v0.0.28

Features

• T0 foundation model: Added support for T0, an open-weights time series foundation model from The Forecasting Company. Use it via the T0 class. See #348.

  import pandas as pd
  from timecopilot.models.foundation.t0 import T0
  
  df = pd.read_csv(
      "https://timecopilot.s3.amazonaws.com/public/data/air_passengers.csv",
      parse_dates=["ds"],
  )
  
  model = T0()
  fcst_df = model.forecast(df, h=12)

• timecopilot-tirex bump: Bumped timecopilot-tirex to >=0.1.1 and updated the TiRex integration to align with the new API. TiRex now returns the model's default quantile knots (0.1–0.9) and validates that custom quantile requests match that set. See #346.

Fixes

• Relax dependencies: Relaxed several pinned core dependencies (lightning, logfire, opentelemetry-api, opentelemetry-sdk) to minimum-version constraints, moved ray to the distributed optional extra, and removed wandb from core dependencies to reduce install conflicts. See #355.

New Contributors

• @GeoffNN made their first contribution in #348

---

Full Changelog: v0.0.27...v0.0.28