CS337 Project 2 Part 2: LLM-Based Cooking Assistant

An AI-powered cooking assistant that helps users follow online recipes using Google's Gemini API. The assistant can answer questions about ingredients, steps, timing, substitutions, and equipment while tracking the user's progress through a recipe.

Features

• Recipe Scraping: Automatically extracts recipe content (title, ingredients, steps) from web URLs
• Interactive Chat: Natural language conversation with an AI assistant about the recipe
• Smart Question Answering: Handles questions about:
  • Ingredient substitutions
  • Step clarifications
  • Cooking times and temperatures
  • Equipment requirements
  • Progress tracking through recipe steps
• Context-Aware: Maintains conversation history for follow-up questions
• Safety-Focused: Provides food safety guidance and warnings

Project Structure

.
├── assistant.py          # Main application with chat loop
├── scrape.py            # Web scraping functionality for recipes
├── prompt.py            # System prompt for the AI assistant
├── recipe_api.py       # Use the parse function
├── gemini_test.py       # Test script for Gemini API
├── environment.yml      # Conda environment configuration
└── README.md            # This file

Installation

Prerequisites

• Python 3.11
• Conda (Anaconda or Miniconda)
• Google Gemini API key

Setup

1. Clone the repository

   git clone <repository-url>
   cd CS337Proj2Part2

2. Create the Conda environment

   conda env create -f environment.yml

3. Activate the environment

   conda activate cs337proj2

4. Set up environment variables

   Create a .env file in the project root:

   touch .env

   Add your Gemini API key to the .env file:

   GEMINI_API_KEY=your_api_key_here
   

   To get a Gemini API key:

   • Visit Google AI Studio
   • Sign in with your Google account
   • Create a new API key

Usage

Running the Assistant

python assistant.py

Example Session

Paste a recipe URL: https://www.allrecipes.com/recipe/10813/best-chocolate-chip-cookies/
Scraped recipe:

Title: Best Chocolate Chip Cookies
Ingredients:
- 1 cup butter, softened
- 1 cup white sugar
- 1 cup packed brown sugar
...

--- (truncated preview) ---

You: Can I use whole wheat flour instead of all-purpose flour?
Assistant: Yes, you can substitute whole wheat flour, but it will make the cookies denser and give them a nuttier flavor. I recommend using half whole wheat and half all-purpose flour for better texture...

You: What temperature should I bake them at?
Assistant: According to the recipe, bake the cookies at 375°F (190°C)...

You: quit

Testing Gemini API

To test your API connection:

python gemini_test.py

Dependencies

• python-dotenv: Load environment variables from .env file
• google-genai: Google Gemini API client
• requests: HTTP library for web scraping
• beautifulsoup4: HTML parsing library

See environment.yml for specific versions.

How It Works

1. Scraping: The scrape.py module fetches the recipe webpage and extracts structured information (title, ingredients, steps)

2. Context Building: The assistant combines:

   • System prompt (defines behavior and capabilities)
   • Scraped recipe content
   • Conversation history
   • Current user question

3. LLM Processing: Sends the combined context to Gemini API for natural language understanding and response generation

4. Response: The assistant provides helpful, context-aware answers

System Prompt

The assistant is guided by a detailed system prompt that defines:

• Capabilities (answering questions, clarifying steps, tracking progress)
• Interaction style (concise, friendly, step-by-step)
• Safety guidelines (food safety, warnings)
• Formatting preferences

See prompt.py for the complete system prompt.

Limitations

• Recipe scraping may not work on all websites due to varying HTML structures
• The assistant relies on the quality of the scraped content
• API rate limits may apply based on your Gemini API tier
• Conversation history grows with each exchange (may hit token limits on very long conversations)

Environment Management

Update environment:

conda env update -f environment.yml --prune

Remove environment:

conda deactivate
conda env remove -n cs337proj2

Export current environment:

conda env export > environment.yml