A CLI tool that simplifies **pyinfra** script management and execution, providing an interactive workflow for deploying infrastructure configurations ("cubes") to remote hosts.
## Overview
Nopy wraps pyinfra with structure, validation, and an interactive experience for managing complex infrastructure deployments. It organizes deployments into self-contained "cubes" with dependency management, schema validation, and lifecycle hooks.
- **Python deployment script**: `deploy.py`, a plain pyinfra script
Configuration variables are declared in the manifest and validated with Zod schemas before the deployment script runs.
```
cubes/
├── .npcubes
└── apt/
└── install/
├── manifest.mjs
└── deploy.py
```
Any directory holding both files is treated as a cube, so cubes can be nested as deeply as you like to group them by topic. Discovery is recursive; directories starting with `.` and `node_modules` are skipped. Additional files in the cube directory (a `README.md`, config templates, and so on) are ignored by the loader and can be referenced from the deploy script — the script runs with its cube directory as the working directory.
The prefixed forms `<cube-name>.manifest.mjs` and `<cube-name>.deploy.py` are also still recognized, but plain `manifest.mjs` / `deploy.py` is the current convention.
A cube's identity comes from the manifest's `id` field (see below). If `id` is omitted, nopy falls back to an `[id]` prefix in the manifest `name`, and finally to the directory's own name. Note that the id does not have to mirror the folder path — `cubes/network/tailscale` declares `id: 'net:tailscale'`.
The matching `deploy.py` is a plain pyinfra script. Nopy passes each schema variable to pyinfra as `--data KEY=value`, so they are available on `host.data`:
Every key defined in the manifest `schema` is guaranteed to be present on `host.data` — either from the Zod `.default()`, from `.nopyrc.json`, from a recorded session, from a dependency, or from a user prompt.
**Value types**: pyinfra parses `--data` values before your script sees them. `"true"` / `"false"` become booleans, numeric strings become `int`, valid JSON becomes the parsed structure, and everything else stays a string. This is why `UPDATE` can be handed straight to pyinfra's `update=` argument, while `PACKAGES` is wrapped in `str(...)` before splitting.
Variable defaults are defined directly in the Zod schema using `.default()`. This ensures that every cube has a predictable starting state and provides type-safe default values.
A variable can be set from several places in one run. Every assignment is kept, tagged with where it came from — its **origin** — and the highest-ranked origin wins.
This allows cubes to ship with reasonable defaults while still allowing users to override them globally via `.nopyrc.json` or interactively during deployment. Because `env` outranks the schema, `.nopyrc.json` is also what steers a run started with `--use-defaults`, which never prompts.
`prompt` and `param` rarely compete: a key a dependency supplies is left out of the prompt entirely, so the user is only ever asked about the keys nothing else has set.
Ranking by origin rather than by arrival order is what makes replay work: a recorded value is applied *before* the cube would be prompted for, and prompting can still override it, but a `--data` value pushed in by a dependency is never clobbered by a stale recording.
A field declared without `.default()` has no `default` origin to fall back on. It is prompted for like any other, with an empty initial value — but a run that cannot prompt (`--use-defaults`) fails on it unless `env` or a dependency provides it.
#### Secrets
A manifest can name schema keys that hold sensitive values:
```javascript
exportdefaultcubes.Manifest({
id:'user:add',
name:'Add a user account',
secrets:['PASSWORD'],
schema:z.object({
USERNAME:z.string().describe('Username for the new account').default('deploy'),
PASSWORD:z.string().describe('Password for the new user account').default('changeme'),
})
})
```
Every entry must be a key of `schema`; naming anything else is a manifest error and aborts the run, so a typo fails loudly instead of silently leaving a value unprotected.
- **It is never written to a session file or to the history.** Everything else the run settled on is recorded — including values that came from a `.default()` — but declared secrets are left out.
- **It is masked wherever a command or a plan is printed** — `--dry-run`, `--print-only`, and the debug log all show `********` in place of the value, in the variable list *and* in the `pyinfra` command line above it. The SSH password passed via `--password` is masked the same way, whether or not any cube declares secrets.
- **It is re-prompted on replay**, since there is nothing recorded to replay from (see [Session Recording and Replay](#session-recording-and-replay)).
- **It stops travelling.** Ordinary `env` values are seeded onto every cube in the run, because a cube may read a key off `host.data` that its own schema never declared. A secret is the exception: it reaches only the cubes whose `schema` names it. Otherwise putting a password under `env` — which is what unattended replay asks you to do — would put it on the command line of every unrelated cube, where nothing masks it because that cube never called it a secret.
Declaring is global, masking is global. A key any manifest calls a secret is masked and kept out of sessions on every cube it lands on, even one whose own manifest forgot to list it. What is *not* global is the guess: a key called `PASSWORD` that no manifest declares anywhere is an ordinary variable — broadcast, recorded, and printed in the clear.
