TID-AD-ASTRA Decoding the Universe, One Planet at a Time

Built for the 2025 NASA Space Apps Challenge – Team TangoisdownHQ

🛰️ Overview

TID-AD-ASTRA is an explainable AI system that analyzes and interprets planetary data from NASA’s Exoplanet Archive and the Open Exoplanet Catalogue. It predicts potential exoplanet habitability while explaining the why behind every decision — giving scientists, educators, and explorers transparent insight into how AI understands alien worlds.

🧩 Note: Users can upload new datasets to ml/app/data/uploads/ or fetch NASA updates automatically using python fetch_data.py --source nasa.

The name means “To the Stars”, symbolizing our mission to make deep-space data more interpretable, accessible, and open to everyone.

🧠 Core Features

1. Explainable AI for Planetary Habitability

Predicts exoplanet classification and habitability index.

Generates interpretable explanations for each prediction.

Detects missing or incomplete data and provides contextual diagnostics.

2. Data Lineage & Provenance

Integrates both NASA Exoplanet Archive and Open Exoplanet Catalogue.

Tracks the dataset origin for each trained model artifact.

Ensures transparency in AI learning sources and evolution.

3. Model Management Console (FastAPI)

Full model registry with:

Metadata (registry.json)

Lineage traceability

Explainability endpoints (/models/explain, /models/lineage)

Auto-loads the latest trained model and exposes prediction APIs.

4. CLI Mission Console

Terminal-based “Mission Control” interface for analyzing planets.

Displays:

Prediction summary

Habitability index

Missing-data diagnostics

Dataset provenance

🪐 Example Output 🧠 Summary: BD+20 2457 b is predicted as class 1 with confidence 0.77. Top influencing factors: feature_1, feature_0. Habitability index: 0.00 — unlikely to support Earth-like life.

🧩 Diagnostics: Missing or incomplete data for fields: discovery_year, radius_earth, host_star.temperature. Default estimates were used where possible.

🧬 Architecture Layer Description Data NASA Exoplanet Archive (nasa_exoplanets.csv) + Open Exoplanet Catalogue (open_exoplanet_catalogue.csv) Model Scikit-learn classifier with feature explainability Backend FastAPI service exposing /models, /planets, and /explain endpoints Interface CLI console (make run-console) and cURL API examples Storage Local model registry and artifacts (models/artifacts/registry.json)

🚀 Quick Start

1. Clone the Repository git clone https://github.com/TangoisdownHQ/TID-AD-ASTRA.git cd TID-AD-ASTRA/ml

2. Set Up Environment python3 -m venv .venv source .venv/bin/activate pip install -r requirements.txt

3. Run the Backend make run

4. Test the Explainability API curl -X POST "http://127.0.0.1:8000/models/explain"
   -H "Content-Type: application/json"
   -d '{"planet_name":"Kepler-442b","features":[]}' | jq

5. Run the Mission Console make run-console

📤 Add Your Own Datasets

TID-AD-ASTRA is designed for open exploration — users can plug in their own exoplanet datasets, telescope data, or simulated planetary environments.

-- 1. Upload CSV Files

Place your dataset(s) inside the upload directory:

ml/app/data/uploads/ └── your_exoplanet_data.csv

Your CSV should follow the base schema below (flexible — missing values are handled automatically):

Column Description planet_name Planet identifier mass_earth Planetary mass (in Earth masses) radius_earth Planetary radius (in Earth radii) orbital_period_days Orbital period semi_major_axis_au Semi-major axis equilibrium_temperature_k Temperature in Kelvin eccentricity Orbital eccentricity discovery_year Discovery year distance_from_earth_ly Distance from Earth (light-years) host_star_temperature Host star temperature (K) host_star_spectral_type Host star type (e.g., G2V)

-- 2. Auto-Fetch from NASA or Open Exoplanet Catalogue

Run the data fetcher to automatically download datasets:

python fetch_data.py --source nasa

or

python fetch_data.py --source open

This will download and store fresh copies under:

ml/app/data/datasets/

Custom URLs are also supported:

python fetch_data.py --url https://example.com/exoplanets.csv

-- 3. Train Using Custom Data

Once your CSV is added or fetched, train the model with:

python -m app.models.classifier --train ml/app/data/uploads/your_exoplanet_data.csv

Model metadata and dataset lineage will automatically appear in:

models/artifacts/registry.json

Check lineage with:

curl http://127.0.0.1:8000/models/lineage | jq

-- 4. Large Dataset Handling

For CSVs larger than 50 MB:

Use Git LFS to store them efficiently:

git lfs install git lfs track "*.csv" git add .gitattributes git commit -m "Track large datasets with Git LFS"

Or host them externally (e.g., Zenodo, Hugging Face Datasets, NASA Open Data) and add the URL to fetch_data.py.

🪙 Dataset Sources Source Description NASA Exoplanet Archive https://exoplanetarchive.ipac.caltech.edu/

Open Exoplanet Catalogue https://github.com/OpenExoplanetCatalogue/open_exoplanet_catalogue/ 🔭 Technology Stack Category Technology Language Python 3.11 Framework FastAPI ML Library Scikit-learn Explainability SHAP / Feature importance Data Layer CSV + JSON registries Interface CLI (Rich-based), REST API

🌍 Vision

TID-AD-ASTRA was designed as a foundation for autonomous, explainable AI agents that can:

Adaptively reason about planetary data across star systems.

Integrate with future NASA APIs and spaceborne sensors.

Support real-time decision support for interplanetary logistics.

Team TangoisdownHQ Role Name / Handle Focus Founder & Engineer @TangoisdownHQ Cybersecurity, AI Infrastructure, Explainability, System Design

🛰️ Contact

For collaboration, testing, or academic exchange:

📫 Email: tangoisdown1@proton.me

💻 GitHub: TangoisdownHQ

🌐 Live Demo (optional): https://tid-adastra.fly.dev

🏁 Submission Info Field Value Region NASA Space Apps Challenge 2025 Team Name TangoisdownHQ Project TID-AD-ASTRA

---

⚡ License

This project is open source under the MIT License. NASA datasets and related content are used under the NASA Open Data policy.

“To the stars — and beyond the noise.”