secure-ai-chatbot/ │

├── backend/

│ ├── app/

│ │ ├── api/

│ │ │ ├── endpoints/

│ │ │ │ ├── upload.py # Upload data files

│ │ │ │ ├── generate_code.py # Interact with Gemini API

│ │ │ │ ├── execute_code.py # Run code securely in Docker/E2B

│ │ │ │ └── status.py # Optional: status updates

│ │ │ └── init.py

│ │ │

│ │ ├── core/

│ │ │ ├── config.py # Settings, Gemini API keys, etc.

│ │ │ └── utils.py # Common functions (e.g., validate files)

│ │ │

│ │ ├── services/

│ │ │ ├── gemini_service.py # Gemini API logic

│ │ │ ├── docker_service.py # Docker/E2B execution logic

│ │ │ └── file_service.py # File handling logic

│ │ │

│ │ ├── main.py # FastAPI entry point

│ │ └── requirements.txt

│ │

│ ├── Dockerfile # Backend Dockerfile (optional)

│ └── .env # Environment variables (secret keys etc.)

│

├── frontend/

│ ├── public/

│ ├── src/

│ │ ├── components/

│ │ │ ├── Chatbot.jsx # Chat UI with prompts

│ │ │ ├── FileUpload.jsx # File upload interface

│ │ │ └── PlotDisplay.jsx # Display generated plots

│ │ ├── services/

│ │ │ ├── api.js # Axios config for backend calls

│ │ ├── App.js

│ │ └── index.js

│ ├── package.json

│ └── .env # Frontend API URLs etc.

│

├── sandbox/

│ ├── Dockerfile # Dockerfile for sandbox

│ ├── script_template.py # Template script Gemini fills

│ └── run_code.py # Executes Gemini-generated code safely

│

├── shared/

│ └── uploads/ # Temporarily stored user files

│

├── docker-compose.yml # Compose for backend, frontend, and sandbox

└── README.md

CODE PLAN:

README.md

Secure AI Chatbot for Data Processing

This project is a secure AI chatbot that:

• Accepts data file uploads from users
• Uses Gemini API to generate plotting/fitting Python code
• Executes code in a sandbox (Docker) for security

Project Structure

• backend/: FastAPI backend handling uploads, Gemini, Docker exec
• frontend/: React UI to upload, prompt, and display results
• sandbox/: Isolated Python execution environment
• shared/uploads: Shared volume for user files

Run Locally with Docker Compose

docker-compose up --build

Access:

• Frontend: http://localhost:3000
• Backend API: http://localhost:8000/api

Security Note

• All execution is isolated in the sandbox container with no internet access
• Add additional limitations via Docker or E2B for production

How to run the code:

activate virtualenv : new

backend :(new) C:\Users\aimar\OneDrive\Desktop\interview\secure-ai-chatbot\backend>uvicorn app.main:app --reload

frontend:C:\Users\aimar\OneDrive\Desktop\interview\secure-ai-chatbot\frontend>npm start

Important: need Gemini or any multimodel API key beacuse other multimodel is very heavy in weight