Skip to main content

Python API

Supported surface​

Most callers should enable the plugin through mlia check. Integrations that already receive an MLIA AnalysisRunResult can invoke the artifact writer directly:

from mlia.model_explorer.artifacts import ModelExplorerArtifacts, write_artifacts

generated: ModelExplorerArtifacts = write_artifacts(analysis_run_result)
print(generated.manifest)

The supported import surface of mlia.model_explorer.artifacts is:

  • The write_artifacts(result) function writes graph, overlay, navigation, and manifest files under <result.output_dir>/model-explorer.
  • The ModelExplorerArtifacts class provides immutable paths describing the generated root, manifest, standardized output, node-data files, generated models, sync-navigation data, and entity-node mappings.

write_artifacts does not build bundle/index.html; the normal analysis plugin does that after writing artifacts.

Input expectations​

The supplied AnalysisRunResult must follow the MLIA analysis-plugin contract. In particular, its standardized output should already have been validated and filtered by MLIA core. Direct callers are responsible for validation if they bypass normal plugin invocation.

The writer accepts an absent output by producing a minimal empty result. A model that is present must be supported and must map unambiguously to source-operator identities.

Stability​

The path container and writer imports above are the intentionally supported package API. Helpers in bundler, graph_data, and _artifacts support the plugin implementation and may evolve with the embedded visualizer and artifact contract.