🤖 sktime-agentic-forecaster

An LLM agent that reads your prompt and picks its own forecasting model.

Upload a CSV (or use the built-in airline example), describe your data in English, and the agent will select, evaluate, and commit to the best forecaster automatically.

Source: kpal002/sktime-agentic-forecaster

Data

Upload CSV (leave empty for airline example)

Date column name

Target column name

Agent settings

Describe your data (English prompt)

Backend

Metric

Forecast horizon (steps)

1 36

Holdout window

4 36

CV strategy

'expanding' is more reliable but slower

holdout expanding

Forecast

Backends: auto detects from env vars (GEMINI_API_KEY → OPENAI_API_KEY → ANTHROPIC_API_KEY). mock runs offline with no API key — good for testing.