# Olist External Benchmark

This benchmark validates that the existing data-agent pipeline can run on an external public relational dataset using the same task structure as the main benchmark.

## Tasks

| Task ID | Description | Difficulty |
|---|---|---|
| `task_olist_01` | Top 5 customer states by delivered revenue | Medium |
| `task_olist_02` | Late vs on-time delivery review scores | Medium |
| `task_olist_03` | Product categories with highest freight-to-price ratio | Medium |

## Required raw CSV files

Download the Olist Brazilian E-Commerce dataset (https://www.kaggle.com/datasets/olistbr/brazilian-ecommerce) and copy the relevant CSVs into each task's `context/csv/` directory.

### task_olist_01

```text
olist_orders_dataset.csv
olist_customers_dataset.csv
olist_order_items_dataset.csv