torchcell.knowledge_graphs.kg_manifest.bootstrap_manifest

bootstrap_manifest(*, repo_root: Path, commit: str, dataset_classes: list[str], n_experiments: dict[str, int], database: str, store_host: str, neo4j_version: str, biocypher_version: str, biocypher_out: str, built_at: str, organisms: frozenset[Organism], generation: GenerationRecord | None = None) → KgBuildManifest

Reconstruct the manifest of a store built by a FULL build at commit.

Every fingerprint is computed from the sources AT THAT COMMIT (git show), so the manifest describes what was serialized into the store, not the working tree. The adapter files of each dataset are resolved through the working tree’s dataset_adapter_map and then hashed at the commit.

organisms is the set the build was asked for (--organisms). A served dataset whose loader declares an organism outside it raises: the store would then hold data the recorded set says it does not.