fact_checkChoosing a process

Two questions decide your chain: what output do you need, and does the graph fit one backend for the single-threaded steps.

The two questions

1. What is the output?
   • Just read the data → Query.
   • Read the entailed data → Reason, then Query.
   • A conformance verdict → Validate (optionally after Reason).
2. Does the graph fit your box for Reason / Validate?
   • Yes → a simple head→tail chain, no carving.
   • No (the source is larger than one backend can close over) → add the carve waist.

Decision table

Your goal	Graph fits the box	Graph larger than the box
Query only	Load → Query	Load → Query — Query is not the single-threaded constraint; the full 8.2 B graph is queryable in place
Inferred query	Load → Reason → Query	Ingest → Carve → Reason
Validation gate	Load → Validate → Query	Ingest → Carve → Reason → Validate the slice
Reason + validate	Load → Reason → Validate → Query	Ingest → Carve → Reason → Validate the slice

The one rule that drives it

Import and Seal scale to the full 8.2-billion-triple ceiling; Reason and Validate are single-threaded and run on a graph sized to your box. So:

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

The carve waist is on the roadmap (C1–C4); until it lands, the carve pattern shows the manual path.

See also

• Building a chain — the design method once you've chosen.
• Overview — the verbs and the two scaling classes.