Perplexica Installation Guide
This guide will help you install Perplexica—your own private, AI-powered search engine—on almost any Linux system.
** Part 1: Choose Your "Brain" (Model Selection)**
Before we start, decide which AI model fits your hardware.
- Low End (2GB - 4GB VRAM):
qwen3:1.7b(Fast & Light) - Standard (4GB - 6GB VRAM):
qwen2.5:3b(Best Balance) - High End (8GB+ VRAM):
qwen2.5:7b(Smartest)
** Part 2: The Foundation (Docker)**
We need Docker to run the application. Choose the command for your OS.
Option A: Ubuntu / Debian / Fedora / Linux Mint
Option B: Arch Linux / Manjaro / CachyOS
** CRITICAL STEP FOR EVERYONE:
Run this command to allow your user to control Docker, then REBOOT YOUR COMPUTER**.
** Part 3: The Brain (Ollama)**
Once you have rebooted, install Ollama and your chosen model.
1. Install Ollama:
2. Download the Model:
(Replace 3b with 1.7b or 7b if you chose a different size)
3. Build the Bridge:
By default, Ollama ignores Docker. We need to tell it to listen.
- Edit the service file:
- Scroll down to
[Service]and add this line:
- Save (Ctrl+O, Enter) and Exit (Ctrl+X).
- Restart Ollama:
** Part 4: Find Your IP**
Docker needs your computer's real IP address to talk to Ollama.
Run this command:
- Write down the number you see (e.g.,
192.168.1.50). - If that command gives an error, try this backup:
ip route get 1 | awk '{print $7}'
** Part 5: Install Perplexica**
Choose ONE command based on your privacy needs.
** Option A: Laptop Mode (Secure)
Best for coffee shops & public Wi-Fi. Only accessible from THIS computer.
** Option B: Server Mode (Open)
Best for home servers. Allows access from your phone or other PCs on your LAN.
To make Perplexica start automatically, you can add --restart always to the Docker install command
** Part 6: Final Configuration**
- Open your browser to:
http://localhost:3000 - Go to Settings > Components.
- Chat Model Provider: Select
Ollama. - Ollama Api URL:
- Type:
http://YOUR_IP_ADDRESS:11434 - (Example:
http://192.168.1.50:11434) - ** IMPORTANT:** Do NOT use
localhosthere!
- Select Chat Model: Choose the Qwen model you downloaded.
- Select Embedding Model: Choose
all-MiniLM-L6-v2. - Click Save / Finish.
** Troubleshooting**
If Perplexica won't load after you restart your computer, run these three commands to kickstart everything:
If your system has a firewall (UFW or Firewalld), it will block the connection. Run the command for your firewall:
For Ubuntu/Debian (UFW):
For Fedora/CentOS (Firewalld):