Async Python client for the Finnish Meteorological Institute's open data.
It fetches FMI's official point forecast, the one shown on fmi.fi and in the FMI weather app, as a typed time series. The forecast covers Finland and nearby areas and reaches about ten days ahead.
- Built on
aiohttp, its only dependency. You pass in theClientSession. - Fully typed, ships
py.typed. - Requires Python 3.14.
pip install aiofmiimport asyncio
import aiohttp
from aiofmi import FmiClient, daily_forecasts
async def main() -> None:
async with aiohttp.ClientSession() as session:
client = FmiClient(session)
forecast = await client.forecast(60.1699, 24.9384)
print(forecast.location.name, forecast.location.timezone)
for point in forecast.points[:3]:
print(point.time, point.temperature, point.wind_speed, point.smart_symbol)
for day in daily_forecasts(forecast):
print(day.date, day.temperature_min, day.temperature_max, day.precipitation)
asyncio.run(main())The client never creates or closes a session.
Returns a PointForecast:
| Attribute | Contents |
|---|---|
location |
name, region, country, timezone (IANA name), latitude, longitude. The name is that of the nearest known place and is not a stable identifier. |
issued |
Analysis time of the forecast. |
points |
ForecastPoints in chronological order. |
By default the series starts at the beginning of the current hour and runs
hourly to the end of the forecast. start and end must be timezone-aware.
A ForecastPoint has a timezone-aware UTC time and these values, each
None when FMI does not provide it:
| Attribute | Unit |
|---|---|
temperature, feels_like, dew_point |
°C |
humidity, cloud_cover |
% |
pressure |
hPa |
wind_speed, wind_gust |
m/s |
wind_direction |
degrees |
precipitation |
mm during the preceding hour |
precipitation_probability, thunder_probability |
% |
smart_symbol |
FMI SmartSymbol code, with 100 added at night |
point.symbol is the code without the night offset and point.is_daytime
tells which half it came from.
Groups the points by calendar day and returns DailyForecasts with
temperature_max, temperature_min, precipitation (sum),
precipitation_probability (highest), wind_speed (highest),
wind_direction (at the highest wind speed), wind_gust (highest),
smart_symbol (at 15:00, as on FMI's own services; symbol is the code
without the night offset) and the day's points.
Days are those of tz, by default the time zone of the forecast location.
Looking that up reads the system's time zone database, so pass tz when
calling from an event loop.
The first day covers only the hours the series contains, and its symbol is the one closest to 15:00. The last day is left out when the series ends before 15:00 on it, so that every day is described by the same hour. Each precipitation value covers the preceding hour and counts towards the day in which that hour began.
All derive from FmiError.
| Exception | Meaning |
|---|---|
FmiConnectionError |
The service could not be reached. |
FmiTimeoutError |
No answer within request_timeout. Subclass of FmiConnectionError. |
FmiRateLimitError |
HTTP 429. retry_after holds the wait in seconds, if given. |
FmiResponseError |
An unexpected HTTP status (status) or an unreadable response. |
FmiRequestError |
FMI rejected the request. code holds the OWS exception code. |
FmiNoDataError |
No forecast for the location (it is outside the forecast area) or for the time range. Subclass of FmiRequestError. |
Invalid arguments raise ValueError before any request is made.
FMI's open data is licensed under CC BY 4.0; credit the Finnish Meteorological Institute when you publish it. The service allows 20 000 requests a day and 600 per five minutes per user. The forecast is updated a few times a day, so there is little to gain from polling more often than every half hour.
uv sync
uv run pytest
uv run ruff check . && uv run ruff format --check .
uv run mypyTests run against responses recorded from the live service, in
tests/fixtures.