Skip to content

[Feature] Add built-in sktime tool for time series forecasting workflows #211

Description

@Saloni-0465

Summary

I'd like to add a built-in sktime tool to ml-intern that lets the agent
run time series ML workflows directly, without requiring an external MCP server.

Motivation

sktime is a widely-used Python framework for time series tasks (forecasting,
classification, anomaly detection). Right now, if a user asks ml-intern to
"forecast this time series using ARIMA" or "compare forecasting models on
airline data", the agent has no direct way to do this.

A built-in sktime tool would let the agent:

  • Discover available estimators by task type
  • Get detailed parameter info for any estimator
  • Run end-to-end forecasting on built-in datasets
  • Evaluate and compare model accuracy

Proposed Design

Following the same pattern as papers_tool.py, I'd add:

  • agent/tools/sktime_tool.py — tool spec + handler with operations:
    • list_estimators — browse estimators filtered by task/tags
    • describe_estimator — get parameters and capabilities
    • forecast — fit and predict on built-in datasets
    • list_datasets — show available datasets

Why I'm the right person for this

I'm an active contributor to sktime-mcp (9 PRs) and have deep familiarity
with sktime's registry and API. I have a working implementation ready.

Happy to open a PR if this direction looks good to maintainers!

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions