Skip to content

OpenEnv Dispatch Demo

OpenEnv Dispatch Demo is an Uber-style ride-dispatch RL environment built on OpenEnv.

Code Demo Video Cerebral Valley

Core trick

PPO agent learns to call repair actions after injected API and schema drift breaks a mock ride-dispatch workflow.

Tags

agentsproductivity

Description

OpenEnv Dispatch Demo is an Uber-style ride-dispatch RL environment built on OpenEnv. It simulates a workflow (request → match driver → fare → confirm → payment) while exposing drift (schema, API, and policy changes). An RL agent is trained with Stable-Baselines3 (PPO) to repair failures and restore the workflow. The repo includes a FastAPI mock API, Gymnasium env, drift engine, and a demo UI for running episodes with different drift settings.
Anantaverma20/Uber-Simulation-★ 00 forksPythonCreated Mar 8, 2026 · last push Mar 8, 2026

README excerpt

---
sdk: docker
app_port: 7860
---

# Ride-Sharing Dispatch RL Environment (OpenEnv)

Uber-style ride-dispatch reliability environment: workflow (request → match → fare → confirm → payment) with **drift** (schema, API, policy). The RL agent learns to repair failures and restore the workflow. Built with **OpenEnv** (mandatory), Gymnasium, Stable-Baselines3, and FastAPI.

## Quick start

### Local (API + training)

1. **Install**
   ```bash
   pip install -r requirements.txt
   ```

2. **Run mock API** (terminal 1)
   ```bash
  python -m uvicorn backend.api.app:app --host 0.0.0.0 --port 8000
   ```

3. **Train** (terminal 2; API must be running, or use built-in thread)
   ```bash
  python -m backend.train --steps 50000 --drift schema
   ```
   Training starts the API in a background thread by default, so you can run `train` alone. Logs to W&B if `WANDB_API_KEY` is set.

### Demo (episode runner + Next.js)

1. **Run demo API** (from project root)
   ```bash
  python run_demo_api.py
   ```
   Listens on port 8001 and starts the mock API on 8000 in a thread if not using Docker.

2. **Run frontend**
   ```bash
  cd demo/frontend && npm install --legacy-peer-deps && npm run dev
   ```
   (Use `npm run dev:win` if you see NODE_OPTIONS errors on Windows.) Open http://localhost:3000, pick a drift type, and click "Run episode".

   **If you see** `--r= is not allowed in NODE_OPTIONS`:
   - **Dynatrace OneAgent** injects this. Use Docker (below) to run the frontend in an isolated contain

Team