Expose reusable RankQuant bucket-code API for downstream evidence systems

[Fieldnote-Echo]

Context

ordgraph-proto currently carries local RankQuant-shaped bucket-code primitives (CompositionSpec, RankQuantSpec, BucketCode) in src/code.rs. That code is kernel-level ordinal encoding machinery, not graph/event semantics.

OrdGraph should consume declared codes/evidence, not own the implementation of how RankQuant bucket codes are produced or validated.

Proposed ownership

Move/replace this functionality in ordvec, which already owns ordinal/sign quantization and RankQuant.

Acceptance criteria

• Expose a reusable API for deriving/validating bucket codes from ranks or vectors without requiring callers to build a full retrieval index.
• Cover fixed-composition validation: dim, bits, bucket count, per-bucket count, malformed bucket payloads, duplicate/out-of-range ranks.
• Provide stable error types suitable for downstream crates.
• Add parity tests against the current ordgraph-proto/src/code.rs behavior before the local copy is removed.
• Document intended downstream use by evidence systems such as OrdGraph.

Migration note

Once this lands, ordgraph should depend on the ordvec API for code production/parity checks or only ingest already-declared code artifacts.

---

Draft source: /home/ndspence/Downloads/ordgraph_extraction_issue_queue.md

[Fieldnote-Echo]

Reconciled during Ordinal Systems spinout tracking. The reusable RankQuant bucket-code API landed on main via merged PR #226; verified clean origin/main contains commit b46d855. Downstream OrdGraph/OrdinalDB adoption remains follow-up integration work, but the ordvec kernel/API extraction issue is complete.