skforecast-ai
OfficialServer Configuration
Describes the environment variables required to run the server.
| Name | Required | Description | Default |
|---|---|---|---|
No arguments | |||
Instructions
Guidance the server publishes about itself, which clients place ahead of the tool catalog so the model reads it before choosing anything.
This server publishes no instructions, or was last inspected before Glama recorded them.
Capabilities
Features and capabilities supported by this server
Protocol revision2025-11-25
| Capability | Details |
|---|---|
| tools | {
"listChanged": false
} |
| prompts | {
"listChanged": false
} |
| resources | {
"subscribe": false,
"listChanged": false
} |
Tools
Functions exposed to the LLM to take actions
| Name | Description |
|---|---|
| profileA | Profile a CSV file: frequency, series, exogenous columns and the recommended forecaster and estimator. Returns a profile id. |
| planA | Build a forecasting plan (lags, window features, metric, preprocessing, interval) from a profile and a horizon. Returns a plan id. |
| refine_planA | Build a new plan from an existing one, changing some of its decisions. Returns a new plan id; the original plan is kept. |
| create_cvB | Build a time series cross-validation strategy (TimeSeriesFold) for a plan. The response states its cost in |
| backtestA | Backtest a plan with a cross-validation strategy: metrics over the folds, with the predictions and metrics in CSV files. Returns a backtest id. |
| compareA | Backtest several configurations on the same folds and rank them by |
| forecastA | Run a plan: forecast the future, or evaluate on a test set with |
| get_codeA | Return the Python script of an object: the one that ran, or that would run for a plan. |
| get_failureB | Return the traceback and the code of a failed run or of a failed candidate of a comparison. |
| list_objectsC | List the ids of the objects the server keeps. |
| describe_objectB | Return the response that created an object (summary, links, notices, files). |
Prompts
Interactive templates invoked by user choice
| Name | Description |
|---|---|
No prompts | |
Resources
Contextual data attached and managed by the client
| Name | Description |
|---|---|
No resources | |
TDQS
Scored across 11 tools
Most tools have clearly distinct purposes: profile builds a profile, plan builds a plan, create_cv builds a CV strategy, and forecast/backtest/compare each run in distinct modes. The main risk is soft overlap between forecast, backtest, and compare (all execute a plan), but their outputs and semantics differ enough to distinguish them.
Names are readable and mostly follow a verb_noun pattern (refine_plan, create_cv, get_code, get_failure, list_objects, describe_object), but several are bare verbs (forecast, backtest, compare, profile, plan), which is a minor deviation. No chaotic mixing of case or conventions.
11 tools is well-scoped for a forecasting workflow server. Each tool maps to a distinct stage (profile, plan, CV, run, compare, inspect) and none feels redundant.
The lifecycle is largely covered: profile -> plan -> create_cv -> forecast/backtest/compare, plus inspection via list_objects, describe_object, get_code, and get_failure, and refinement via refine_plan. Minor gaps exist (no explicit cancel or delete tool despite mentions of cancellation and object storage), but agents can work around these.