Self-Hosted Perplexity: Run Perplexica for Private AI Search

Self-Hosted Perplexity: Run Perplexica for Private AI Search

TThumper Team2026-02-18T10:00:00Z7 min read
perplexicasearchragprivacy

Run Perplexica locally for AI-powered search with citations and no tracking. Your queries stay private.

What is Perplexica?

Perplexica is an open-source AI search engine inspired by Perplexity. It combines web search with a local LLM to provide cited, conversational answers to your questions. Unlike Perplexity's cloud service, Perplexica runs entirely on your machine.

Your search queries are never logged, tracked, or used for advertising. This is AI search with full privacy.

How It Works

Perplexica uses a RAG (Retrieval-Augmented Generation) pipeline:

  1. Your question is sent to a search engine (SearXNG, self-hosted)
  2. Relevant web pages are fetched and parsed
  3. The content is chunked and ranked by relevance
  4. A local LLM synthesizes an answer with inline citations

The LLM sees real web content, so answers are grounded in sources rather than hallucinated.

Installation

Install Perplexica from the Thumper-Run catalog. Thumper sets up SearXNG as a dependency and configures the LLM connection through Ollama. The install pipeline handles all wiring.

Choosing a Model

For search-and-summarize tasks, you want a model with strong instruction following and factual accuracy:

  • Llama 3.1 8B – fast, good enough for most queries
  • Qwen 3 14B – better at synthesizing complex multi-source answers
  • Mistral Nemo 12B – excellent at following citation formatting instructions

Use Cases

  • Research – get cited answers without wading through SEO spam
  • Competitive analysis – search without your queries being tracked by Google
  • Legal/medical research – sensitive queries that should not be logged anywhere
  • Offline-capable – pre-index local documents for air-gapped search

Privacy Benefits

Traditional search engines track every query. Even Perplexity's cloud service stores your search history. With Perplexica running locally:

  • No query logging – searches exist only in your local history
  • No tracking – SearXNG proxies requests without identifying you
  • No ads – results are ranked by relevance, not revenue
  • Optional sync – current supported relay-assisted paths are server-assisted; private E2EE remains gated

See our privacy post for the full data protection model.

Ready to try it? Download Thumper-Run free →

Share this article

About the Author

T

Thumper Team

The team behind Thumper-Run.

Related Articles