Note

... i wanted to explore the rag stuff, so this is the simplest implmentation of rag.

... it is my attempt project to understand how search engine works.

Important

... WhaleSearch is a RAG-powered search engine that combines Google search results with local LLM processing using Phi-3-mini to provide detailed answers with source citations. It supports both regular web searches and image searches, with results enhanced through vector embeddings and ChromaDB for efficient information retrieval.

Installation

Prerequisites

1. Python: Version 3.8 or higher is required.
2. SerpAPI API Key: get API key from SerpAPI.
3. Model: Download a model like Phi-3-mini-4k-instruct-fp16.gguf and place it in an accessible location.

Steps

Important

... gpu poor ?

... you can run here :

1. Clone the repository:

   git clone https://github.com/your-repo/whalesearch.git
   cd whalesearch
   

2. Download the llm weights

   !wget https://huggingface.co/microsoft/Phi-3-mini-4k-instruct-gguf/resolve/main/Phi-3-mini-4k-instruct-fp16.gguf
   
   

3. Install all required libraries

   pip install -r requirements.txt
   

4. Run the app through gradio

   cd src
   python3 app.py 
   

5. Enter your serapi-key and explore your curiosity.

How to use

1732274576503487.mov

Future work (TODO)

• Add support for more search APIs for broader data coverage.
• Enable multilingual query handling.
• Optimize performance for low-resource environments.
• also make it useful for complet local search
• implement llamaCPP from scratch to inference phi-3
• should solve math problem
• keyword search need be added

How i build it ? (so simple lol)

work in process...

Resources (must go through stuff)

• [1] What is RAG ?
• [2] Semantic Search and RAG(9th chapter)
• [3] What is Vector Database ?
• [4] Understanding how LLM inference works with llama.cpp