Show HN: GribStream – Query Weather Forecasts Like a Database
I shared GribStream here a while ago when it was just a NOAA forecast API. It’s grown into something bigger: a full weather data archive you can query just like a database, by time (with as-of/time-travel!), location, or more interestingly by specific weather conditions.
Use built-in expressions directly in your API requests to calculate derived metrics like wind chill, dew point, or more complex like crop-specific photothermal units capture (see demos below). Computation is fully server-side, ideal for web apps, dashboards, or mobile apps with limited resources.
Demos ( https://gribstream.com/demo ): NBM Snow Accumulation: Real-time ski conditions across North America HRRR Corn Growth Simulator: Predict crop development stages precisely GFS Storm Chaser: Track storms globally using dynamic filtering NBM Wind Field Explorer: Explore wind patterns interactively
Other Improvements: Added models: NBM, GFS, HRRR, RAP, GEFS, CFS, and Google's GraphCast GFS Bounding box queries at custom resolution. Perfect for interactive maps Significant price reduction: over 50% cheaper, plus 90% discount for cached requests Efficient bulk queries (up to 500 points per request) at same quota cost Comprehensive Quickstart guide and OpenAPI spec for easier onboarding
Sneak peek at what I'm currently doing: Realtime streaming of grib files as video. This is raw, not productionized at all, just an alpha-toy but it is cool so I'd like to share a few examples. This does no caching at all, it is generated on the fly. So you can play with the times, try other weather variables, tweak the scaling and framerate. I'll be monitoring networking in case I need to shut it down in a hurry, please be patient. Demo for Hurricane Milton
Wind speed at 80m
https://gribstream.com/video?fromTime=2024-10-08T00:00:00Z&UntilTime=2024-10-12T00:00:00Z&name=WIND&level=80%20m%20above%20ground&info=&scaleMin=0&scaleMax=40&fps=6
Convective available potential energy at the surface
https://gribstream.com/video?fromTime=2024-10-08T00:00:00Z&UntilTime=2024-10-12T00:00:00Z&name=CAPE&level=surface&info=&scaleMin=0&scaleMax=3000&fps=6
# wave height milton
https://gribstream.com/video?fromTime=2024-10-08T00:00:00Z&UntilTime=2024-10-12T00:00:00Z&name=HTSGW&level=surface&info=&scaleMin=0&scaleMax=7&fps=6
Side-project hunting? If you're looking for your next indie-hack, here are a few ideas that GribStream makes ridiculously easy: Agriculture: Crop growth modeling and irrigation planning Renewable Energy: Forecasting for solar and wind energy production Logistics: Weather-informed routing and delivery scheduling Insurance: Risk modeling based on historical weather patterns Event Planning: Scheduling and resource allocation for outdoor events
Hopefully DOGE will let NOAA keep running smoothly so I can ship faster!
Would appreciate your feedback, feature requests, or any ideas on what you'd love to see next.
Thank you!
The embedded expressions and derived metrics are really powerful—especially for domains like agriculture or renewable energy. The real-time GRIB video streaming demo is wild too. Are you thinking of exposing that as an API in the future?
Also, are you planning any kind of data quality control or validation layer on top of the raw model outputs, or is the assumption that users are familiar with the limitations of each model?
Amazing work—I'll definitely be following along with future updates!
Herbie support for the GribStream API might be worthwhile; https://news.ycombinator.com/item?id=42470493 :
> Are there error and cost benchmarks for these predictive models?
ENH: TODO list / PIM integration potential: tasks notifications for specifically labeled tasks for when it's going to rain, going to be dry for days; when to plant, mow, work on construction projects in various states.
todo.txt labels: +goingtorain +goingtobedry
FarmOS is an open source agtech product that could have Historical and Predictive weather built-in.
"Show HN: We open-sourced our compost monitoring tech" https://news.ycombinator.com/item?id=42201207 ; compost and mulch businesses can save money by working with the rain
https://news.ycombinator.com/item?id=42201251 ; "crop monitoring system" site:github.com , digital agriculture, precision agriculture, SIEM
"Satellite images of plants' fluorescence can predict crop yields" https://news.ycombinator.com/item?id=40234890
Having been inundated with ads on weather apps that must afford notifications somehow, I've tried a few weather apps; TWC, AccuWeather; Wx, WeatherMaster, WeatherRadar
Wx is too complex for the average bear. (Where is animated local radar, for example?)
WeatherMaster is nice and open source, but isn't on FDroid and doesn't have animated radar
WeatherRadar is open source, on FDroid, has a chart that shows {Precipitation and Hi and Lo} all on the one chart, and already requires users to get and configure a uniquely-identifying third-party API token.