Skip to content
AggregaAggrega
Select language: English

aggrega config

Inspect and manage the workspace configuration

Synopsis

aggrega config

Description

Inspect and manage the configuration Aggrega reads.
Most projects need two files.
One .aggrega-config.yaml at your project root, carrying a "workspace:" block —
that block is what marks the root — holds how this project builds: its catalog
directories, its provider bindings, which lint rules it relaxes. It is
committed, so a teammate who clones the project builds it the same way.
One global file is shared across every workspace on this machine:
~/.config/aggrega/config.yaml, or on Windows aggrega\config.yaml inside the
directory the APPDATA environment variable names. It holds what belongs to the
machine rather than to any one project: the registries you pull from and the
signature trust policy for each, how large the compiled-module store may grow,
any command extension you have installed. It loads automatically when it is
there, and nothing breaks when it is not.
Your project's file wins over the global one, an environment variable wins
over both, and a flag wins over everything.
Secrets are not configuration. A registry password or a signing identity token
is never a field in either file; it is supplied by environment variable or flag
at the moment it is needed.
Because the effective configuration is assembled rather than written anywhere,
"aggrega config show" is how you see what actually applies, and --origin is how
you find out which layer each setting came from. "aggrega config init" seeds a
working configuration, and "aggrega config capture" freezes the effective
build-critical settings into this workspace so a build does not depend on a
global file that is not committed.
Larger projects have more to work with. None of it is needed to start, and each
piece exists for the shape it solves:
More than one file inside a project. Every .aggrega-config.yaml from the
project root down to the directory you run in contributes, a nearer file
overriding a farther one. Only a file carrying "workspace:" marks the root;
one without it contributes settings without anchoring anything.
Extending a list instead of replacing it. A nearer file's list replaces the
farther one's. The companions catalog_dirs_append, governance_append,
linter_append, importer_append and providers_append concatenate onto the
resolved list at each layer instead of shadowing it.
A file your pipeline owns. .aggrega-config-append-ci.yaml may sit beside the
hand-authored file in any of those directories, so automation can overwrite
it wholesale without touching what a person wrote. It is additive only: it
removes nothing and overrides no scalar, and a scalar key written in it is
ignored and named in a warning. It never marks a root.
A relative path anchors to the file that declared it. A catalog directory, a
local plugin path, or an entry under a plugin's with.sources resolves against
the directory of the config file it was written in — not the project root,
and not the directory you ran the command from. So the same "./catalog" names
a different directory depending on which file carries it. Two settings sit
outside this rule: artifact_cache_dir is interpreted relative to the project
root wherever you run from, and AGGREGA_CATALOG_DIRS has no declaring file at
all, so supply absolute paths in it.
The other root markers. A "workspace:" block wins outright at whatever depth
it sits. Failing one, the nearest .aggrega/ directory or aggregate*.yaml,
release*.yaml or manifest*.yaml file anchors the root, in that order within a
directory; failing everything, your working directory is used.
--workspace-root skips discovery entirely.
Naming the global file. --config <path> wins over the AGGREGA_CONFIG
environment variable, which wins over the default location. Both name a file
deliberately, so a path either gives that does not exist is a loud error
rather than a silent fallback. Whichever file resolves is also the one a
config-writing command writes when it writes globally.

Subcommands

CommandDescription
aggrega config captureFreeze the effective build-critical configuration into this workspace
aggrega config initSeed a working configuration with a registry and its trust policy
aggrega config showShow the effective configuration and where each setting came from

aggrega config capture description

Write the effective build-critical configuration into this project's workspace
config file, so the workspace builds identically without relying on the
machine's global configuration.
The build-critical subset is the configuration a build participates in: a set of
plugin socket attachments and the catalog directories. By default the captured
sockets are only those that determine the
rendered output — the provider bindings and the transform-role governance attachments — since only those must
travel with the project for byte-identical output. Pass --all-plugins to also
freeze the verdict-role governance, linter, and importer sockets, which otherwise stay free to
vary by environment. Each field is written at its effective value — the value
that resolves right now with the global configuration and the workspace tree
merged together — so a binding that today comes only from the global config is
materialized into the workspace and travels with the project.
Machine-wide tooling settings, such as a registry's trust policy, are not
captured: they stay in the global config by their nature. Capture always writes
this project's workspace config file. Running it again on an unchanged
configuration rewrites the same file byte-for-byte. Run in a folder that is not
inside a workspace, with no --workspace-root naming one, it writes nothing and
exits with an error saying so.

