Knowledge check Choose an answer to see why it is right or wrong.
JavaScript is off, so every explanation is shown at once.
01 Why does Cogitave Core use one typed property graph instead of a docs CMS, an IDP catalog, a CMDB, and a vector store as separate systems?
Because a store per domain is the default industry shape, and it guarantees drift and duplicated identity - one graph lets docs, IDP, governance, infra, and agent knowledge all be projections of the same source instead of copies that can disagree. Correct. This is the problem statement ADR-0001 opens with: per-domain stores are the default, and they guarantee drift because nothing forces them to agree once they diverge. Because a single graph is cheaper to host than four separate databases. Incorrect, and not the stated reason. The decision drivers are one identity, one query layer for humans and agents, and a traversable provenance model for certification - not hosting cost. Because property graphs are the only data model that can store text. Incorrect. Any of the considered stores can hold text. The distinguishing property is that nodes and edges carry properties directly, which matches faceting and ranking without join gymnastics. 02 ADR-0001 considered a relational schema, an RDF/triple store, a federated per-domain store, and a single labeled property graph (LPG). Why was the LPG chosen?
It is now an ISO standard (GQL) with a mature open lineage (openCypher), and its nodes/edges carrying properties directly matches faceting and ranking needs - while relational buries the graph being traversed, RDF is higher-ceremony with weaker property ergonomics, and federation re-introduces the drift the decision exists to remove. Correct. That is the decision outcome and the stated reasons each alternative was rejected. Because relational databases cannot represent relationships at all. Incorrect. Relational databases represent relationships via foreign keys; the ADR's objection is that this buries the graph you actually want to traverse, not that it is impossible. Because RDF and SPARQL are deprecated technologies with no active users. Incorrect and not the ADR's claim. RDF was rejected for being higher-ceremony with weaker property ergonomics for this use case, not for being unused. 03 Cogitave Core gives every node both a uid and a contentHash. What is each one for?
uid is immutable identity, globally unique across the whole estate, and never changes even if the URL does; contentHash is a content-addressed digest of the canonicalized payload that changes whenever the content changes, so unchanged content is deduplicated and stored once.Correct. Identity and version are deliberately orthogonal: uid is what edges and @uid references point to; contentHash is what determines whether two materializations are byte-identical. They are two names for the same thing, kept for backward compatibility. Incorrect. They answer different questions - which node is this, versus which version of its content is this - and are tracked separately for exactly that reason. uid changes on every edit; contentHash is the permanent identifier.Incorrect - this is backwards. The uid is what stays permanent; the contentHash is what changes when the content does. 04 Cogitave authors Core's graph model against GQL/openCypher, an external ISO standard, but the serving engine is described as "full-scratch." What does that combination mean?
The property-graph model is used as a vendor-neutral spec to design against, so concepts and vocabulary transfer and interoperate - but the actual query and storage engine that serves it is built in-house rather than running an off-the-shelf graph database. Correct. Off-the-shelf systems are treated as reference implementations and specs, not dependencies - the standard is authored against, not installed. It means Cogitave runs an unmodified openCypher database in production. Incorrect. The standard is the spec being authored against; the serving engine itself - the single Rust binary co-locating BM25, HNSW, and the graph - is built from scratch. It means the GQL standard is irrelevant to how Cogitave Core actually works. Incorrect. The standard is exactly why the LPG model was chosen over the alternatives - it is load-bearing for the decision, even though the engine is not a copy of any existing implementation.