SwiftShelf++ Inventory Control Terminal
SYSTEM ONLINE
Meta PyTorch Hackathon
OpenEnv Compliant
   SWIFTSHELF++  ◆  FEFO DISPATCH — FIRST EXPIRED FIRST OUT  ◆  20 SKU DARK STORE SIMULATION  ◆  3 EVALUATION TASKS (EASY / MEDIUM / HARD)  ◆  PYTORCH LLM AGENT: facebook/opt-125m  ◆  DEEPMIND-LEVEL REWARD SHAPING  ◆  GYMNASIUM.ENV COMPLIANT  ◆  FASTAPI REST SERVER  ◆  DOCKER DEPLOYMENT  ◆  META PYTORCH OPENENV HACKATHON SUBMISSION  ◆ 
Budget
—
Init: $1000.00
Step
—
Max: 200
Total Reward
—
Cumulative
Near Expiry
—
≤ 3 days
Expired
—
Wasted items
Orders
—
In queue
Shelf Inventory — 20 SKUs LOADING…
▶ Interactive Agent Console LIVE API
Last Reward — Episode Total 0.0
SYS > Terminal ready. Press RESET to start a new episode, then choose an action.
Heuristic Grader POST /grade
PASS
Avg reward: 0.00
Reward History 0 pts
Order Queue 0
No orders
Eval Tasks Live Score
Loading…
OpenEnv Compliance Matrix 11 / 11 REQUIREMENTS MET
✅
openenv.yaml (spec_version: 1)3 tasks · easy / medium / hard
✅
models.py — Pydantic typed modelsAction · Observation · State · StepResult
✅
client.py — SwiftShelfClientreset() · step() · state() interface
✅
__init__.py with __all__ exportsClean importable package
✅
inference.py in repo root[START] / [STEP] / [END] stdout format
✅
Task scores in (0.01, 0.99)Strictly within hackathon bounds
✅
FastAPI REST API (port 7860)/health · /reset · /step · /state · /schema · /mcp
✅
Docker deploymentpython:3.11-slim · EXPOSE 7860
✅
pyproject.toml [project.scripts]server = "server.app:main"
✅
Meta PyTorch Integrationfacebook/opt-125m LLM agent via torch
✅
11 PyTest tests — all passingInc. seed reproducibility · FEFO compliance
API Endpoints
API Docs/docs — Swagger Health/health Metadata/metadata Schema/schema Tasks/tasks State/state MCP/mcp