aggrega config capture flags

FlagDefaultDescription
--all-pluginsfalseCapture all plugin sockets, including the verdict-role governance, linter, and importer gates. Without it, capture freezes only the sockets that determine the rendered output (the provider bindings and the transform-role governance attachments).

aggrega config init description

Seed a working configuration by registering a registry and its signature trust
policy, so a fresh machine can pull and verify signed artifacts without editing
YAML by hand.
By default it registers the official hub (hub.aggrega.ai) with signatures
required and the official trusted identities. The registry's advertised trust
policy is fetched once from its own origin and pinned locally — trust-on-first-
use; when the hub is unreachable, advertises nothing, or the workspace is
offline, its built-in official baseline is written instead, so init still
produces a usable config offline. Pass --registry to seed a different registry
(for an air-gapped or private hub). Such a registry has no built-in baseline,
so seeding it while the workspace is offline is refused: init takes no
identities or key to pin by hand, so register it once while online, or add it
with "aggrega registry add --identity" or "--key".
The entry is written to the global config shared across every workspace by
default; pass --workspace to write it to this project's workspace config
instead. If the target config file already exists, init refuses rather than
overwrite it — pass --force to write anyway (the registry entry is merged into
the existing file, other settings preserved).
Before anything else, init says where things are: the configuration file it
created — or, under --force, the one already there that it wrote into — and
the directory the machine-wide caches are kept in. On Windows that directory is
aggrega\cache inside the directory the LOCALAPPDATA environment variable names;
elsewhere it is a directory named aggrega under the per-user cache directory:
~/.cache/aggrega on Linux unless XDG_CACHE_HOME names another, and
~/Library/Caches/aggrega on macOS.
A seeded configuration is a document you own and go on editing, so --dry-run does
everything init does except write: it hands the content back, creating and
modifying nothing. The answer names the file the content is for — which matters
here, because the target is the global config unless --workspace selects this
project's own. A caller that keeps its own document model puts the content there,
where it is visible before it is saved and can be amended or undone. The registry
is still contacted for its advertised trust policy, because the policy in that
content is the policy a write would have pinned.
Because a dry run does everything else init does, it refuses an existing config
file exactly as a write refuses it, and for the same reason. So its exit code
predicts the write's: if the dry run refuses, init would have refused. Pass
--force alongside it to see the merge an overwriting write would produce.
Under --dry-run standard output carries that content and nothing else, so it stays
parseable. A registry seeded with no trust policy, because it advertises none,
still produces its notice, on standard error: read that stream too, or that
configuration arrives without the advice to pin identities by hand. A pinned
policy, whether advertised or the hub's built-in baseline, is in the content
itself and produces no notice.

aggrega config init flags

FlagDefaultDescription
--dry-runfalseDo everything except write: hand back the seeded configuration as content, naming the file it is for, and create or modify nothing. Every other behaviour is the write’s, so an existing config file is refused exactly as a write refuses it — pass —force alongside it for the merge an overwriting write would produce.
--forcefalseOverwrite an existing config file. Without it, init refuses when the target file already exists.
--globalfalseWrite the entry to the global config shared across every workspace. This is the default.
--registryhub.aggrega.aiRegistry host to seed with its trust policy. Defaults to the official hub.
--workspacefalseWrite the entry to this project’s workspace config instead of the global config.

aggrega config show description

Show the configuration that applies right now.
By default it leads with the cascade of config files and environment overrides
that contributed, lowest precedence first, then prints the effective merged
configuration — the global config, the workspace-tree cascade, and any
environment override combined into what applies right now.
Pass --global or --workspace to print a single layer's own file bytes verbatim,
preserving its comments and formatting; when that layer's file does not exist it
says so rather than failing. The two are mutually exclusive.
Pass --origin to annotate each effective setting with the layer it came from —
a command-line flag, an environment variable, a config file, or the built-in
defaults, highest precedence first. A flag is named as the reader typed it, so
a setting a flag decided reports that flag as its source. A setting is
attributed to the highest-precedence layer that sets it, whatever value that
layer writes: a file restating the default is the file named, not the built-in
defaults. List and map settings are attributed as a whole. The signing
requirement is the one exception: any layer may turn it on and none can turn it
off, so it names the first layer that turns it on.

