zihanxing / LLM_RAG_from_scratch

Implementation of RAG technique for LLM from scratch

Interior Designer with RAG from scratch

This project uses Retrieval Augmented Generation (RAG) based on GPT-3.5 to summarize "THE TOP 6 INTERIOR DESIGN TRENDS" by FEATHR. The aim is to provide a detailed and accurate analysis of the key interior design trends while enhancing content depth and accuracy through RAG.

Objective

To analyze and summarize key interior design trends, enhancing content depth and accuracy through RAG.

Data

The data for this project consists of a comprehensive report on "THE TOP 6 INTERIOR DESIGN TRENDS" by FEATHR. This report is in PDF format and contains detailed descriptions and analyses of prevailing interior design trends.

Pipeline

1. Text Extraction: Extract text from the PDF using pdfplumber.
2. Preprocessing: Chunk and preprocess text for analysis.
3. Data Storage: Store preprocessed data in CrateDB for efficient retrieval.
4. RAG Integration: Augment GPT-3.5 with data retrieval for enriched analysis.
5. Insight Generation: Generate summaries of interior design trends.

Model

The project uses GPT-3.5, enhanced with RAG, to improve the relevance and detail of generated summaries by incorporating specific data from the trends report.

Performance

• Without RAG: Baseline setup with GPT-3.5 provided initial insights but lacked specific trend details.
• With RAG: The addition of RAG improved the summaries' depth and relevance.

Conclusion

RAG's integration with GPT-3.5 offers significant benefits for generating detailed and contextually relevant content, showcasing its potential in specialized domains like interior design.

Acknowledgments

Inspiration, code snippets, etc.

https://github.com/architkaila/Chat-With-Documents

How to use

pip install -r requirements.txt
streamlit run app.py

Note: Please ensure that you have the required API keys and access to the CrateDB instance for the complete functionality of the application, and update the environment variables accordingly.