> ## Documentation Index
> Fetch the complete documentation index at: https://lab.pollack.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# Lesson 3: Read and Transform

> Read files, perform data transformation, and write structured output

<Card title="Source Code" icon="github" href="https://github.com/markpollack/agent-client-tutorial/tree/main/02-read-and-transform">
  `agent-client-tutorial/02-read-and-transform`
</Card>

## What You'll Build

A Spring Boot application that asks an agent to read log files, count severity levels (ERROR, WARNING, INFO), and produce a summary CSV. This demonstrates agents working with existing files and producing structured output.

## Step 1: Set Up Sample Data

Clone the tutorial repository and navigate to the lesson:

```bash theme={null}
git clone https://github.com/markpollack/agent-client-tutorial.git
cd agent-client-tutorial/02-read-and-transform
```

The application creates sample log files automatically if they don't exist. Three service logs (`auth.log`, `api.log`, `worker.log`) with mixed severity levels.

## Step 2: Understand the Goal

The goal is a multi-line prompt that specifies the input, the transformation, and the expected output format:

```java theme={null}
String goal = """
    In the logs/ directory there are multiple .log files from different services.
    Scan all .log files and count how many lines contain "ERROR", "WARNING", and "INFO".

    Output your results to summary.csv with the following structure:
    severity,count
    ERROR,<total ERROR lines>
    WARNING,<total WARNING lines>
    INFO,<total INFO lines>

    The output should be a valid CSV file with exactly four lines.""";
```

Notice the specificity: file format, column names, exact line count. Agents perform better with precise output specifications.

## Step 3: Run It

```bash theme={null}
# With Claude (default)
./mvnw spring-boot:run

# Or with a different provider
./mvnw spring-boot:run -Dspring.profiles.active=codex
```

## Step 4: Verify

Check the output CSV:

```bash theme={null}
cat summary.csv
# severity,count
# ERROR,4
# WARNING,4
# INFO,7
```

## What's Different from Lesson 1

* **Agent reads existing files** — not just creating from scratch
* **Multi-file input** — the agent scans a directory of logs
* **Structured output** — the result is a well-defined CSV, not free text
* **Data transformation** — counting, aggregating, formatting

This pattern — read existing state, transform it, write structured output — is the core of most agent tasks in production.

## Next

[Lesson 4: Git Operations →](/docs/agent-client/tutorial/04-git-operations) — Use an agent to inspect git history and recover lost changes.