module.toml Manifest Reference
The module.toml file is the manifest at the root of every rescile module. It defines the module’s identity,
declares the parameters it accepts, specifies the asset and input schemas it expects from the consuming project,
and configures data generators that fetch or scaffold data automatically before the graph is built.
When a module is loaded via --module, rescile reads this file first to validate parameters, assets, and inputs
before any graph processing begins. It also drives the bootstrap command and dependency resolution via
rescile-ce mod.
[module]
name = "my-module"
version = "1.0.0"
description = "A custom module"
rescile_version = ">=0.1.85"
schemas = ["./schemas/infrastructure.schema.json"]
[dependencies.other_module]
url = "https://github.com/rescile/other.git"
resource_aliases = { "old_type" = "new_type" }
[params.env]
description = "Environment variable"
default = "dev"
required = false
[generators.fetch_data]
target_input = "config.json"
command = ["script.sh"]
env = ["TOKEN={{ env.TOKEN }}"]
ttl = "1h"
condition = "on_missing"
[generators.fetch_data.secrets.my_token]
collection = "my-app-secrets"
secret = "API_TOKEN"
[module] — Module Identity
Declares the module’s name, version, human-readable description, and the minimum rescile version it requires. These fields are displayed in the rescile dashboard and used during dependency resolution.
| Key | Required | Description |
|---|---|---|
name |
Yes | A short identifier for the module (e.g., "standard-webapp"). |
version |
Yes | Semantic version string (e.g., "1.0.0"). |
description |
No | A human-readable summary shown in the module index. |
url |
No | Canonical URL of the module’s repository. |
rescile_version |
No | Minimum rescile version constraint (e.g., ">=0.1.85"). |
For detailed usage see Using Modules.
[dependencies] — Module Dependencies
Declares other modules this module depends on. Three declaration formats are supported: a simple Git URL string,
a detailed { git = "...", tag = "..." } table, or a release archive { url = "...", version = "..." } table.
An optional resource_aliases map transparently remaps resource type names from the imported module to avoid
naming conflicts in the global graph.
Dependencies are resolved recursively. Use rescile-ce mod lock to pin them to exact commits for reproducible
builds.
For the full reference including lockfile commands see Module Dependencies & Lockfiles.
[params] — Module Parameters
Declares the named parameters a module accepts at runtime via --module-params. Each parameter entry can specify
a description, a default value, whether it is required, and an allowed_values list for validation.
Declared parameters are made available as {{ params.<name> }} in the Tera context of all models/,
compliance/, output/, and actions/ files of the module.
For backward compatibility the same parameters are also still exposed as root-level template variables
({{ <name> }}). This legacy form works but emits a deprecation warning; new authoring should prefer
{{ params.<name> }}.
Parameters are also exposed to action constraint templates as {{ params.<name> }} (for example,
hostname:{{ params.runner_hostname }}) and returned in the /api/engine-info endpoint as module_params.
| Key | Description |
|---|---|
description |
Human-readable explanation of the parameter’s purpose. |
required |
Boolean. If true, rescile will error when the parameter is not supplied. |
default |
Fallback value used when the parameter is not supplied and required is false. |
allowed_values |
Array of permitted string values. rescile validates the supplied value against this list. |
For usage examples see The Data Directory — Parameters Make Modules Reusable.
Schema Declarations
Modules can bind their resources to central schemas. When schemas are bound, the validator checks required properties, relations, capability coverage, and CSV value conformance. See Schema Validation for the full validation behavior.
schemas = [...]
