marketplace:index

Builds the marketplace index, one index.json with every released feature’s contract and every published SPL model. marketplace:search, feature:info and the marketplace web app read it. Run it with --local during development to index your workspace offline.

Example

Index the default organisation from the forge.

splent marketplace:index

Index the workspace offline.

splent marketplace:index --local

Build the published index in CI from a curated list of sources.

splent marketplace:index --registry registry.toml --output ./public/index.json

Options

Option	Description
`--org ORG`	Forge organisation to index. Repeatable. Defaults to `SPLENT_DEFAULT_ORG`, then `splent-io`.
`--repo ORG/REPO`	An extra repository to index, such as a third-party feature outside the organisations. Repeatable.
`--registry FILE`	A `registry.toml` that lists the organisations and repositories to index.
`--local`	Index the editable features and SPL models of the workspace instead of the forge and the marketplace. Features need no network.
`--output PATH`	Where to write the index. Defaults to `.splent_cache/marketplace/index.json`.
`--no-pypi`	Skip the PyPI check, which is faster.

What it does

Features. For every splent_feature_* repository it reaches, it takes the latest semver tag, downloads pyproject.toml at that tag and turns [tool.splent.contract] into an entry. The entry also lists the settings declared under [tool.splent.config], each with its default, whether only the deployment can supply it (<SET>) and its comment (see Architecture). A repository without a released tag is reported and skipped.
SPL lines. It reads every line from the marketplace’s public GET /api/v1/spls at SPLENT_MARKETPLACE_URL (https://marketplace.splent.io by default) and indexes each at the DOI it publishes. The model is read by that DOI, from .splent_cache/spls/ or UVLHub, so unpublished edits in a working copy never reach the index. The line’s version fills uvl.version.
Models. It parses each model (mandatory and optional features, alternative groups, constraints). A model that cannot be fetched is indexed without one, with a warning, and never fails the build.
Relations. It computes used_by (reverse dependencies) and the collisions, features that provide the same route, service or model.
Writes the index atomically, so an interrupted write leaves the previous index readable.

The forge is the one the CLI is configured for, git.splent.io by default (see Tokens). Set SPLENT_FORGE_TOKEN, or GITHUB_TOKEN, to raise the API rate limit and reach private repositories.

With `--local`

It indexes every SPL model the workspace knows, meaning the working copies in splent_spl_*/, the models in .splent_cache/spls/, the ones products pin by DOI, and the one the CLI ships a pin for (sample_splent_spl). A model missing from disk is downloaded by DOI when possible. When that fails the SPL is indexed without its parsed model and ` could not fetch UVL for : ` is printed. An SPL with no DOI is indexed without a model and without a warning.

The registry file

[registry]
orgs  = ["splent-io"]
repos = ["some-org/splent_feature_billing"]

Third-party authors are listed by adding their line here through a reviewed pull request. The marketplace never accepts pushed data.

Special cases

A rate limit, a refused request or an unreadable marketplace. The build stops before writing anything, so a partial build never replaces a good index.
An organisation that is not found or your token cannot see. The build stops the same way and exits 1. Names are not normalised, so splent_io fails where splent-io works.
```
  ❌ Org 'splent_io' not found or not accessible.
  Check the org name and that your token can see it.
  Index NOT written.
```
An SPL line with a name that is not lowercase letters, digits and underscores. It is skipped with a warning, since the name becomes part of a cache path.
```
  skipped an SPL line with an invalid name 'Bad-Name' (expected lowercase letters, digits and underscores)
```