For a sensitive `env` value that no cube declares at all — a token only a hook reads, say — name it in the config instead:
```json
{
"secrets":["DEPLOY_TOKEN"],
"env":{"DEPLOY_TOKEN":"..."}
}
```
Entries here behave exactly like a manifest's: masked, never recorded, and delivered only to cubes that declare them.
Three limits are worth knowing, because `secrets` keeps a value out of the files nopy writes and nothing more:
- **It is on the command line.** pyinfra takes its data as `--data KEY=value`, so the real value is visible in `ps` for as long as the deployment runs. Masking covers nopy's own output, not the process table.
- **The prompt shows it.** The variable form displays and pre-fills what it is asking about, so a secret is on screen while it is being entered or confirmed.
- **A `.default()` is not protected.** A default lives in the manifest, in plain text, wherever the manifest is checked in. Give a secret a placeholder default like `changeme` if it needs one at all, never a real credential.
`history` controls automatic session recording (see [Deployment History](#deployment-history)), and `execution.continueOnError` sets the default for `--continue-on-error`.
`secrets` names `env` keys to treat as sensitive that no manifest declares — it is the config-side half of a manifest's `secrets`, and behaves identically. See [Secrets](#secrets).
`hosts` seeds the target picker; see [Target hosts](#target-hosts) for what else that picker offers.
`cubeDirs` holds paths, `cubePackages` holds installed npm packages that ship cubes — see [Cube Discovery](#cube-discovery) below and [CUBE-BUNDLES.md](docs/CUBE-BUNDLES.md) for publishing your own. Both are additive, and both resolve relative to the config file that named them, not to the working directory: a `.nopyrc.json` two levels up may name a package that only exists in *its*`node_modules`.
| `false` | (none) | No debug logs (default) | Normal operation |
| `true` | `--debug` | Enable pyinfra debug logs | Deep debugging of pyinfra internals |
**Recommendation:** Start with `"info"` for typical troubleshooting, use `"trace"` when investigating command failures, and enable `debug: true` only when debugging pyinfra itself.
### Session Recording and Replay
Nopy supports recording deployment sessions to JSON files for later replay. This is useful for:
- Repeatable deployments
- CI/CD pipelines
- Documentation and auditing
- Sharing configurations across teams
#### Session File Format
Sessions are stored in `.nopysession.json` files with the following structure:
- **`version`**, **`timestamp`**, **`name`**: stamped on every session nopy writes — the format version, the ISO 8601 record time, and a one-line description in the same `date - cubes → hosts` form the history list uses
Only `cubes` and `auth` are required. A hand-written session may omit the rest, and one that predates the stamp still loads; a `version` this build does not recognise is a warning on stderr, never a refusal.
`auth.method` has a third value the picker never offers: **`ssh`**, meaning the connector owns authentication and nopy supplies none. It is what an `@vagrant/` or `@docker/` host gets, which is why replaying one asks for nothing.
**What is recorded:** every value each cube settled on, regardless of where it came from — a value the user typed, one inherited from `.nopyrc.json``env`, one a dependency supplied, and one that fell through to the schema's `.default()` are all written out the same way. A session is therefore a full snapshot rather than a diff, and a `--use-defaults` run produces a session with real values in it instead of an empty one.
The consequence is that replay is faithful rather than re-derived: the recorded value outranks the current `.nopyrc.json``env` and the current schema default, so editing either one does not silently change what a replay does. To pick up a new default, record a fresh session.
**Security Note**: Passwords are never stored in session files. This covers both the SSH password — a session records the auth *method* and username, never the credential — and any schema key a cube's manifest lists under [`secrets`](#secrets). Both are re-prompted on replay. The rule applies to the session's `env` block as well as to each cube's `variables`, so a declared secret set in `.nopyrc.json` is left out of the recorded copy rather than written back out in plaintext.
A replay runs straight through without asking anything, with three exceptions. Password authentication always re-prompts. A session with no recorded host falls back to the host picker. And a cube is re-prompted for its declared secrets, plus for any required variable the session has no value for — which happens when the cube's schema has gained a field since the session was written.
Those re-prompts are what a session cannot supply, so `--use-defaults` cannot paper over them: combining `-D` with a replay that needs either fails with a message naming the keys rather than deploying with a placeholder. Put the values under `env` in `.nopyrc.json` — or pass them from a dependency — to make such a replay unattended. A secret supplied that way reaches only the cubes that declare it, so this does not broadcast it across the run; see [Secrets](#secrets).
A schema `.default()` is deliberately *not* accepted in its place. The recorded answer is gone on purpose, so falling back to the manifest would deploy a different credential than the run being replayed, and say nothing about it.
2. Cube directories of every package listed in `.nopyrc.json``cubePackages`
3. Directories containing a `.npcubes` marker file (searching upwards from current directory)
All three are unioned and scanned the same way. A directory is a cube when it holds both a manifest (`manifest.mjs` or `*.manifest.mjs`) and a deploy script (`deploy.py` or `*.deploy.py`); dotted directories and `node_modules` are skipped during the scan.
