docs(rfc): introduce automatic memory as independent artifacts

[frf12]

Which issue or RFC does this PR close?

RFC #1809 proposal; no issue is closed by this documentation PR.

Depends on #1803, the public per-Family search wide-table contract. Review and establish that foundation first; this RFC conforms to it.

Amends RFCs 0014, 0019, 1345, 1652, and 1718. Related contracts include RFCs 1417 and 1549.

Rationale for this change

Legacy Memory versions a Scope-level collection containing a full manifest of entry-version references. It already supports entry revision and incrementally maintained projections, but each effective collection change still stores a full directory and shares a collection CAS. Extraction supplies all active entries to the model instead of retrieving a bounded set of related memories after candidate generation.

Representing each logical memory as an independent Artifact aligns identity and history with Scope and removes new full-manifest snapshots.

What changes are included in this PR?

• Bilingual Automatic Memory RFC using the repository's nine-section template, plus the retained detailed Chinese implementation design.
• New canonical automatic-memory family; independent identity, revisions, evidence, active/inactive/retired lifecycle, and explicit restoration/compensation.
• Bounded candidate generation and related retrieval, create/revise/merge/noop/conflict actions, review rules, narrowly scoped preauthorization, CAS, idempotency, Scope generation checks, and atomic capacity accounting.
• A Family-owned current search wide table conforming to docs(rfc): define per-family artifact search wide tables #1803; no public-head content expansion or separate competing retrieval contract.
• Centralized, operation-aware API compatibility for the old memory name. Authorization and execution use the same resolved identity; exact old references and collection contracts are not blindly rewritten.
• A versioned offline migration tool that replays all old history, preserves compact semantics and evidence, records interval aliases, and preserves tags, authorization, Source progress, and accepted work.
• Maintenance-window migration with backups, resumable checkpoints, readiness gates, validation, and an explicit downgrade boundary after new writes. No runtime dual writes or online backfill.
• Future commercial online upgrades may use a mandatory barrier release; this is outside the initial implementation scope.

The two RFC implementations may ship together and migrate during one maintenance window: convert authoritative Memory history before building its search projection under #1803.

Are there any user-facing changes?

Documentation only. No Runtime, API, schema, or migration implementation changes in this PR.

The proposal is a breaking identity/content/API change for future releases. Legacy history and exact citations remain readable. Old collection IDs, ETags, manifests, and pagination cursors require explicitly supported adapters or a clear upgrade-required error. Name compatibility does not imply unchanged response structures.

How was this change tested?

• Checked all relative Markdown links in the three delivered documents.
• Checked the nine RFC sections, balanced code fences, UTF-8 content, and trailing whitespace.
• git diff --cached --check.
• Standard repository commit hooks.

No application tests or database benchmarks were run for this documentation-only proposal. Migration rehearsals, concurrency behavior, and real-backend search acceptance are requirements for implementation.

AI usage statement

Prepared with OpenAI Codex (GPT-6), including document drafting and source-grounded design review, under maintainer direction.