fact_checkChoosing a process

Two questions decide your chain: what output do you need, and does the graph fit your hardware for reasoning and validation.

The two questions

1. What is the output?
   • Read the data → Query.
   • Read the data plus what follows from it → Reason, then Query.
   • A conformance verdict → Validate, after Reason if the shapes constrain entailed facts.
   • A frozen, verifiable copy to hand on → Seal, then Unload if it should leave the database.
2. Does the graph fit your hardware for Reason and Validate?
   • Yes → load it and run the chain on it directly.
   • No → load all of it, carve the slice you need into its own graph, and run Reason and Validate on the slice.

Decision table

Your goal	Graph fits your hardware	Graph too big to reason over
Query only	Load → Query	Load → Query — Query isn't bound by the one-session limit
Inferred query	Load → Reason → Query	Ingest → Carve → Reason
Validation gate	Load → Validate → Query	Ingest → Carve → Reason, then validate the slice
Reason + validate	Load → Validate → Query, with its Reason step	Ingest → Carve → Reason, then validate the slice
Publish a frozen copy	Any of the above, then Seal and Unload	The same, on the slice

The one rule that drives it

Import scales to the complete 8.2-billion-triple Wikidata dump. Reason and Validate run in one session over one graph, and that graph has to fit your hardware. So:

Query at any scale. Reason and validate at your scale — carve a slice when the source is bigger than one session can hold.

See also

• Building a chain — the design method once you've chosen.
• Overview — the seven verbs and how each one scales.