Storage and paths¶
The Fuseki triple store, seed ontologies and on-disk caches.
Fuseki¶
FUSEKI_URI¶
Fuseki HTTP server root (e.g. http://localhost:3030), not a dataset path or #/dataset/... UI URL; use FUSEKI_DATASET for the dataset name.
FUSEKI_AUTH¶
Fuseki credentials as user/password or user:password. Optional: FUSEKI_URI alone selects Fuseki, unauthenticated.
FUSEKI_DATASET¶
Facts dataset name; defaults to the name derived from 'ontocast'/'test'. ontocast serve and ontocast process replace it with the name derived from the tenant and project; only an embedded ToolBox reads it.
FUSEKI_ONTOLOGIES_DATASET¶
Ontologies dataset name; derived like FUSEKI_DATASET, and replaced the same way by ontocast serve and ontocast process.
FUSEKI_SHAPES_DATASET¶
SHACL shapes dataset name; derived like FUSEKI_DATASET, and replaced the same way by ontocast serve and ontocast process. Kept apart from the ontologies dataset because catalog discovery claims every named graph carrying an owl:Ontology subject, and a shapes document declares one.
Paths and caches¶
ONTOCAST_ONTOLOGY_DIRECTORY¶
Directory of seed ontology *.ttl files, read once at startup. Read-only: ingestion never writes here and deletion never removes files from it
ONTOCAST_CACHE_DIR¶
Cache directory for LLM responses and tool outputs
ONTOCAST_CACHE_MAX_BYTES¶
Size ceiling for the whole cache directory. Once exceeded, least-recently-used entries are deleted until the total fits. Accepts a byte count or a human size such as '1GB' or '500MB'. Set to 0 to disable automatic pruning.
ONTOCAST_CACHE_TTL_DAYS¶
Delete cache entries not used for this many days, applied before the size ceiling. None disables the age cut.
ONTOCAST_CACHE_PRUNE_EVERY¶
Re-check the cache size ceiling after this many writes. The check walks the cache tree, so it is amortised rather than run per write.