From Curiosity to Cosmos: Building a Custom GPT with NASA’s Multi-API Suite
Date Started: 03/17/2025
Overview
In an era where AI interfaces are quickly evolving beyond static Q&A bots, the next frontier is conversational agents that dynamically integrate with real-time data. That’s exactly what I set out to build with JedAI Space — a custom GPT designed to serve as a cosmic guide, fully connected to NASA’s unified API suite.
This project wasn’t just about exploring space data. It was about proving that large language models can be architected as interactive, domain-specific agents that combine natural language interaction with live, structured information.
Purpose and Use Case
The intent behind this build was clear: to create an AI assistant that allows users to explore NASA’s public datasets — not through static endpoints or documentation — but through a conversational interface.
Primary objectives:
- Turn natural language queries into API calls
- Serve rich, user-friendly outputs across multiple data types
- Maintain a coherent personality that contextualizes results with scientific insight and accessible language
Whether a user wants the Astronomy Picture of the Day, Mars rover imagery, current solar activity, or archival footage from the NASA image library, JedAI Space acts as the bridge — translating everyday questions into structured outputs.
Architectural Design
Integration Backbone
At the heart of this system is NASA’s publicly available API ecosystem. I consolidated the following endpoints into a unified OpenAPI schema:
- /planetary/apod: Astronomy Picture of the Day
- /mars-photos/api/v1/rovers/{rover}/photos: Mars Rover surface imagery
- /mars-photos/api/v1/manifests/{rover}: Mission manifests and milestones
- /donki endpoints: Solar events (coronal mass ejections, solar flares, SEP events)
- /search and /asset/{nasa_id}: NASA Image and Video Library metadata and access
The schema was uploaded as a plugin spec and linked directly to the GPT instance, allowing it to make authenticated calls to NASA’s servers in response to user prompts.
Persona Layer
To ensure contextual and consistent tone, I crafted a custom persona prompt:
“You are a space-enthusiast assistant with access to NASA’s unified API suite… You act like a passionate cosmic guide, ready to fetch space media, science, and rover insights on request.”
This shaped not only the model’s vocabulary and tone, but also how it interpreted ambiguous prompts — guiding it to clarify intent before querying the appropriate endpoint.
Response Formatting Each data type was rendered through a consistent visual logic:
- Section headers per category (e.g., APOD, Mars Rover Images, Solar Activity Report)
- Concise field labels (title, date, explanation, image URL)
- Clean, structured text responses that feel informative, not overwhelming
This balance of clarity and excitement was intentional — designed to make complex data feel accessible to any user.
Results and Reflection
JedAI Space is now capable of:
- Fetching daily APOD entries with explanations and media previews
- Surfacing Mars Rover photos by sol or Earth date
- Returning manifest data for mission timelines and exploration milestones
- Reporting on recent or active space weather events
- Searching the NASA media library by keyword, filtered by media type
But more importantly, it demonstrates what’s possible when three components converge:
- Reliable public APIs
- Well-scoped custom GPT personas
- Structured, intentional output formatting
This isn’t just about NASA. It’s a replicable blueprint for any domain-specific agent that needs to combine live data with high-context AI interaction.
What This Means for AI Builders
This project reinforces a few key lessons:
- APIs are the new nervous system for AI tools. With thoughtful integration, GPTs become gateways to real-world data.
- Persona design is non-optional. The clarity and consistency of output hinges on how well the model’s role is defined.
- AI is moving toward operational assistance. These aren’t just chatbots — they’re becoming knowledge agents capable of research, reporting, and synthesis.
As we scale this thinking into business and enterprise environments, it’s easy to imagine how AI agents like this could assist in everything from market research to contract review, data analysis, or compliance audits.