Skip to content

Workspace

A project’s workspace is its live, mutable file tree — what the next run reads, and what the agent is working in. A snapshot is a frozen copy of it; a deployment is the company-level original the project was seeded from. aura workspace reads and changes that live tree, which is what a CI job needs to swap in today’s input data and solve again.

Terminal window
aura workspace file list # tree view (visible dirs)
aura workspace file list --prefix data/ # just the inputs
aura workspace file list --flat # flat table (old style)
aura workspace file read run/solve.py # bytes to stdout
aura workspace file upload ./orders-june.csv data/orders.csv
aura workspace download --prefix data/ --out inputs.zip
Command Does
aura workspace file list Tree of workspace files (--prefix <path>, --flat)
aura workspace file read <path> Print one workspace file to stdout
aura workspace file upload <local> <remote> Upload one local file into the workspace (--conversation <id>)
aura workspace file delete <path> Remove one workspace file (y/N prompt; --yes to skip; --conversation <id>)
aura workspace download Download the workspace as a zip (--prefix <path>, --out <file>)

upload spells out both paths, one file per call. Nothing is inferred: aura deploy maps your model folder onto run/, so a module you edit at ./src/model.py lives at run/src/model.py in the workspace — defaulting to the local path would quietly write the wrong one. A workspace destination must also name a directory (data/orders.csv, not orders.csv).

Dot-prefixed path segments are rejected on every container: listings skip dotfiles, so writing one would create a file nothing can read back.

A listing with no --prefix shows define/, data/, run/ and solutions/ — the same set a snapshot freezes. Pipeline internals under modeling/ are left out of it; --prefix modeling/ still reaches them, so the default narrows what you stumble into rather than what you can ask for. The chat session’s file browser shows the same four.