Introduction

Therminal provides programmatic access to Kalshi temperature prediction market data and NWS weather observations through a unified REST API.

Base URL

https://api.mostlyright.xyz

All endpoints return JSON by default. Add ?format=csv or ?format=parquet for bulk downloads. All timestamps are UTC unless a tz parameter is specified.

Response limits

Format	Max rows	Notes
JSON	50,000	Paginated via ?limit= and ?offset= (default 1000, max 10000 per page)
CSV	500,000	Full result set, no pagination
Parquet	500,000	Full result set, no pagination

For 1-minute ASOS data, 500K rows covers ~347 days for a single station. Use ?format=parquet for bulk ML/analytics workloads.

Key concepts

Concept	Description
Series	A temperature metric for a location (e.g., KXHIGHNY = NYC daily high)
Event	A specific date’s prediction (e.g., KXHIGHNY-26MAR20 = NYC high on March 20)
Market	A strike within an event (e.g., KXHIGHNY-26MAR20-T50 = “Will NYC high exceed 50°F?”)
Candle	OHLCV price data at 1-minute intervals. Prices are in cents (0–100).
Station	A 3–4 letter NWS weather station code (e.g., NYC, LAX, ORD)

Data freshness

Dataset	Latency	Source
Candles (recent 14 days)	~5 minutes	Kalshi API → Supabase
Candles (historical)	~1 hour	R2 parquets via DuckDB
Observations (recent 14 days)	~5 seconds	AWC → Supabase
Observations (historical)	~1 hour	R2 parquets via DuckDB
Observations (1-minute ASOS)	Historical only	NCEI DSI-6406 (2000–2026)
Climate reports	~24 hours	IEM → Supabase

Python SDK

pip install therminal-py

The Python SDK (v1.0.7) provides domain-split clients, a config file, and WeatherLive for real-time METAR observations from AWC with schema-identical output to historical data.

from therminal.weather import WeatherHistory, WeatherLive

weather = WeatherHistory()                             # Historical (via API)
live = WeatherLive()                                   # Real-time (via AWC)
current = live.current(["NYC", "ATL"])                 # Same Observation type

Full documentation

• API Reference — All endpoints with interactive playground
• Python SDK — Domain clients, WeatherLive, config, ML features
• Data Sources — Where the data comes from
• Rate Limits & Errors — Error codes and rate limiting