Korean Personal Color Analysis (PCA) AI Assistant

This project provides an AI-powered assistant for Korean Personal Color Analysis (PCA). It analyzes an image of a person using Google's Gemini 1.5 Flash model, determines their most suitable PCA seasonal palette (Spring, Summer, Autumn, Winter, with subtypes), and generates a visual color palette based on the recommendations.

Features

• Detailed Image Analysis: Utilizes Gemini 1.5 Flash to extract comprehensive visual details from an input image, including skin tone, undertone, facial features, hair, clothing, and background.
• Korean PCA Determination: Based on the detailed analysis, the AI model recommends the most suitable Korean PCA seasonal palette.
• Specific Color Recommendations: Provides concrete color suggestions for makeup, clothing, and accessories relevant to the determined palette.
• Visual Color Palette Generation: Automatically creates a .png image showcasing the recommended colors, saved locally.
• LangChain & LangGraph Integration: Leverages LangChain for model interaction and LangGraph for orchestrating the multi-step PCA workflow.
• Interactive CLI: A simple command-line interface for easy interaction.

How It Works

The application uses a multi-step workflow orchestrated by langgraph:

1. User Input: The user provides a command analyze <image_path>.
2. Image Loading: The load_image_as_bytes function reads the specified image file and converts it into a byte stream suitable for API submission.
3. Detailed Image Analysis (analyze_image_detailed):
   • The ChatGoogleGenerativeAI model (Gemini 1.5 Flash) receives the image bytes and a prompt asking for a detailed analysis of the person and their features, including skin tone, hair, clothing, etc.
   • The model returns the analysis in a structured JSON format.
4. PCA Prompt Generation (generate_pca_prompt):
   • The raw analysis content is parsed.
   • A specific prompt tailored for Korean Personal Color Analysis is constructed, incorporating all relevant details from the image analysis.
5. PCA Result Generation (generate_pca_result):
   • The ChatGoogleGenerativeAI model receives the PCA-specific prompt.
   • It processes the prompt and generates a PCA report, including the recommended palette, an explanation, and a list of specific color names, all in JSON format.
6. Color Palette Image Generation (generate_color_palette_image):
   • The list of recommended color names is extracted from the PCA result.
   • Using the Pillow (PIL) library, a visual representation of these colors is created as a color_palette.png image, with each color represented by a swatch.
   • Note: The current implementation has a predefined color_map to convert color names to RGB values. For colors not in this map, a default grey is used. You might need to expand this map for more diverse recommendations.
7. Response: The final PCA report (text) and the path to the generated color palette image are displayed to the user in the console.

Setup

Prerequisites

• Python 3.9+

Environment Variables

You need to set your GOOGLE_API_KEY as an environment variable.

1. Obtain API Key: Get your Google API Key from Google AI Studio or Google Cloud.
2. Create .env file: In the root directory of your project, create a file named .env and add your API key:

   GOOGLE_API_KEY="your_google_api_key_here"
   

Installation

1. Clone the repository (if you haven't already).

2. Navigate to the project directory.

3. Install the required Python packages:

   pip install langchain==0.3.0 langgraph==0.4.8 langchain-google-genai==2.1.5 pillow==11.2.1 python-dotenv==1.0.1 pydantic==2.9.2 typing-extensions==4.12.2

   (Or, if you create a requirements.txt file from the above list, use pip install -r requirements.txt)

Usage

1. Ensure you have followed the Setup instructions and your .env file is correctly configured.

2. Run the Python script from your terminal:

   python your_script_name.py

   (Replace your_script_name.py with the actual name of your Python file, e.g., pca_analyzer.py)

3. The program will prompt you to enter a command. To perform PCA, use the format:

   Enter 'analyze <image_path>' or 'quit' to exit: analyze /path/to/your/image.jpg
   

   Replace /path/to/your/image.jpg with the actual path to an image file on your system (e.g., images/my_profile.png).

4. The assistant will process the image and print the PCA report in the console. A color_palette.png file will also be generated in the same directory where you run the script.

5. To exit the program, type quit or exit.

Example Interaction