Perplexica: The Complete Guide to the Open Source AI Search Engine
What if you could run your own Perplexity.ai — on your own hardware, with your own models, and with total privacy? That's exactly what Perplexica delivers. With 31,600+ GitHub stars and a growing community, Perplexica is the leading open-source, privacy-focused AI answering engine that combines web search with LLM intelligence to deliver accurate, cited answers without sending your queries to a third party.
What Is Perplexica?
Perplexica is a self-hosted AI-powered answering engine. Think of it as Perplexity.ai, but:
- Open source (MIT License)
- Self-hosted — runs entirely on your hardware
- Privacy-first — your queries never leave your machine
- Provider-agnostic — works with local LLMs (Ollama) or cloud providers (OpenAI, Claude, Gemini, Groq)
Under the hood, Perplexica uses SearxNG as its search backbone — a privacy-respecting metasearch engine that aggregates results from multiple sources without tracking you.
Key Features
🤖 Multi-Provider LLM Support
Use the AI provider that fits your needs:
- Local models via Ollama — complete privacy, zero API costs
- OpenAI (GPT-4o, o1, o3)
- Anthropic (Claude Sonnet, Opus)
- Google (Gemini Pro, Flash)
- Groq — for ultra-fast inference (3-4 second answers)
Mix and match models for different tasks.
⚡ Smart Search Modes
Three modes optimized for different needs:
- Speed Mode — quick answers from top results
- Balanced Mode — everyday search with good depth
- Quality Mode — deep research with comprehensive analysis
🧭 Multiple Source Types
Search isn't limited to generic web results:
- Web — general internet search
- Academic — scholarly papers and research
- Reddit — community discussions
- YouTube — video content
- Wolfram Alpha — computational data and analysis
🧩 Smart Widgets
Contextual UI cards that appear automatically:
- Weather forecasts
- Calculator results
- Stock prices
- Quick lookups and data cards
📷 Image & Video Search
Find visual content alongside text results. Modern search isn't just about articles.
📄 File Upload & Analysis
Upload documents and ask questions about them:
- PDFs
- Text files
- Images
- Perplexica understands them all
🌐 Domain-Specific Search
Limit searches to specific websites when you know where to look. Perfect for searching within technical documentation, Wikipedia, or research databases.
💡 Smart Suggestions
Intelligent search suggestions as you type, helping you formulate better queries.
📚 Discover Feed
Browse trending articles and interesting content throughout the day — no search required.
🕒 Search History
Every search is saved locally. Your research is never lost.
Installation
Docker (Recommended — One Command)
docker run -d -p 3000:3000 \
-v perplexica-data:/home/perplexica/data \
--name perplexica \
itzcrazykns1337/perplexica:latest
This single command pulls and starts Perplexica with the bundled SearxNG search engine. Open http://localhost:3000 and configure your API keys in the setup screen.
Using Your Own SearxNG Instance
docker run -d -p 3000:3000 \
-e SEARXNG_API_URL=http://your-searxng-url:8080 \
-v perplexica-data:/home/perplexica/data \
--name perplexica \
itzcrazykns1337/perplexica:slim-latest
Building from Source
git clone https://github.com/ItzCrazyKns/Perplexica.git
cd Perplexica
docker build -t perplexica .
docker run -d -p 3000:3000 \
-v perplexica-data:/home/perplexica/data \
--name perplexica perplexica
Non-Docker Installation
git clone https://github.com/ItzCrazyKns/Perplexica.git
cd Perplexica
npm i
npm run build
npm run start
Open http://localhost:3000 to complete setup.
Architecture
Perplexica's system consists of five key components:
- User Interface — Clean, modern web UI built with Next.js
- Agents/Chains — Interpret user queries and route to appropriate search sources
- SearxNG — Privacy-respecting metasearch engine that aggregates results from multiple engines
- LLM — Processes and summarizes fetched information into coherent answers
- Embedding Models — Re-rank search results for relevance and accuracy
The result: you ask a question in natural language, and Perplexica searches the web, processes the results through an LLM, and returns a summarized answer with cited sources — all running on your own infrastructure.