aggrega config show flags

FlagDefaultDescription
--globalfalsePrint the global config file’s own bytes verbatim, preserving its comments and formatting.
--originfalseAnnotate each effective setting with the layer it came from (a command-line flag, an environment variable, a config file, or the built-in defaults).
--workspacefalsePrint this project’s workspace config file’s own bytes verbatim, preserving its comments and formatting.

What it does

The config subcommands seed, inspect, and freeze the configuration that governs a build. For which file a setting belongs in and how the layers combine, see Configuration — it starts with the two files most projects need and keeps the rest behind that.

config init — seed a working configuration

aggrega config init seeds a config on a fresh machine by registering a registry and its signature trust policy, so signed artifacts pull and verify without hand-editing YAML. By default it registers the official hub (hub.aggrega.ai) with signatures required and its official trusted identities. The registry’s advertised trust policy is fetched once from its own origin and pinned locally (trust-on-first-use); when the hub is unreachable, advertises nothing, or the workspace is offline, its built-in official baseline is written instead, so init still produces a usable config offline.

Before anything else, init says where things are. Seeding a fresh Linux machine opens with:

Created the global configuration file at /home/you/.config/aggrega/config.yaml
Machine-wide caches are kept in /home/you/.cache/aggrega

and a fresh Windows machine with:

Created the global configuration file at C:\Users\you\AppData\Roaming\aggrega\config.yaml
Machine-wide caches are kept in C:\Users\you\AppData\Local\aggrega\cache

The first line names the configuration file init wrote: Created when the file was not there, and Using when it was and --force let init write into it; under --workspace it names the workspace configuration file instead of the global one. The second names the directory the machine-wide caches — compiled plugins and their descriptions — are kept in, which is where each platform keeps a user’s caches; Which caches belong to this project, and which to the machine gives the rule per platform. Each line ends with its path and nothing after it, so the path can be copied as it stands. The lines that follow are the registration’s own: what the trust-policy fetch found, and which file the registry was registered in. A seeding that is refused prints neither line, since it created nothing, and neither does --dry-run, whose standard output holds only the content it hands back.

  • --registry <host> seeds a different registry — an air-gapped or private hub — instead of the official one. Such a registry carries no built-in baseline, so where it advertises nothing the entry is written with no policy, and in an offline workspace seeding it is refused (short code E0310): the policy cannot be fetched, and init takes no --identity or --key to pin one by hand. Seed it once while online, or add it with aggrega registry add and a policy you supply.
  • If the target file already exists, init refuses rather than overwrite it; --force writes anyway, merging the registry entry into the existing file and preserving other settings.
  • --dry-run does everything init does except write. It hands the seeded configuration back as content, creating and modifying nothing, and the answer names the file the content is for — which matters here, because the target is the global config unless --workspace selects this project’s own. A caller that keeps its own document model, such as an editor, puts the content there, where it is visible before it is saved and can be amended or undone.
  • Because a dry run does everything else init does, its exit code predicts the write’s. It refuses an existing config file exactly as a write refuses it, and for the same reason, so you can use it to ask whether init would succeed before committing to the change. To see the document an overwriting write would produce, pass --force alongside it: --force --dry-run reports the merge and still writes nothing.
  • The registry is still contacted for its advertised trust policy, because the policy in the reported content is the policy a write would have pinned. A dry run makes the same single unauthenticated request to that registry’s own origin that a write makes.
  • Under --dry-run, standard output carries that content and nothing else, so it stays parseable. The one thing the content cannot express is an absence: when a registry advertises no trust policy and is seeded with none, the notice saying so — and telling you to pin identities with --identity or --key — goes to standard error. A configuration document with no trusted identities looks the same whether a policy was pinned and happened to be empty or none was advertised at all, so read both streams; a program that discards standard error is told nothing about it. When the registry does advertise a policy, or the official hub’s built-in baseline is pinned in its place, the pinned policy is in the reported content itself and no separate notice is produced.

