API reference¶

This reference is generated from the source docstrings. Use it when you know which symbol you need and want its exact signature, parameters, exceptions, or return type. For task-oriented walk-throughs, see the guides. For design rationale, see the concepts.

Records and data¶

• Records: BlobRef, the view helpers, and the generated pub.layers.* record namespaces.
• Shared types: the JsonValue alias and the type variables.
• Dataset API: Corpus and load_corpus, the lazy Dataset, and the Features description derived from the generated models.
• Dataset discovery: discover_datasets, list_datasets, and table_of_contents, the DatasetFilter, DatasetSummary, and RepoTableOfContents result shapes, and the searchable discovery index.

ATProto access¶

• ATProto access: identity resolution, the PDS and appview record clients, blob fetch, and the firehose surface.

Store¶

• Store: the in-memory pool, the schematic Repository, the Arrow views, and the blob cache.

Media¶

• Media: media resolution, anchor resolution, and the audio, video, and neural decode and slicing helpers.

Authoring¶

• Authoring: the LayerBuilder and anchor builders (span, temporal, bbox, and the rest), the WriteClient, apply_writes, pull, and the PublishPlan diff against a PDS.

Integrations¶

• Integration ports: the Codec, Exporter, KnowledgeBase, and StorageBackend protocols that adapters bind to.
• Adapter registry: the registry and the top-level get_codec, get_exporter, and get_knowledge_base lookups by name.
• Format codecs: the shared corpus-fragment models and the CoNLL-U and brat codecs.
• Knowledge bases: the entity, candidate, and edge models and the Wikidata, reconciliation, and glazing connectors.
• Exporters: the framework-native exporters over an Arrow view (HuggingFace, PyTorch, TensorFlow, WebDataset).
• Experiment tracking: logging a Repository revision and its provenance bundle to Weights & Biases or MLflow.

Code generation¶

• Code generation: the lexicon-to-model pipeline, the schema-to-spec mapping, the emitter, and the vendoring manifest.

Tools¶

• Explorer TUI: the lairs.tui Textual application and the standalone DuckDB-backed QueryEngine.
• CLI: the lairs console script and its vendor, gen, pull, materialize, publish, inspect, datasets, toc, search, index (with build, update, search, and diff), tui, login, logout, and whoami subcommands.