Stephenthompson002/froggytalk

A personalized AI assistant (FroggyTalk Prototype) built with FastAPI and GPT to provide immigrants and the African diaspora with warm, actionable, and location-specific relocation advice and action plans. # FroggyTalk Prototype ## Overview This is a prototype for FroggyTalk, a personalized AI assistant designed to help users (with a focus on African diaspora experiences) with purposes like immigration, job search, education, visa renewal, or general inquiries. It collects user data via a simple React frontend, stores it in a SQLite database, generates personalized responses using an LLM (OpenAI or fallback), and supports a "training" mechanism via synthetic examples and few-shot prompting. The prototype fulfills the take-home assignment requirements: - User data collection and storage. - Personalized LLM responses with base and trained models. - Synthetic training examples generation. - API endpoints for collection, generation, training, and health checks. - Minimal frontend with form, model selection, and response display. The project is structured into two main folders: - **backend/**: Contains the Python/FastAPI server, database logic, LLM integration, and training script. - **frontend/**: Contains the React application for the user interface. ## Architecture - **Frontend (frontend/ folder)**: Built with React (using Vite for build tooling and `react-lottie-player` for animations). Key files: - `src/App.jsx`: Main component with tabs for "Create Profile" (form to collect user data), "Generate Response" (input user ID, select model, generate), and "History" (view past responses). - Form fields for full name, countries, purpose, language, and additional info. - Model selection (base or trained) via clickable cards. - Displays formatted responses (parses markdown for bold, lists). - `src/assets/`: Lottie animations (frog-success.json, frog-wave.json). - Quick-select for saved users and a fun frog animation. - Communicates with backend via fetch to `/collect`, `/generate`, etc. - Other files: `index.html`, `main.jsx`, `App.css`, `index.css`, `.gitignore`, `eslint.config.js`, `package.json`, `vite.config.js`. - **Backend (backend/ folder)**: FastAPI application. …