firefinder

A small, standalone FireFinder+ implementation with the same core logic used in at Geocene, plus lightweight preprocessing helpers.

Install

From PyPI:

pip install firefinder

Local editable install while developing:

pip install -e .

Usage

import pandas as pd
from firefinder import prepare_timeseries, fire_detector_v2, group_events

df = pd.read_csv("metrics.csv")
df = prepare_timeseries(df, correction="false")
df = fire_detector_v2(df)

events = group_events(df)
print(events.head())

If your timestamps are epoch milliseconds:

import pandas as pd
from firefinder import prepare_timeseries, fire_detector_v2, group_events

df = pd.read_json("data.json")
df["timestamp"] = pd.to_datetime(df["timestamp"], unit="ms", utc=True, errors="coerce")
df = prepare_timeseries(df, correction="false")
df = fire_detector_v2(df)
events = group_events(df)
print(events.head())

Testing

pip install -e .[test]
python -m pytest

Required columns

• timestamp
• value
• sensor_type_id (required if correction="true")

Notes

• prepare_timeseries mirrors the cleaning behavior from the original lambda processor.
• correction="true" uses ambient sensor_type_id=9 and stove sensor_type_id=1 when computing ambient-corrected values.
• You can pass a custom sensors mapping into prepare_timeseries to support additional stove sensor IDs.