Captain Exoplanet: An Open, Browser-Based Interface for Exoplanet Candidate Classification Using Machine Learning Pipeline

Captain Exoplanet v1.0.0 - Initial Public Release

Live Demo Hunt for exoplanets with AI.

Main Features

• Browser-based interface: Easily classify exoplanet candidates using any modern browser.
• Next.js + FastAPI stack: Frontend built in React (Next.js), backend in FastAPI.
• ML-powered predictions: Make use of the latest trained model for candidate classification.
• No local install needed: Input feature values via form or file upload—get instant results.
• Minimal, research-focused UI: Designed for clarity; supports fast review and collaboration.
• Provenance tracking: Each prediction returns metadata including the model version.

Technical Overview

This release introduces the first stable pipeline for exoplanet light curve deconfusion:

• Flow: [User] → Next.js (apps/web) → FastAPI (apps/api) → Trained Model (pipeline/artifacts)
• The web client does not run models locally; it sends requests to post /predict for inference.
• API retrieves the latest exported artifact and returns both prediction and model_version.

Typical Prediction Usage:

• Request (JSON): { "features": { "<feature_name>": 0.0, "<feature_name_2>": 1.23 } }

• Response (JSON): { "prediction": ["<class_or_value>"], "model_version": "<version>" }

• Change API endpoint via MODEL_API_CLASSIFY in apps/web/.env.local (e.g., http://localhost:8000/predict).

• Health check endpoint available at GET /health (returns {"status":"ok"}).

Developed for NASA Space Apps 2025. For contributing credits and setup instructions, see the README.