Using as a Browser Search Engine
You can replace Google in your browser:
- Open browser settings → Search Engines
- Add:
http://localhost:3000/?q=%s - Set as default
Now every search from your address bar goes through Perplexica.
Perplexica API
Developers can integrate Perplexica's search capabilities into their own applications:
- Run searches programmatically
- Use multiple models
- Get structured answers with citations
Full API documentation is available in the docs.
Perplexica vs Alternatives
Category: This repo is an open-source, self-hosted AI search/answering engine. Alternatives must also be self-hosted AI search tools.
| Feature | Perplexica | Khoj | SurfSense | Scira |
|---|---|---|---|---|
| Focus | Privacy-first AI search engine | AI second brain + research agent | NotebookLM alternative for teams | Minimalistic AI search UI |
| Stars | 31.6K ⭐ | 33.2K ⭐ | 13.1K ⭐ | 11.5K ⭐ |
| Open Source | ✅ MIT | ✅ AGPL-3.0 | ✅ Apache-2.0 | ✅ AGPL-3.0 |
| Self-Hosted | ✅ Docker one-liner | ✅ Docker | ✅ Docker | ✅ Vercel/Docker |
| Search Backend | ✅ SearxNG (meta-search) | Web + personal docs | ✅ Slack, Notion, Gmail, GitHub, YouTube | Exa AI, Tavily |
| Local LLM Support | ✅ Ollama | ✅ Ollama, llama.cpp | ✅ Ollama | ❌ Cloud only |
| Search Modes | ✅ Speed/Balanced/Quality | Web + docs + research | Deep research | General search |
| File Upload | ✅ PDF, text, images | ✅ Docs, Obsidian, Emacs | ✅ Multi-source | ❌ |
| Smart Widgets | ✅ Weather, stocks, calc | ❌ | ❌ | ✅ Weather, flights, stocks |
| Domain Search | ✅ | ❌ | ❌ | ✅ Reddit, academic, YouTube |
| Image/Video Search | ✅ | Image generation | ❌ | ✅ |
| Browser Extension | ❌ | Obsidian + Emacs plugins | ✅ Chrome extension | ❌ |
| Mobile Access | Via browser | ✅ Phone, WhatsApp, desktop | Via browser | Via browser |
| Custom Agents | 🔜 Coming soon | ✅ | ✅ | ❌ |
| API | ✅ | ✅ | ✅ | ❌ |
| Discover Feed | ✅ | ❌ | ❌ | ❌ |
| Language | TypeScript | Python | Python | TypeScript |
When to choose Perplexica: You want the closest self-hosted replacement for Perplexity.ai with a polished web UI, SearxNG-powered meta-search, smart widgets, Docker one-liner install, and complete privacy. The most comprehensive AI search engine experience on your own hardware.
When to choose Khoj: You want an AI "second brain" that searches both the web AND your personal documents (Obsidian, Emacs, PDFs). Multi-platform access (phone, WhatsApp, desktop), custom agents, and automated research. More than search — it's a personal AI assistant.
When to choose SurfSense: You need to search across team data sources — Slack, Notion, Gmail, GitHub, YouTube — not just the web. Built for teams with a Chrome extension and multi-source integration. The NotebookLM alternative for organizations.
When to choose Scira: You want a beautiful, minimalistic AI search interface with specialized searches (flights, movies, weather, stocks). Lightweight and deployable on Vercel. Great for quick prototyping or as a starting point.
Upcoming Features
The Perplexica team is actively developing:
- More widgets and integrations
- Custom agent creation
- Authentication system
- Tavily and Exa AI integration
Conclusion
Perplexica has earned its 31,600 stars by proving that you don't need to surrender your privacy for AI-powered search. With SearxNG as its search backbone, support for local LLMs via Ollama, three smart search modes, file uploads, widgets, and a Docker one-liner install, Perplexica delivers a Perplexity.ai-level experience entirely on your own hardware. Whether you're a privacy advocate, a researcher, or simply someone who wants control over their search infrastructure, Perplexica is the open-source answer.