timegpt plugins

by Jeremy Longshore

Real Apache Airflow tooling for TimeGPT forecasting pipelines. Generate production-ready DAGs (BigQuery / Snowflake / Postgres / S3 sources), AST + DagBag-based DAG validation with credential lint, real Airflow Connection JSON for 5 source types, and pytest test modules that load the DAG via DagBag.

by Jeremy Longshore

[PROOF OF CONCEPT] Anomaly-detection API surface for streaming protocol metrics (Kafka/Kinesis/webhooks). All tools return illustrative fixtures — not live stream data. Production requires real consumers + Nixtla detect_anomalies() + alert delivery (~6-10 wks).

by Jeremy Longshore

Run Nixtla statsforecast / TimeGPT forecasting on BigQuery time-series data with SQL-injection-validated identifiers and exponential-backoff retry on transient GCP errors. Pull from BigQuery, forecast, write results back.

by Jeremy Longshore

Real cost optimization for Nixtla TimeGPT customers. Analyzes API call patterns, detects batching/caching/single-series/short-horizon opportunities, generates ranked recommendations with USD savings estimates, simulates batching impact, produces hybrid TimeGPT+statsforecast routing strategies as code, and exports markdown/json/csv reports.

by Jeremy Longshore

dbt package with macros for Nixtla TimeGPT forecasting + anomaly detection across data warehouses. BigQuery canonical (BQML ML.FORECAST); Snowflake / Databricks adapters ship as honest PoCs (depend on Nixtla Native App or external UDF registration).

by Jeremy Longshore

[PROOF OF CONCEPT] Anomaly-detection API surface for DeFi protocol monitoring (TVL/APY/volume). Built as exploration after a real DeFi exploit. All tools return illustrative fixtures — not live protocol data. Production requires DeFiLlama integration + Nixtla detect_anomalies() + alert webhook delivery (~6-8 wks).

by Jeremy Longshore

[WORK IN PROGRESS] Embeddable JS widget that renders a TimeGPT forecast inline in any web page. Not production-ready — every tool response includes a `_disclaimer` field. Tracked for v1.0 build (Epic 5.3).

by Jeremy Longshore

[WORK IN PROGRESS] Generate an audit-grade report from a TimeGPT forecast for compliance + governance review. Not production-ready — every tool response includes a `_disclaimer` field. Tracked for v1.0 build (Epic 4.3).

by Jeremy Longshore

Post-hoc explainability for TimeGPT forecasts. STL decomposition (statsmodels), Pearson correlation driver analysis with lag scanning (scipy), executive narrative generation, and markdown report composition.

by Jeremy Longshore

[WORK IN PROGRESS] Curated TimeGPT forecasting workflow templates for retail, finance, energy, ops. Not production-ready — every tool response includes a `_disclaimer` field. Tracked for v1.0 build (Epic 4.2).

by Jeremy Longshore

Real migration assistant for moving from Prophet, statsmodels, or sklearn forecasting to Nixtla TimeGPT or StatsForecast. AST-based code analysis, pandas-based data transformation to canonical (unique_id, ds, y) schema, code generation across 5 source/target combinations, real side-by-side accuracy comparison, and phased migration plans.

by Jeremy Longshore

Calculate 3-year ROI of moving forecasting to Nixtla TimeGPT, generate a real PDF report, compare scenarios with sensitivity analysis, and post the result to Salesforce as an Opportunity (live or dry-run).

by Jeremy Longshore

[WORK IN PROGRESS] Generate enterprise-ready Nixtla TimeGPT sales demos from a customer's data shape. Not production-ready — every tool response includes a `_disclaimer` field. Tracked for v1.0 build (Epic 4.1).

by Jeremy Longshore

Automated content discovery for Nixtla / time-series practitioners. Searches the web + GitHub for forecasting content, generates AI summaries, and publishes formatted digests to Slack on a schedule.

by Jeremy Longshore

Head-to-head benchmark of TimeGPT vs open-source StatsForecast on user data. Real model fits (AutoETS, AutoTheta, SeasonalNaive + optional NixtlaClient), real metrics (sMAPE, MASE, MAE, RMSE), wall-time comparison, rule-based recommendations.