ridegenie 0.1.0

One-line installer for the India RideGenie Databricks demo (data + Genie + AI/BI dashboard + app).

🛺 ridegenie — India RideGenie demo, installed in one line

A dbdemos-style installer for the India RideGenie demo: natural-language geospatial analytics on synthetic Indian ride-hailing data — 200K trips · 160+ cities · ₹ fares · UPI payments.

One call provisions the whole thing into your Databricks workspace:

#	Asset	What it does
1	Unity Catalog data layer	rides_enriched, ride_zones, pickup_hotspots (~200K rides, PIN-code geo)
2	Genie Space	Ask in plain English — Genie writes governed SQL (₹, cities, surge, payments)
3	AI/BI Dashboard	INR KPIs, demand-by-hour, surge, vehicle/payment mix + a live pickup-hotspot map
4	RideGenie App	FastAPI + Leaflet — draw a region on an India map and ask Genie about it

---

Quick start (Databricks notebook)

%pip install git+https://github.com/anuj1303/ridegenie-demo.git
dbutils.library.restartPython()

import ridegenie
ridegenie.install("india-ridegenie")

That's it. The installer prints links to your Genie Space, Dashboard, and App when it finishes.

Prerequisite: a running SQL Warehouse (the Serverless Starter Warehouse is enough). Works on Databricks Free Edition and any AWS/Azure/GCP workspace.

Prefer not to write code? Import notebooks/00-START-HERE-RideGenie.py into your workspace and Run All.

---

What install() does (à la dbdemos)

1. Resolves context — current user, workspace host, and a SQL warehouse (auto-picks a running one, or starts/creates one).
2. Builds the data layer — creates the schema + volume, uploads the bundled (gzipped) CSVs, and runs a templated setup.sql to build the tables/views.
3. Creates the Genie Space — registers the tables with format-assistance + entity-matching and applies domain instructions (INR, surge, payments).
4. Creates & publishes the AI/BI dashboard — one shared dataset drives cross-filtering across KPIs, charts, and the hotspot map.
5. Deploys the app — creates the app, uploads the source, attaches sql-warehouse + genie resources (IDs auto-injected), grants the app's service principal data access, and deploys.

Everything is workspace-portable — table names, warehouse, Genie/app resource IDs are resolved at install time via {{CATALOG}} / {{SCHEMA}} / {{VOLUME}} templating. No hard-coded IDs.

---

Options

ridegenie.install(
    "india-ridegenie",
    catalog="workspace",              # target UC catalog
    schema="india_rides_workshop",    # target UC schema
    warehouse_id=None,                # default: first running/available warehouse
    install_app=True,                 # set False to skip the FastAPI app
    profile=None,                     # only when running locally (Databricks CLI profile)
    overwrite=False,
)

Run locally (outside a notebook) against a CLI profile:

import ridegenie
ridegenie.install("india-ridegenie", profile="FreeEdition")

---

Package layout

ridegenie/
├── conf/india-ridegenie.json     # demo manifest (the "bundle_config")
├── installer.py                  # the Installer orchestrator + public install()
├── sql/setup.sql                 # templated data-layer DDL
├── genie/instructions.md         # Genie domain instructions
├── dashboards/india_ridegenie.lvdash.json   # AI/BI dashboard definition
├── data/*.csv.gz                 # pre-built synthetic data (uploaded to a Volume)
└── app/                          # FastAPI + Leaflet app source
notebooks/00-START-HERE-RideGenie.py   # one-click bootstrap notebook
scripts/                          # build_zones.py / generate_rides.py (regenerate the data)

Regenerating the data (optional)

The repo ships pre-built data. To rebuild it from source (GeoNames India PIN codes):

cd scripts
python3 build_zones.py       # downloads GeoNames, builds ride_zones.csv
python3 generate_rides.py    # builds rides.csv (200K trips)
gzip -c ride_zones.csv > ../ridegenie/data/ride_zones.csv.gz
gzip -c rides.csv      > ../ridegenie/data/rides.csv.gz

---

Built for Databricks Field Engineering. Synthetic data — no PII. PIN codes from GeoNames (India).