# Project export: Buenia AI This document was generated by HackStack to give an AI agent context about a hackathon project. Sections are labeled with their provenance; content marked as truncated was cut to keep this document small. ## Project metadata - Hackathon: UC Berkeley AI Hackathon 2025 - Tagline: Buenia AI provides real-time mission control for satellites, using advanced AI agents to continuously monitor performance, optimize operations on the fly, and flag anomalies immediately. - Devpost: https://devpost.com/software/buenia-ai - GitHub: not linked - Team: contributor stats unavailable ## Devpost submission (written by the team) ### Inspiration Space is full of dynamic challenges—satellites experience anomalies (power glitches, orbit deviations, sensor failures) that require timely intervention. We wanted to bring that excitement to developers and learners: an autonomous mission-control system that not only monitors satellites in real time but also lets you “talk” to a virtual astronaut to troubleshoot issues. ### What it does Real-time monitoring & anomaly detection: Continuously ingests simulated telemetry (e.g., orbit parameters, power/thermal readings) and flags anomalies as they occur. Interactive astronaut assistance: Offers a chat interface with a simulated astronaut persona—powered by an LLM—that explains the situation and guides you through diagnostic steps or fixes. Optimization engine: Recommends adjustments (e.g., attitude tweaks, power rebalancing) based on the detected anomaly, simulating how mission control would respond. Visualization dashboard: Live graphs/maps showing satellite status, alerts, and the current “fix plan” in progress. ### How we built it Frontend: Next.js app with WebSocket or SSE for streaming telemetry and alert updates; React components to render live charts (e.g., using a chart library) and chat UI for astronaut dialogue. Backend: Fastify Runs the simulation engine: produces synthetic real-time telemetry streams and injects anomalies at random or preset times. Hosts an anomaly-detection module: simple rules or lightweight ML model that watches incoming data and emits alerts. Manages LLM integration: when an anomaly is detected or the user asks, sends context (telemetry summary + anomaly details) to the LLM with a prompt framing the “astronaut” persona, then relays responses back to the frontend. Real-time plumbing: WebSocket-based channel for pushing telemetry and chat messages immediately to the UI. Simulation logic: A simplified orbital/thermal/power simulation coded in the backend; can be extended later but currently generates plausible data and anomalies using cesiumjs. Environment & APIs: No external APIs needed for core simulation; optionally fetch basic space weather data (e.g., real solar activity) if connectivity allowed. All hosted locally or on a hackathon-friendly cloud. ### Challenges we ran into Time constraints: Tight hackathon schedule meant we prioritized core real-time streaming and chat interface, leaving deeper physics models for later. Frontend–backend integration: Setting up WebSocket/SSE so telemetry updates and chat replies flow smoothly; handling reconnection and UI state when streams drop. Prompt engineering for astronaut persona: Crafting prompts so the LLM reliably explains anomalies in an accessible way and suggests plausible troubleshooting steps without going off-topic. Simulating believable telemetry: Designing a simple engine that feels realistic enough but remains lightweight to run in real time. ### Accomplishments we're proud of Working real-time data pipeline: Live telemetry stream in the UI with anomaly alerts popping up as soon as they occur. Interactive simulation: The user can click on an alert and immediately chat with the “astronaut” to diagnose and decide corrective actions. End-to-end flow: From simulated sensor data → anomaly detection → LLM-driven guidance → user action → updated simulation state, all in one seamless loop. Demo-ready visualization: Clean charts and status indicators that make it easy to see how the satellite behaves over time and how fixes resolve issues. ### What we learned Building real-time streams: Hands-on with WebSocket/SSE integration in Next.js and backend, managing state updates and reconnections. Prompt design for domain-specific personas: How to frame context (telemetry summary, anomaly details) so the LLM acts like an informed astronaut rather than generic chat. Simulation basics: Creating a lightweight model that balances realism with performance—injecting anomalies in a controllable way. ### What's next Gamified educational mode: Turn this into a game for kids or learners: present missions with increasing difficulty, track “mission success” scores, unlock new modules (e.g., multi-satellite coordination). Enhanced simulation fidelity: Integrate more accurate orbital mechanics or space-weather data via public APIs when available, so scenarios reflect real-world conditions. Team collaboration mode: Let multiple users join a mission control session, each with a role (e.g., telemetry analyst, comms officer, power engineer), collaborating via chat with the astronaut. Offline/remote astronaut training: Package scenarios that educators can use in classroom settings to teach systems thinking and problem-solving under pressure. Voice interface: Add speech-to-text for user commands and text-to-speech for astronaut responses, making the experience more immersive. Analytics & feedback: Track how users respond to anomalies, which fixes they choose, and refine the simulation or astronaut guidance based on common mistakes. Deploy for real data hookup: Eventually connect to live (or near-live) satellite telemetry for demonstration satellites or cubesats, so users can practice on actual missions in a sandbox. ## README (from the GitHub repository) No README available. ## Detected evidence (automated analysis) No repository was indexed for this project. Claimed technologies below could not be checked against code. - Next.js (technology) — claimed on Devpost, not found in the code - PostgreSQL (technology) — claimed on Devpost, not found in the code ## Codebase structure No repository index available. ## Key source files No repository index available; no source files included.