✉ Snowflake Summit 2025 — Say Hello from Summit!

Interactive experience where attendees write and address a physical postcard, then use Snowflake to maps its journey from Moscone Center to anywhere in the USA — powered by Snowflake geospatial functions, Cortex AI, and Streamlit.

---

Project Structure

postcard-summit/
├── init_db.sql        ← Run once in Snowflake to set up DB, table, views
├── cli_app.py         ← CLI app for 6 booth laptops
├── tv_map.py          ← Streamlit real-time map for the TV display
├── requirements.txt
├── .env.example       ← Copy to .env and fill in credentials
└── README.md

---

Setup

1. Credentials

cp .env.example .env
# Edit .env with your Snowflake account/user/password/role/warehouse

Your role must have:

• USAGE on ZIP_CODES_DB.POSTALADMIN (Marketplace dataset)
• ALL on SUMMIT_APP (created by the SQL script)
• USAGE on POSTCARD_WH

2. Install dependencies

python -m venv .venv
source .venv/bin/activate      # Windows: .venv\Scripts\activate
pip install -r requirements.txt

3. Initialize Snowflake

Run init_db.sql in a Snowflake worksheet (or via SnowSQL):

snowsql -f init_db.sql

---

Running the Apps

Booth Laptops (6×) — CLI

python cli_app.py

Each attendee:

1. Enters their destination zip code
2. Sees distance calculated via Snowflake ST_DISTANCE
3. Record inserted into postcard_entries
4. Can ask Cortex AI a question about today's data
5. Type done → next attendee

TV Display — Streamlit Map

streamlit run tv_map.py

• Auto-refreshes every 4 seconds
• Shows KPI metrics: total postcards, total miles, top state
• Arc map from SF to all destinations
• State leaderboard + recent entries sidebar

---

Snowflake Features Highlighted

Feature	Where Used
ST_MAKEPOINT	Build point geometry from lat/lon
ST_DISTANCE	Great-circle distance in metres → miles
ST_MAKELINE	Build LineString flight path
GEOGRAPHY column type	Stores the arc path in postcard_entries
SNOWFLAKE.CORTEX.COMPLETE	AI Q&A grounded in live data
Marketplace data	ZIP_CODES_DB.POSTALADMIN."zcr_usa_zip_centroids"
MVCC (no row-locks)	Safe concurrent INSERTs from 6 laptops

---

Cortex Model

The CLI uses mistral-large2 by default (widely available across regions). To change it, edit the model name in cli_app.py → ask_cortex().

Available alternatives: snowflake-arctic, llama3-70b, gemma-7b

---

Concurrency Note

Snowflake uses MVCC (Multi-Version Concurrency Control) — INSERT statements never block each other. All 6 laptops can write simultaneously without any application-level locking needed.