consider standalone SearXNG
• High privacy requirements: then connect Ollama, but the machine needs an upgrade

The greatest value of Perplexica is not "installing yet another AI tool," but turning search results into answers with traceable sources. For people doing technical research, SEO, competitor analysis, and operations troubleshooting, this saves more time than ordinary search.

But it is not omnipotent either. Poor search source quality, models summarizing randomly, and inaccurate citations can all happen. The real prerequisite for long-term usability is that you are willing to maintain SearXNG, model configuration, reverse proxy, and backups, rather than installing it and forgetting about it.