chore(terminology): adopt Granite Switch / Mellea glossary across Epic #929 and repo #1192

Parent epic: #929

Background

The team agreed a canonical glossary for Granite Switch / Mellea naming. Most of Epic #929's issue bodies, code comments, and docs predate that agreement — they use "intrinsic", "intrinsic function", "activated-LoRA", and "Project Granite Switch" throughout.

The timing matters: Phase 1 starts as soon as #1157 and #1158 land. If we don't update the issue bodies first, contributors picking up those issues will carry old vocabulary into commits, PR descriptions, and reviews. Fixing it now costs one focused pass; fixing N stale artefacts later costs much more.

This issue is the pass.

What we're committing to the repo

The agreed glossary needs a durable home. This issue adds docs/glossary.md and sweeps AGENTS.md to use the preferred terms. No logic changes; no Python symbol renames (those are Phase 1 and Phase 4 work).

Glossary

Note on aLoRA wording: avoid "cache rewrite" — the right framing is cache hits/misses. aLoRA generates more cache hits (fewer new KV entries) than separate LoRA loads, which reduces latency and memory pressure. This benefit extends to non-adapter requests too, since less cache eviction means plain model generation can also hit the cache more often.

Terms to retire:

References to include in the glossary

The glossary should end with a short references section pointing to authoritative sources. At minimum:

• Granite Switch repo / docs — link TBD; confirm with the team
• aLoRA cache efficiency demo — visual illustration of the cache-hit argument

Code changes (PR)

The two Phase 0 PRs (#1157 / #1158) introduced new symbols under mellea/backends/adapters/ using "adapter" — that's already clean. The gap is prose: comments and docstrings in those files still say "intrinsic" in places.

Files to sweep:

• mellea/backends/adapters/adapter.py
• mellea/backends/adapters/_core.py
• mellea/backends/adapters/catalog.py
• mellea/backends/adapters/capabilities.py
• mellea/stdlib/components/adapter_based_component/__init__.py

Constraint: do not rename Python symbols here (IntrinsicAdapter, call_intrinsic, mellea/stdlib/components/intrinsic/). Those renames live behind Phase 1 shims (#1136) and the Phase 4 cleanup (#1144).

Issue body edits (administrative, no PR)

Sweep the body of each open #929-series issue and replace deprecated terms with preferred ones. Old class names that appear as context ("today's IntrinsicAdapter") are fine to leave — they're describing what the issue is replacing, not advocating for the term. Update narrative prose only.

Issues:

• Epic: Fix Adapter Function Lifecycle & Consistency in Mellea #929 (epic — title and body)
• refactor(intrinsics): internal migration with shims — old classes inherit from Adapter; call_intrinsic rewritten (Epic #929 Phase 1) #1136 (Phase 1.A bottleneck)
• refactor(intrinsics): rag.py whole-file migration to new Adapter types (Epic #929 Phase 1) #1137 (Phase 1.B)
• refactor(intrinsics): requirement_check migration — loud schema-mismatch, no more silent False (Epic #929 Phase 1) #1138 (Phase 1.C)
• refactor(intrinsics): guardian.py whole-file migration + documents= kw-only + auto-context discovery (Epic #929 Phase 1) #1139 (Phase 1.D)
• refactor(backends): AdapterMixin verb rename/narrow + resolve_model_options + IntrinsicMetricsPlugin (Epic #929 Phase 2) #1140 (Phase 2.1)
• feat(backends): LocalFileBinding implements verbs (PEFT/aLoRA path) + from_catalog() (Epic #929 Phase 2) #1141 (Phase 2.2)
• feat(backends): EmbeddedBinding implements apply_activation (Granite Switch path); remove render_controls + set_request_adapter (Epic #929 Phase 2) #1142 (Phase 2.3)
• refactor(intrinsics): remove deprecation shims; rewrite intrinsics_and_adapters.md; write 3 tutorials (Epic #929 Phase 4) #1144 (Phase 4)
• refactor(intrinsics): decouple capability registry from catalog entry names (Epic #929 Phase 0) #1186 (Phase 0.3)
• design: adapter output-schema versioning (deferred from #929) #1111 (deferred versioning)

Identify any additional work

The sweep may turn up terminology that doesn't fit neatly into "prose comment" or "issue body" — e.g. a tutorial that needs updating, a label that should be renamed, a test fixture name that leaks the old vocabulary. Track any findings here as a comment rather than expanding this PR's scope. Items that are non-trivial become follow-up issues; obvious one-liners can go in the same PR.

Out of scope

• Renaming Python modules, classes, or public functions — Phase 1 (refactor(intrinsics): internal migration with shims — old classes inherit from Adapter; call_intrinsic rewritten (Epic #929 Phase 1) #1136) and Phase 4 (refactor(intrinsics): remove deprecation shims; rewrite intrinsics_and_adapters.md; write 3 tutorials (Epic #929 Phase 4) #1144)
• Renaming the area/intrinsics label — separate housekeeping if the team wants it
• Full rewrite of intrinsics_and_adapters.md — that's refactor(intrinsics): remove deprecation shims; rewrite intrinsics_and_adapters.md; write 3 tutorials (Epic #929 Phase 4) #1144

Done when

• docs/glossary.md exists, renders, and includes authoritative links
• aLoRA entry uses cache hit/miss framing, not "cache rewrite"
• AGENTS.md uses preferred terms throughout
• No "intrinsic" prose comments remain in the five Phase 0 files above
• All eleven listed issues have updated bodies
• Any extra findings are noted as comments here (or tracked as new issues)
• No Python symbols renamed; CI green