Your First Project
Build a small file-based task and check the result yourself.
Before you start
Use the desktop app with a model that supports tools. Prepare an empty folder named csv-summary-practice and a small CSV file containing non-sensitive practice data. Keep your original files elsewhere.
For example, create category and amount columns with Books / 20, Food / 30, and Books / 10. The expected category sums are Books 30 and Food 30. This is an input-check example, not a claim about a generated result.
1. Choose a project folder
Enter Work mode
Choose Work near the app logo. Use Add project folder in the project area, or choose a workspace from the composer’s plus menu.
Select the practice folder
In the system folder picker, select csv-summary-practice. Check the project name in the workspace before sending the task.
Keep the task in that project
Continue in the conversation associated with that folder. A different conversation may use a different directory and permissions.
2. Describe a testable result
Give the task a purpose, an input, an output, and a stopping point. Start with a local page that can be opened in a browser; avoid combining the first exercise with accounts, payments, or deployment.
In this project folder, make a browser-local CSV summary tool. Let me select a CSV and choose the category and amount columns. Group and sum the amounts, show a bar chart, and export a summary CSV. Report rows with invalid amounts. Do not upload the file or modify the original. Explain how to open the result and test it with the practice data.
3. Review the work and permissions
Read the proposed steps and file changes. When a permission request appears, check the operation and its target before allowing it. If the request exceeds the task, refuse it and explain a narrower scope.
If the task is waiting for approval, repeatedly sending the same prompt will not resolve it. Handle the pending request in its conversation. Stopping execution prevents further work after cancellation completes; it does not undo files already changed.
If a file changed outside Kalamos, ask it to read the current file before continuing. Avoid asking it to overwrite a conflict without checking the new contents.
4. Open and test the result
Open the output
Follow the opening instructions returned for the project. If they require a command, understand where it runs and what it does first.
Check a normal input
Choose the practice CSV. Select category and amount. Compare the chart and summary with Books 30 and Food 30.
Try an incorrect amount
In a copy of the CSV, replace an amount with text. Confirm the page explains the invalid row rather than silently presenting a wrong total.
Read back the export
Export the summary, then open the downloaded CSV and check its column names and values.
A completion message is a prompt to inspect the files. The exercise is complete only after the page opens, the known input produces the expected totals, and the exported file contains the same result.
5. Continue with one change
The practice totals are correct. Now make the column selector easier to use on a phone. Keep the CSV parsing and export behavior unchanged.
Keep a usable copy before a larger revision. You can return from the project list or recent conversations. Removing a project from the list does not delete its folder or earlier conversations.
If a project does not run →