02 — The demo CSV

Duration: 10 min Prerequisites: a text editor or VS Code.

Where the data lives

The repo ships with a ready-to-use CSV at:

csv-llm-ollama/
└── data/
    └── data1-anonymized.csv     ← 734 synthetic transactions

You can open it in any editor. It’s a plain ;-separated file with quoted strings.

File format

Header (first line):

"Date";"Card Number";"Description";"Category";"Debit";"Credit"

A few rows:

"2027-12-27";"************4382";"Late Payment Fee";"Fees";"24.31";"0"
"2027-12-26";"************4382";"Postbox Rental";"Housing";"489.53";"0"
"2027-12-25";"************4382";"City Cab Network";"Transport";"73.34";"0"
"2027-12-24";"************4382";"Travel Booking Refund";"Refunds";"0";"69.15"
"2027-12-24";"************4382";"Harbor Noodle House";"Dining";"59.22";"0"
"2027-12-23";"************4382";"Credit Card Payment Thank You";"Payments";"0";"850.65"

Six columns:

Column	Type	Notes
`Date`	`YYYY-MM-DD`	All transactions in 2027
`Card Number`	masked	Always `************4382`
`Description`	string	The merchant name
`Category`	string	One of 13 English categories
`Debit`	string number	Money out; `"0"` if not applicable
`Credit`	string number	Money in (refunds, card payments)

The numbers

Metric	Value
Rows (excl. header)	734
Expense rows (`Debit > 0`)	675
Credit rows (`Credit > 0`)	59
Total debit	$118,667.69
Total credit	$43,707.67
Date range	`2027-01-02` → `2027-12-27`
Distinct months	12

The 13 categories

Category	Example merchants
`Business`	DesignMarket License, PrintWorks Studio, Invoice Client Deposit
`Dining`	Harbor Noodle House, Coastal Spoon Cafe, Maple Table Bistro
`Education`	Online course bundles, certification fees
`Entertainment`	PuzzleRoom Adventure, Riverfront Music Hall, GameCorner Store
`Fees`	Late Payment Fee, foreign exchange fees
`Groceries`	Wellington Pantry Co, Canterbury Family Foods, Rotorua Organic Store
`Health`	CarePlus Lab Services, ActiveLife Physiotherapy
`Housing`	Rent Payment Willow Street, Postbox Rental, StorageBox Monthly
`Payments`	Credit Card Payment Thank You (statement-level entries)
`Refunds`	Travel Booking Refund and other returns
`Transport`	City Cab Network, NorthLine Bus Card, GreenRide Shuttle
`Travel`	Queenstown Path Pass, Aotearoa Lodge Booking, Pacific Coast Hostel
`Utilities`	ClearNet Fibre, Harbor Internet Co

Top categories by spend

A quick pandas look (you’ll do this in chapter 07):

import pandas as pd
df = pd.read_csv("data/data1-anonymized.csv", sep=";", quotechar='"')
df["Debit"] = pd.to_numeric(df["Debit"], errors="coerce").fillna(0)
print(df.groupby("Category")["Debit"].sum().sort_values(ascending=False).head(5))

Expected output:

Category
Housing      30895.29
Travel       16597.44
Business     14386.39
Groceries     9725.93
Education     8987.85
Name: Debit, dtype: float64

A few useful merchants to memorize

These will show up in the Chat demos in chapter 09:

Merchant	Visits	Total spent
Coastal Spoon Cafe	19	$1,221.34
Harbor Noodle House	18	$1,095.60
NorthLine Bus Card	18	$795.76
Rent Payment Willow Street	8	$6,761.44
Queenstown Path Pass	8	$4,315.40

We’ll use Coastal Spoon Cafe as the recurring merchant for canonical question Q3.

Loading the file in Python

A simple read (we’ll do better in chapter 04 with csv-llm-shared/ingest.py):

from pathlib import Path
import csv

path = Path("csv-llm-ollama/data/data1-anonymized.csv")
with path.open(encoding="utf-8-sig") as f:
    reader = csv.DictReader(f, delimiter=";", quotechar='"')
    rows = list(reader)

print(len(rows))            # 734
print(rows[0])
# {'Date': '2027-12-27', 'Card Number': '************4382', ...}

Takeaways

The repo ships with csv-llm-ollama/data/data1-anonymized.csv — 734 synthetic transactions.
6 columns, ; separator, UTF-8 BOM tolerant.
12 months of 2027, 13 English categories, 1 masked card.
Top category is Housing ($30,895.29). Most-frequent merchant is Coastal Spoon Cafe (19 visits).

Next: Setting up the environment →