beavermaj/FoodPrint

FoodPrint – Preserving African Indigenous Food Knowledge # FoodPrint – Preserving African Indigenous Food Knowledge FoodPrint is an AI-powered web application that identifies African plants, fruits, berries, roots, and leaves from photos using OpenAI's GPT-4o vision model. ## How Codex Was Used I used **Codex** (OpenAI's agentic coding tool) to generate the entire application: - Full Flask backend (`app.py`) with image preprocessing and API integration - Mobile-first frontend (`static/index.html`) with drag-and-drop upload - Deployment configuration (`Procfile`, `requirements.txt`, `runtime.txt`) - The complete project structure and all core functionality Codex accelerated development by generating production-ready code from natural language prompts, allowing me to build a full-stack AI app in hours rather than days. ## How GPT-4o Was Used I used **GPT-4o** (OpenAI's vision model) for: - Plant identification from user-uploaded photos - Generating structured JSON output with: - Common name, local names (Yoruba, Igbo, Swahili, Hausa) - Nutritional profile (vitamins, minerals, sugar, starch, protein, fiber) - Eco-Score (carbon footprint, water usage, sustainability tips) - Health goal insights, traditional uses, and cultural significance I engineered a custom system prompt that positions GPT-4o as an expert African Ethnobotanist, ensuring accurate, culturally-aware identifications. ## Tech Stack - **Backend:** Flask (Python) - **AI Model:** OpenAI GPT-4o (vision) - **AI Agent:** Codex - **Image Processing:** Pillow (PIL) - **Frontend:** HTML, CSS, JavaScript (mobile-first) - **Deployment:** Render + Gunicorn ## Live Demo foodprint-pl9c.onrender.com ## Setup Instructions 1. Clone the repository 2. Create a `.env` file with your `OPENAI_API_KEY` 3. Run `pip install -r requirements.txt` 4. Run `python app.py` 5. Open `localhost`