foundationforecast

Foundation time series forecasting models, extracted from TimeCopilot.

Run state-of-the-art pretrained models (Chronos, Moirai, TimesFM, Toto, TiRex, TimeGPT, and more) through a single unified API.

Installation

pip install foundationforecast

Requires Python 3.10+. Some models have additional version requirements — see the Model Hub.

Optional plotting support:

pip install "foundationforecast[plot]"

Quick example

import pandas as pd
from foundationforecast import FoundationForecast
from foundationforecast.models import Chronos, Toto

df = pd.read_csv(
    "https://timecopilot.s3.amazonaws.com/public/data/air_passengers.csv",
    parse_dates=["ds"],
)

ff = FoundationForecast(models=[Chronos(), Toto(context_length=256)])
fcst = ff.forecast(df, h=12, freq="MS")
cv = ff.cross_validation(df, h=12, freq="MS")

Supported models

Chronos, FlowState, Moirai, PatchTST-FM, Sundial, T0, TabPFN, TiRex, TimeGPT, TimesFM, Toto

Documentation

Build and serve docs locally:

uv sync --group docs
uv run --group docs mkdocs serve

See Getting Started and Examples.

Development

uv sync --group dev --group docs
pre-commit install --install-hooks
uv run pytest

License

MIT