SKU RLE-1205 · Sold by External

AppWorld

Product specifications

SKURLE-1205
Data typeRL environments
Volume750 tasks
Size on diskNot published by source
FormatPython package + SQLite DBs + JSON task specs
Access modelPUBLIC LICENSE
PricingFree · open-source license
Quality score
LicenseApache-2.0
AppWorld is a high-fidelity execution environment of 9 simulated day-to-day apps (Amazon, Spotify, Venmo, Gmail, etc.) exposing 457 documented REST-style APIs over 100+ SQLite database tables populated with ~100 fictitious users. Agents solve tasks by writing and running interactive Python code in a Jupyter-like shell; solutions are graded by state-based unit tests. The benchmark contains 750 tasks (250 scenarios × 3 variants), split into Train 105 / Dev 60 / Test-Normal 168 / Test-Challenge 417. Installed via 'pip install appworld' plus 'appworld download data'.