Inspiration
Drone owners are able to capture impressive aerial imagery, but the useful information often stops at a live feed or a folder of files. As an inventor with claims in this area , U.S. Patent No. 11,414,189 B2 (DeFrank et al.) it includes concepts involving drone reconnaissance, communications, LiDAR terrain analysis, weather awareness, and emergency signaling. My Drone Guardian application explores a software-first commercial product thru creation of a vendor-neutral API kit for people and organizations that already own a compatible drone.
The patent is a product foundation, not a blanket freedom-to-operate opinion. Manufacturer SDK rights, patent ownership and licensing, and third-party technologies will still require appropriate review before commercialization.
What it does
This Build Week project is a “ safe “ browser simulation. A judge can:
open a normalized simulator telemetry sample; see the exact 22% versus 30% configured-reserve calculation; replay synthetic LiDAR-style terrain coverage; acknowledge a battery warning; record a possible terrain observation; generate a clearly labeled mission debrief; and prepare a server-validated JSON simulation-session summary.
The summary recomputes canonical telemetry and safety evidence on the server, identifies client-reported fields, and labels its SHA-256 digest as change detection— this is not a signature or proof of authenticity. The idea is every possible terrain finding requires human review.
How we built it
Codex helped turn my prior patent and product planning into this simulator-first Build Week repository, there was no pre-existing software code, it was all developed from ideas with the result of implementing the end-to-end judge flow, ensure that it would harden safety semantics, add strict route validation, debug rendering and browser sequencing, and prepare the Sites build and submission evidence.
The application uses this stack, React, TypeScript, Vinext, and Cloudflare Workers-compatible routes. A normalized v1 telemetry contract separates the working simulator adapter from planned manufacturer adapters. The optional OpenAI Responses API route is configured for gpt-5.6-sol, bounded structured input, strict JSON-schema output, store: false, a privacy-preserving safety identifier, and a best-effort live-call limit. If a server key is absent or the model is unavailable, the UI explicitly displays Demo fallback and does not claim that GPT produced that result.
Automated evidence includes seven server/route tests plus three Playwright browser tests covering required action order, successful JSON preparation and parsed contents, failed-export messaging, and clearing old decisions when fresh telemetry is loaded.
Challenges
The hardest problem was ensuring the understanding of the simulation , avoiding false certainty. A concern is a polished drone interface can accidentally imply real sensor accuracy, autonomous flight, emergency dispatch, or universal hardware support. We made sure repeatedly that the boundaries are visible in the primary workflow, not just in fine print.
Another challenge was making a contest demo credible without a physical aircraft. The answer was to focus on creation of the necessary claims supported thru an inspectable product loop: simulator adapter → normalized telemetry → explainable safety rule → human decision → possible observation → labeled debrief → bounded session summary.
Accomplishments
These goals were realized. A working, responsive, six-step judge flow. Vendor-neutral normalized telemetry and explainable safety APIs. Explicit warning, equality, rounding, critical, and invalid-threshold semantics. Synthetic terrain mapping, weather, and cloud-cover context. Human-gated observation and simulated emergency actions. Server-validated, explicitly non-audit-grade session summary. Local-only anonymized drone-owner evaluation packet. Desktop and mobile UI verification with no browser console errors.
What we learned
The strongest opportunity is the data gathered and how it can be presented, its not a generic drone controller. It is a decision-support and evidence layer for existing compatible drones. Hikers, ranchers, remote-property owners, guides, inspectors, and search teams need a clear record of what was reviewed, what might need attention, and what remains uncertain. It can provide a safety layer to these and other stakeholders.
We also learned that clear and honest simulation labels improve this product. They let potential users and investors evaluate the workflow without confusing a browser prototype with tested aircraft compatibility or certified sensor performance.
We had one recreational drone owner with one to three years of experience complete all six evidence steps in the synthetic browser simulation on a desktop browser. They rated workflow clarity, practical usefulness, and safety-language clarity 5/5; with selected trail or terrain scouting as the most relevant use case; and expressed interest in a future supervised, read-only drone-data test. The cited highest-priority improvement was a faster visual route replay showing sensor activity and decision points. This is drone owner provided one anonymized usability evaluation of a synthetic simulation—again, not validation of aircraft compatibility, sensor accuracy, real-flight safety, or market demand.
What's next
Next milestones are to build licensed sample flight-log import, read-only MAVLink/PX4/ArduPilot telemetry, supervised hardware telemetry tests, payload-adapter specifications, a supported-hardware matrix, production authentication and durable audit controls, and commercial API plans. Real-aircraft commands remain outside this milestone.
Important. Safety boundary:
No real drone, LiDAR payload, live weather service, emergency service, or outside contact is connected. The project does not launch or control aircraft, identify people, track people, dispatch emergency services, or claim survey-grade measurements.
I am open to comments and critiques, Thanks. !!!
Built With
- cloudflare-workers
- codex
- gpt-5.6
- openai-responses-api
- playwright
- react
- typescript
- vinext