A cube package is an ordinary npm package that ships its cubes in a `cubes/` directory at its root. That is the whole contract — no nopy-specific `package.json` field is required. A bundle whose cubes live elsewhere (compiled into `dist/cubes`, say) overrides the location:
Naming a package is a statement that cubes are expected from it, so anything wrong is an error that aborts the run rather than a silent skip: the package is not installed, it has neither a `cubes/` directory nor a `nopy.cubes` override, its `nopy.cubes` is malformed, or an entry points at a directory that does not exist or lies outside the package.
A cube id such as `apt:essentials` is claimed across every source at once, not per directory or per package. Two cubes with the same id abort the run with an error naming both and where each came from. There is no precedence rule and no shadowing — a local cube does not quietly win over a packaged one, in either direction. Prefix your own cubes distinctly if you point `cubeDirs` at a local tree alongside an installed bundle.
Writing cubes to publish is covered in [CUBE-BUNDLES.md](docs/CUBE-BUNDLES.md).
Every deployment is automatically recorded to a `.nopy.history.json` file in the current working directory, so the last run is always available to `-R` without having to pass `--save-session` first. The default retention is the 10 most recent sessions (configurable via `history.maxSessions`); use `nopy history` to list them and `nopy install -H <id>` to replay any one of them — see [Deployment History](#deployment-history).
The recording happens before the deploy commands run, so a **failed** deployment is recorded too — `-R` is the quick way to retry one after fixing the cause. Replaying a session with `-R` or `-H` does not itself create a new entry, so repeating never pushes the original run out of the list.
A `--load-session` run *is* recorded, and the distinction is the point: a session file has never been in history, so without the entry `nopy history` would report nothing afterwards and `-R` would have nothing to repeat.
Because the history file is resolved against the current working directory, each project keeps its own history — running nopy from a different directory will not find the previous run. As with session files, passwords are never stored and are re-prompted on replay.
**Default behaviour (fail-fast)**: without this flag, nopy stops at the first cube that fails. Cubes are deployed sequentially in dependency order, so the failing cube's output is the last thing you see — every cube still queued behind it is skipped entirely and is never attempted.
This is deliberate: because cubes are topologically sorted, a cube that fails is often a dependency of the ones after it, and continuing would deploy them onto a half-configured host.
Two consequences worth knowing:
- **Cubes that already succeeded are not rolled back.** The host is left in a partial state — the cubes before the failure are applied, the rest are not. Fix the cause and re-run; well-written cubes are idempotent, so re-applying the earlier ones is normally harmless.
- **Skipped cubes are not reported as failed.** They are simply absent from the results, so a summary of "3 successful, 1 failed" out of 6 cubes means the remaining 2 were never run.
Either way, the command exits with code `1` if any cube failed, which is what CI picks up. Use `--continue-on-error` when your cubes are genuinely independent and you would rather collect every failure in one run than stop at the first.
The default can be flipped for a project by setting `execution.continueOnError` in `.nopyrc.json`; the CLI flag takes precedence over it.
History is what makes [Repeat last run](#basic-commands) work, but it holds more than just the last deployment: every recorded run stays replayable until newer runs push it out. `nopy history` (alias `nopy h`) lists them newest first, with `→` marking the entry that `-R` would replay:
Each entry records the selected cubes together with every variable value they ran with, the target hosts, the authentication method, and the username — never the password, and never a key the manifest declared a [secret](#secrets). Pass an ID to `-H` to run that exact combination again:
A replay is non-interactive: cube selection, host, and variable values all come from the entry, so nopy runs straight through without asking anything. It asks only for what the entry cannot hold — the password under password authentication, and any declared secret — plus the host picker when the entry recorded none.
- **Recorded values win over the current configuration.** The entry is a snapshot of everything the run settled on, so editing a cube's `.default()` or the `env` block of `.nopyrc.json` afterwards does not change what the replay does. A variable the schema has gained *since* the entry was written has nothing recorded: if it has a `.default()` the replay quietly takes it, and if it is required the replay prompts for it.
- **A replay fails if a cube no longer exists.** Renaming or deleting a cube id makes every history entry that referenced it unreplayable: nopy logs `Cube from session not found` and then aborts with `Cube not found: <id>`.
The history lives in `.nopy.history.json` in the working directory and uses the same structure as a session file, so trimming the array by hand is a perfectly good way to prune it. It does contain the variable values a run used, which is why it is listed in this repository's `.gitignore` — treat it like any other file holding deployment configuration. A corrupt or unreadable history file is treated as empty rather than raising an error, which looks exactly like a project that has never been deployed from.
For a run you want to keep indefinitely, don't rely on history — it rotates. Use `--save-session` to write it to a file you control (see [Session Recording and Replay](#session-recording-and-replay)).
