A flying car dropped at a random height and attitude, landed by Laya (Ollaya's System One API). Laya reads plain-language flight reports, and its answers set which of the 8 thrusters fire and how hard.
-
Create
.envnext toserver.py:OLLAYA_URL="http://<ollaya-host>:11435" OLLAYA_API_KEY="<your key>" OLLAYA_MODEL="laya:latest"
-
Start the server (Python 3.9+, no installs):
python3 server.py
-
Open http://localhost:8000/car. Use
http://localhost:8000/car?squarefor a square frame to record.
More pilots: every <NAME>_URL + <NAME>_API_KEY (+ <NAME>_MODEL) in .env is a pilot, e.g. JEV_URL="https://api.typesafe.ai", JEV_API_KEY, JEV_MODEL="jev-latest". All of them fly the identical drop at once, overlapping on one pad; pick some with /car?pilots=ollaya,jev.
The page loads three.js from a CDN, so the browser needs internet access.
npm install
npm run flight-test # 40 headless drops replaying recorded Laya answers
MODE=laya npm run flight-test # real Laya, real time
MODE=laya PILOT=jev npm run flight-test # any pilot from .env