Skip to content

MODULES.md

Latest commit

 

History

History
87 lines (62 loc) · 4.22 KB

MODULES.md

File metadata and controls

87 lines (62 loc) · 4.22 KB
Raw

MODULES.md — ucns_kit module registry

Source of truth for the ucns_kit dependency skeleton. Modules divide into Category 1 (frame-independent) and Category 2 (frame-dependent protocols).

Category 1 — frame-independent utilities

Module Responsibility Status
coherence_primes.py Sequence registry: nth(k), is_coherence_prime(n), sequence_up_to(limit). Recursive definition fully pinned (see below). ✅ implemented
encoder.py Archived lexical encoder skeleton; no current UCNS mapping authority. ⛔ fail-closed
pool.py UCNS object intern table. Encode-once-refer-many. Frame-independent. ✅ implemented
disk_flip.py disk_flip(obj) → UCNSObject. Provisional: swap n_dec/n_min. Spec law not yet verified against multiplication rule. ✅ implemented (provisional)
theta_gate.py gate(obj, capability) → UCNSObject. Class-only view when capability ungrouped; full object when granted. ✅ implemented (taxonomy pending)
audit.py UCNS-keyed audit records for S9-sentinel-style logging. Append-only, in-memory v0. ✅ implemented
orchestrator.py Six-step pipeline orchestrator. Operates on Category 2 protocols only; no frame-specific code. ✅ implemented

Category 2 — frame-dependent protocols

Module Responsibility Status
protocols.py RingState, PropagationRule, CoherenceMeasure, RewardMechanism, Serializer. Protocol interfaces only; no implementations. ✅ implemented

Coherence prime definition (fully pinned)

Canonical source: interdependent_lib.coherence_primes (The-Interdependency/interdependent-lib) is the single source of truth for this sequence — its shared test oracle pins the early sequence, the scarcity count table, and the p=4373 recursive-ancestry regression. coherence_primes.py in this kit is the behaviourally identical in-tree copy; keep the two in lockstep.

Base: C₀ = {3, 5, 7}

A prime p ∈ C iff:

  1. (p - 1) % 4 == 0 — p ≡ 1 mod 4
  2. (p - 1) // 4 is squarefree — no repeated prime factors
  3. every prime factor of (p - 1) // 4 is already in C

Verified sequence: 3, 5, 7, 13, 29, 53, 61, 157, 349, 421, …

Rejection examples:

  • 17: kernel = (17−1)/4 = 4 = 2² — not squarefree
  • 19: (19−1) % 4 = 2 ≠ 0 — fails condition 1

PCNA ring topology note

PCNA ring sizes {53, 29, 19, 17}:

  • Φ=53 and Ψ=29 are coherence primes ✅
  • Mem-L=19 and Mem-S=17 are not coherence primes: 17 fails squarefree (kernel 4 = 2²); 19 fails divisibility ((19−1)/4 = 4.5)

These two rings predate the coherence-prime constraint. Ring-size slice selection (hmmm 2) will need to address this discrepancy.

Frame × Reward compatibility matrix

R1 (bandit over UCNS arms) R2 (discrete anchor promotion) R3 (separate numerical control plane)
Frame A (tensor of UCNS objects) ✅ valid ✅ valid ✅ valid
Frame B (ring as one UCNS object) ✅ valid ✅ valid ✅ valid
Frame C (parallel tensor + UCNS audit) ✅ valid ❌ INVALID ✅ valid

Frame C × R2 is invalid: Frame C's primary state has no anchor structure to promote. Enforced nominally in orchestrator.py::_check_compatibility — becomes executable once Frame/Reward implementations are identifiable by type (see hmmm below).

Open hmmms (do not resolve without Erin)

  1. Frame: A (tensor of UCNS objects), B (ring as one UCNS object), C (parallel tensor + UCNS audit).
  2. Ring-size slice: which subset of coherence primes maps to which ring.
  3. Anchor payload depth: depth-2 vs arbitrary recursive.
  4. Reward mechanism: R1/R2/R3 selection.
  5. Disk-flip verification: spec law not verified against ucns_v04.multiply.
  6. Coresidence: in-place rename of pcna.py etc, or parallel pcna_ucns.py files.
  7. Canonical UCNSObject key: pool._key breaks at recursive anchor payload depth > 0.
  8. Capability taxonomy: theta_gate allowlist not yet defined.
  9. Entry-to-UCNSObject mapping: encoder._entry_to_ucns provisional.
  10. Lexical mapping: a lawful mapping into the current producer-owned UCNS API remains unresolved.