tabular-examples

What problem does it solve?

This Skill helps you map messy structured/tabular data into high-quality LLM training samples with correct labels, prediction dates, and resolution criteria.

Core Features & Use Cases

• Map rows to Sample fields: Convert CSV/BigQuery/API outputs into Sample() components like question_text, label, prediction_date, and resolution metadata.
• Compute labels from outcomes: Define outcomes (e.g., shock vs no shock) from time-series or derived columns, while avoiding leakage.
• Generate questions and add real-world context: Use TemplateQuestionGenerator for consistent question text and optionally NewsContextGenerator plus a renderer to enrich prompts.
• Production-oriented walkthrough: Includes a supply-chain shock detection example that you can adapt to other tabular forecasting setups (including time splits for train/test).

Quick Start

Ask the AI to adapt the supply chain shock detection pipeline by mapping your tabular fields into create_sample(), generating questions with TemplateQuestionGenerator, and (optionally) enriching prompts with NewsContextGenerator for your forecasting dataset.