Module cml_context 

Rust-native CML context graph + COF HTML+RDFa for the hypermedia Library.

Goal: every ingested document (legislation, notes, policies, contracts) gets a TEXT → CONCEPT → LOGIC layer that is honest about provenance:

• All machine output is cml:Proposed (never cml:Attested).
• Structure + signal extraction is deterministic (no Python, no Ollama required).
• Deontic cues compile into real NQuin norms (OP_OBLIGATE / OP_PERMIT / OP_FORBID).
• Privacy / GDPR-family / rights / temporal / cross-ref signals become topics + graph edges.
• COF (cof_html) serialises the same graph as HTML+RDFa for agent windows, with token-bounded segments (index + body packs) when a document is large.

Optional local LLM enrichment can attach later on the same ContextUnit surface; the product path stays pure Rust.

Structs

CmlConcept

A proposed concept in the CML layer.

CmlContextGraph

Full context graph for a document or provision.

CofPackage

A complete COF package: index + ordered body segments.

CofSegment

One COF HTML segment (self-contained HTML document, RDFa-complete).

ContextUnit

One structural unit (section, heading block, or paragraph cluster).

SignalHit

A named signal hit (privacy family, rights, etc.).

Enums

CofStyle

DeonticClass

Proposed deontic class for a provision / paragraph.

PrivacySignal

Privacy / data-protection family (GDPR-like and cognates).

Constants

COF_PROFILE

Official COF HTML+RDFa profile (must match cof.n3 / legis2cml).

DEFAULT_SEGMENT_MAX_CHARS

Default agent body-segment budget (~6–8k tokens at ~4 chars/token; leave headroom).

MEDIA_TYPE_COF

Functions

build_cof_package

Build a full COF package (index + body segments) for agent token optimisation.

build_document_context

Build a multi-unit CML context graph for a document.

build_unit_context

Build CML context for a single unit (provision or paragraph).

classify_deontic

Classify deontic force from English legislative / policy phrasing.

extract_cross_refs

Cross-references to other provisions / instruments.

extract_privacy_signals

Extract GDPR-like / privacy-family signals.

extract_rights_signals

Human-rights / civil-rights cues (broader than privacy).

extract_temporal_signals

Temporal / LTL-ish cues (commencement, deadlines).

pack_units_into_segments

Pack units into COF segments under a character budget (section-aligned).

render_cof_document

Single-document COF (no multi-segment) — still valid html-rdfa-1.

render_unit_fragment

Render one unit as a COF section fragment (no document chrome).

units_from_headings

Split plain text into heading-aware units (Markdown # / ALL-CAPS lines / blank-line blocks).

units_from_paragraphs

Fallback: split on blank lines into paragraph units (max 64).