A seeded configuration is a document you own and go on editing, which is why it can be handed back rather than written. Freezing the effective configuration, below, produces a record of what the configuration resolved to — state the compiler keeps about itself rather than content anyone composes — so it always writes and offers no --dry-run.

config show — inspect what applies right now

aggrega config show reports the configuration that applies right now. In its default view it leads with the cascade of config files and environment overrides that contributed, lowest precedence first, then prints the effective merged configuration — the global config, the workspace-tree cascade, and any environment overlay combined into a single result. This loaded-file cascade is where you confirm which files and environment overrides shaped the current configuration.

Two focused views narrow the output:

  • --global or --workspace prints a single layer’s own file bytes verbatim, comments and formatting preserved; when that layer’s file does not exist it says so rather than failing. The two are mutually exclusive.
  • --origin annotates each effective setting with the layer it came from — a command-line flag, an environment variable, a config file, or the built-in defaults, highest precedence first. A flag is named as you typed it, so a setting --offline decided on this run reports --offline as its source; an environment variable is named behind an env: prefix and a config file by its absolute path.

Under --origin, a setting is attributed to the highest-precedence layer that sets it, whatever value that layer writes. So a project file that writes default_registry: hub.aggrega.ai is named for that setting even though the built-in default is the same registry, and a project file restating a value the global file already sets is named rather than the global file: in both cases it is the file holding the value in place, and the one whose removal changes the answer the day the default or the file beneath it moves. A setting no layer sets reports the built-in defaults.

  • What counts as setting it. A config file sets a setting when it carries that setting’s key, and a list setting also when it carries the list’s _append companion. An environment variable sets its setting when it is exported with a non-empty value, and a flag when you type it — --offline=false sets the posture as surely as --offline does.
  • A configuration file your pipeline owns is named only for the keys it is read for, which are its list additions and the signing policy. A key it ignores, and warns about, is never reported as where a value came from.
  • List and map settings are attributed as a whole, to the highest-precedence layer that sets them; which layer contributed which entry of a list is not reported.
  • The signing requirement is the one exception. Any layer may turn it on and none can turn it off, so it names the first layer that turns it on. A layer writing required: false turns nothing on, so where no layer turns it on the setting reports the built-in defaults.

config capture — freeze build-critical settings

aggrega config capture writes the effective build-critical configuration into this project’s workspace config file, so the workspace builds identically without relying on the machine’s global configuration. Machine-wide tooling settings such as a registry’s trust policy are not captured; they stay in the global config by their nature. Running it again on an unchanged configuration rewrites the same file byte-for-byte.

The captured subset is a set of plugin socket attachments and the catalog directories:

  • By default capture freezes only the sockets that determine the rendered output — the provider bindings and the transform-role governance attachments — since only those must travel with the project for byte-identical builds. The catalog directories are always captured.
  • --all-plugins widens the captured sockets to include the verdict-role governance, linter, and importer sockets as well. Without it those three stay free to vary by environment.

Capture always writes, and offers no way to be handed the content instead. What it produces is a record of what the configuration resolved to, not a document anyone composes, so there is nothing for an author to review and amend — reporting it would invite editing a snapshot of a resolution.

Capture writes only into a workspace. Run in a folder that is not inside one, with no --workspace-root or AGGREGA_WORKSPACE_ROOT naming one, it writes nothing — a configuration file written there would turn the folder into a workspace — and exits with an error saying so. aggrega config init --workspace makes the folder a workspace first.

See also

  • configuration.md — which file a setting belongs in, the precedence ladder, and the mechanisms a larger project reaches for.
  • global-flags.md — the --config and --workspace-root flags.
  • environment-variables.md — the AGGREGA_* variables that override configuration at runtime.
  • aggrega-registry.md — manage registries and their trust policy directly.
  • aggrega-cache.md — what the directory artifact_cache_dir names actually holds, and why an entity sitting there that no lockfile pins is not found by a build.