nanoPerplexityAI - Minimal Open Source Implementation

The simplest open-source Perplexity AI implementation in just 200 lines of Python. Combines Google search with LLM responses, featuring real-time streaming, source citations, and VS Code markdown preview integration.

Overview

nanoPerplexityAI demonstrates how a Perplexity-like search+AI system works in minimal code. It implements the full pipeline: query understanding, web search, content extraction, and cited response generation in approximately 200 lines of Python.

How It Works

1. Accept user input
2. LLM determines if search is needed and reformulates queries for Google
3. Google search retrieves relevant web pages
4. BeautifulSoup extracts page content
5. Content combined with system instructions and user query
6. LLM generates response with citations
7. Output streamed to markdown file

Features

• Google search integration with webpage content retrieval
• Real-time LLM response generation with markdown output
• VS Code preview for live response visualization
• Source citation in generated answers
• Stream-based completion
• Save/quit conversation controls

Tech Stack

Python, googlesearch-python, requests, BeautifulSoup4, lxml, OpenAI API

Installation

pip install googlesearch-python requests beautifulsoup4 lxml backoff openai
export OPENAI_API_KEY=your_key
python nanoPerplexityAI.py

Educational project ideal for understanding AI search architecture.