nanoPerplexityAI

The simplest and most intuitive open-source implementation of an open source perplexity.ai, a large language model(LLM) service which cites information from Google. No fancy GUI or LLM agents are involved, just 100 lines of python code. Please check out this video(to be uploaded soon) for more explanations!

Architecture

1. Get the user query
2. Search Google to find relevant webpage URLs and fetch texts
3. Use rerank model(cross encoder) to find relevance scores for each search results
4. Build a prompt using system prompt + webpage context + user query
5. Call the LLM API to generate a response
6. Format citations and save the LLM response into a markdown file to visualize it

Install

pip install googlesearch-python requests beautifulsoup4 lxml sentence-transformers openai 

Quick Start

export OPENAI_API_KEY=<Your OpenAI API KEY>
python nanoPerplexityAI.py

The script will prompt you to type your question, then it will generate a response in response.md

View the LLM Responses:

There are several ways to visualize the responses easily:

• Open in your editor, e.g., VScode
• Open in Markdown Playground
• Push them to your github repo

Check out the responses nanoPerplexityAI has already generated

Acknowledgements

Thank you perplexity.ai for the amazing idea and clarity-ai and Perplexica for coding inspirations on the open-source implementation of perplexity.ai.