RestoRecommender: AI-Powered Restaurant Recommendation System

RestoRecommender leverages the power of artificial intelligence (AI) to suggest restaurants and bars based on user preferences and review analysis. It utilizes Retrieval-Augmented Generation (RAG) to extract meaningful insights from customer reviews.

Functionality Breakdown:

• Data Collection (places.ipynb):

  • This Jupyter Notebook scrapes data from Google Maps's Places API and stores the extracted information in CSV files (places.csv and reviews.csv) within the data directory.

• RAG LLM Model Creation (rag_chatbit.ipynb):

  • This script outlines the process of building a RAG-based large language model (LLM). The steps involve:
    • Data loading from CSV files
    • Vector database construction
    • Embedding model training
    • LLM model creation
  • The notebook also showcases demonstration use cases and evaluations based on synthetically generated questions.

• Answer Generation (experiments.ipynb):

  • This Jupyter Notebook utilizes the questions.txt file containing user queries to generate corresponding answers and saves them in the Question-Answer.txt file. It allows for the inclusion of additional user questions for enhanced personalization.

• Web User Interface (web-UI folder):

  • This folder houses the necessary files to create a web-based user interface (UI) that facilitates interaction with the RestoRecommender chatbot. It employs the Flask framework.
  • To launch the UI:
    1. Navigate to the web-UI directory in your terminal.
    2. Run the command: python app.py
    3. Open your web browser and visit the following URL: http://127.0.0.1:5000/

Setting Up the Environment

1. Create a Virtual Environment (Recommended):

   • Virtual environments isolate project dependencies and prevent conflicts with other Python installations.
   • To create a new virtual environment named venv:
     • Linux/macOS: python3 -m venv venv
     • Windows: python -m venv venv
   • Activate the virtual environment:
     • Linux/macOS: source venv/bin/activate
     • Windows: venv\Scripts\activate

2. Install Required Libraries:

   • Navigate to the project's root directory.
   • Assuming you have a requirements.txt file listing project dependencies, install them using pip:
     • pip install -r requirements.txt

Running the Project

1. Make sure the virtual environment is activated (if used).
2. Navigate to the project's root directory in your terminal.
3. Run Jupyter Notebooks:
   • Open a terminal or command prompt within the project directory.
   • Use jupyter notebook to launch the Jupyter Notebook server.
   • Interact with the notebooks (places.ipynb, rag_chatbit.ipynb, and experiments.ipynb) to perform data collection, model creation, and further experiments.
4. Run Web UI (if applicable):
   • To launch the UI:
     1. Navigate to the web-UI directory in your terminal.
     2. Run the command: python app.py
     3. Open your web browser and visit the following URL: http://127.0.0.1:5000/
   • The UI interface created in this manned is just for development environment. However, this is enough for the purpose of this project.

Using the Gemini API

If you intend to leverage the Gemini API for their embedding and LLM model, you'll need an API key. Here's how to obtain one:

1. Visit the Google AI Studio: https://aistudio.google.com/app/apikey
2. Click on "Get API key" and follow the on-screen instructions.
3. Once you have the API key, store it securely inside a .env file.
4. Update the rag_chatbit.ipynb notebook with your API key where necessary.

Alternative Embedding Models and LLM Models

While the rag_chatbit.ipynb notebook demonstrates using the Gemini API, you can explore alternative embedding models and LLM architectures. Here are some resources to get you started:

• Embedding Models:
  • Sentence Transformers: https://huggingface.co/sentence-transformers
• LLM Models:
  • Hugging Face Transformers Library: https://huggingface.co/docs/transformers/en/index
  • OpenAI API: https://openai.com/ (requires account creation)

By following these instructions, you'll establish a clean environment, install necessary dependencies, and be ready to run the RestoRecommender project effectively.

References:

• https://pypi.org/project/googlemaps/
• https://github.com/googlemaps/google-maps-services-python/blob/master/googlemaps/places.py
• https://python.langchain.com/v0.1/docs/get_started/introduction
• https://ai.google.dev/gemini-api/docs
• https://github.com/facebookresearch/faiss
• https://ai.google.dev/gemini-api/docs/api-key