Load one or more schema sources for the whole module:
schemas = ["./schemas/infrastructure.schema.json"]
[[schema]]
Bind a schema to specific resources:
[[schema]]
schema = "./schemas/infrastructure.schema.json"
resources = ["virtual_machine", "network"]
[[implements]], [[requires]], [[offers]]
Declare how the module relates to a published schema:
[[implements]]
schema = "rescile/infrastructure"
version = "20260729"
resources = ["virtual_machine"]
[[offers]]
schema = "rescile/infrastructure"
resource = "virtual_machine"
verb = "provision"
action = "lxd-provision"
Per-asset override
Override the schema used for a single CSV asset:
[assets."virtual_machines.csv"]
schema = "./schemas/custom-vm.schema.json"
[assets] — Asset Import Directives
Deprecation notice: The legacy
columnsandbootstrapfields inside[assets]are deprecated. Define expectations with central schemas (see Schema Declarations) and ship starter CSV rows in the module’sassets/directory instead, copying them withrescile-ce bootstrap.The following fields remain valid import-time directives with no schema equivalent:
name_template,match_on,resource_type,schema,extends,description, andbootstrap_policy.
Declares import-time directives and optional per-asset schema overrides for CSV asset files.
Each entry can set an explicit resource_type, a name_template that dynamically rewrites the
primary key (name property), match_on rules that filter rows before they are imported, a
per-asset schema override, a satellite extends marker, an optional description, and a
bootstrap_policy that controls how starter CSV rows are merged into the workspace.
The legacy columns validation table and bootstrap string are still parsed but should be
replaced by central schemas and module starter files.
| Field | Type | Description |
|---|---|---|
name |
string | The asset filename, e.g. servers.csv. |
resource_type |
string | Override the resource type inferred from the filename. |
name_template |
string | Tera template that produces the row primary key (name). |
match_on |
array | Filter rows before they are imported into the graph. |
schema |
string | Path or URL to a JSON schema that validates this asset. |
extends |
string | Satellite marker for a derived resource type. |
description |
string | Human-readable purpose. |
bootstrap_policy |
string | once, append, merge (default), replace, or skip. |
For the modern approach see Bootstrapping a Project and Asset Management — Module Asset Configurations.
Shared asset types across modules: if two modules declare column specifications for the same asset file, startup fails with
Asset '<file>' has conflicting column specifications ... Please use resource_alias to resolve the conflict.When your module merely references a type owned by another module (e.g.system,team), ship its reference CSV rows without an[assets."<file>"]block and declare expectations via schemas. See Composing Multiple Modules.
[input] — Input Schema & Bootstrap (Deprecated)
Deprecation notice: The
[input]and[[inputs]]blocks are deprecated. Place starter JSON input files in the module’sinput/directory and copy them withrescile-ce bootstrap. Central schemas can validate structure when needed.
Declares the JSON input files the module expects in the consuming project’s data/input/ directory. Each entry
defines the expected format (currently object_of_objects), optional fields validation, and a legacy
bootstrap string.
For the modern approach see Bootstrapping a Project and Input Management.
[generators] — Data Generators
Configures Phase 0 materialization of HTTP API data before the graph is built. Each generator references a
reusable external/<name>.toml fetcher and specifies one target_input or target_asset.
[generators.catalog]
external = "catalog"
target_input = "catalog.json"
args = { region = "{{ params.region }}" }
condition = "on_missing"
abort_on_failure = true
Arguments support literals, module parameters, and environment values. Request details, secrets, pagination, and cache TTL belong to the external fetcher. Inline HTTP generators remain executable with a deprecation warning for the current major release. Command/script execution is unsupported; obsolete command fields remain parseable only to provide a focused migration error. Both legacy field sets will be removed in the next major release.
For the complete generator reference including security and validation behaviour see Data Generators.
external/ — On-demand model inputs
external/ is a module directory, not a module.toml block. It contains named TOML HTTP fetchers used by
model-header external! declarations. A model reference such as "external!" = "vendor_lookup" resolves
external/vendor_lookup.toml in the same module.
External fetchers differ from generators:
- generators run before asset loading and write files;
- external fetchers run only for argument combinations reached by model rules and return in-memory JSON;
- external fetchers are HTTP-only by schema and have no
runtimefield; - generators can reference the same fetchers to persist Phase 0 data;
- external fetch failure is a hard graph-build error.
Fetcher files support sequential steps, GET/POST, GraphQL queries, headers, response chaining, pagination,
JMESPath, TTL caching, and Vault secret declarations. With --offline, an exact-key cached response is used even
when expired. If no compatible cache exists, the build fails without sending a request.