Project Files

A Haute project keeps its files in two kinds of place. The pipeline folder is versioned with the pipeline. The local folders at the project root stay on your machine.

Where files go

Folder	What goes in it	Versioned
rating/	The pipeline: main.py, its layout file, node settings in config/, shared Python in utility/	Yes
rating/files/data/	Tables and data the pipeline ships with, such as a postcode lookup	Yes
rating/files/models/	Model files with their feature contracts, and optimiser results	Yes
rating/files/params/	Parameter files the pipeline reads; a Workbench column's allowed values can come from one	Yes
rating/workbench/	The workbench's schema, sheets and sample, a file for each table and each sheet (Workbench)	Yes
data/	Local data: portfolios, extracts, anything large or personal	No
outputs/	What runs write: Data Output files and exported data	No

rating/ is the pipeline folder of a new project. A project whose haute.toml names another pipeline file uses that file's folder.

Everything in the pipeline folder goes into version history. Save records the pipeline's own files, and Commit also records every other file you add to the pipeline folder, so a model or table in rating/files/ reaches CI and deploys with the pipeline. A file larger than 100 MiB can't be versioned (GitHub refuses it), so Commit stops and names it. Keep a file that size in data/, or log the model to MLflow.

Data the quote output reads must be in the pipeline folder: a deploy refuses a Data Input on the way to the quote output that reads from data/. Anything else, such as model training or optimisation, can read from data/: a deploy drops every node the quote output doesn't depend on (Deployment).

data/ and outputs/ stay on your machine. Moving to another version, catching up or branching away leaves both as they are, so outputs/ can hold what another version's runs wrote.

Save model to file and an optimiser result's Save to file put a filename without a folder in rating/files/models/.

The Files buttons

Two buttons in the toolbar open the project's files beside the canvas or the workbench:

• Pipeline files opens rating/files/. A file you add here is saved into version history straight away, and a file you delete is recorded the same way.
• Local files opens data/ and outputs/. Files here stay on your machine. You can add and delete files in data/; outputs/ only downloads, because a run writes it.

The panel shows one folder at a time. Click a folder to open it, and click a name in the path under the title to go back. Each file shows its size and when it last changed, with Download and, where files can be added, Delete on the right of its row.

To add files, drag them from your computer onto the panel: onto the list to add them to the folder that's open, or onto a folder's row to add them there. Upload at the top of the panel picks files instead. Dropped folders aren't added; drop the files inside them. If a file with the same name is already there, Haute asks before replacing it. A replaced or deleted pipeline file stays in version history unless it was New or Changed; a local one can't be recovered. The question says which.

In Pipeline files a file marked New isn't in version history yet, such as one copied in with your file explorer; the next Commit adds it. One marked Changed has been edited since it went into version history; the next Commit records the change. One marked Ignored is left out by your .gitignore, so it is never versioned. A file larger than 100 MiB can't be added to Pipeline files: put it in data/.

A hosted Haute app keeps only what version history keeps, so it takes no files in Local files.

Projects made before this layout

Older projects have empty rating/data/ and rating/outputs/ folders that nothing uses: delete them. If your .gitignore has a data/ line, change it to /data/. Without the leading slash it also ignores rating/files/data/. Add /outputs/ and /.cache/ while you're there.