[Zxf-xufeng]

Which issue or RFC does this PR close?

Design proposal based on the text material import research. No implementation issue is closed.

• English RFC
• 中文 RFC
• Dependencies: Source plugins feat(source): add source plugin framework and some examples #2210 at 6ef5c494d701c355686806b6b1042d5729267690, Artifact plugins feat(artifact): add managed plugin framework and some example #2246 at 9fe2e68220a6e0aa8ebe4b78900cd12421db5cc3.
• HTTP identity and response conventions follow RFC docs(rfc): define source and artifact REST APIs #1437.

Rationale for this change

Importing documents needs parsing, private previews and explicit acceptance in an existing Scope. Reuse Source plugins for collection and Artifact plugins for ImportedContent generation, revisions and reads, while retaining the import-specific workflow and a durable complete-document handoff.

Confirmation atomically saves document Sources, a frozen commit manifest, publication intent and an acceptance receipt (T1). Artifact generation separately publishes the Revision, Head, lineage, declared indexes and completion record (T2). Evidence may therefore be visible before the Artifact exists; 202 is acceptance, not completed import.

What changes are included in this PR?

• Matching Chinese and English RFCs define family=importedcontent, source_kind=content, mandatory existing Scope selection, 52 proposed parser profiles and four diagrams per locale.
• A reuse matrix identifies existing Source/Artifact capabilities, import-specific responsibilities and required framework extensions. The plugin PRs do not already provide complete-document input, transactional publication acknowledgement or the collect-to-import adapter.
• Document-scoped input validates exact ordered evidence across journal windows, with bounded part reads and separate input/output budgets. A normal Source window is not treated as a complete document.
• Seventeen proposed import operations and four reused Artifact read operations document parameters, meanings, limits, request examples, state, errors and recovery. New evidence acceptance returns 202; published-current reuse returns 200; commit status supplies the exact published revision.
• Reuse Artifact storage and plugin lifecycle/index infrastructure. Specify nine import-specific logical persistence responsibilities, including immutable commit manifests and publication state, without an ImportedContent main/version table or FileImportJob business resource.
• Define per-document publication reservation, Head CAS, idempotent retries, permission rechecks, TTL-independent accepted commits and checkpoint advancement only after verified publication.
• The explanatory long infographic is a separate deliverable and is not embedded in the RFC.

Are there any user-facing changes?

Documentation only. No parsers, runtime code, OpenAPI output or database migrations are implemented by this PR. The UI distinguishes private preview, evidence accepted, publication pending/failed and completed import. ImportedContent is assembled deterministically without an LLM. Source acceptance/publication does not imply Memory generation or full-text indexing.

The pinned plugin PRs are dependency baselines, not claims of merged availability. Publication readiness must remain unavailable until the specified document-input and transaction-completion extensions are implemented and validated.

How was this change tested?

• git diff --check.
• SKIP=ty-check /Users/zhengxiaofeng/Desktop/code/powercontext/.venv/bin/prek run --files docs/zh/rfcs/2073-text-material-import.md docs/en/rfcs/2073-text-material-import.md: applicable document hooks passed; unrelated always-run type checking was skipped for this documentation-only change.
• Bilingual consistency checks: matching 21-row HTTP inventories (17 proposed import operations and four reused Artifact reads), all 52 original parser profiles retained, 12 valid JSON blocks per language, local documentation links resolve and no obsolete ImportCommitReceipt remains.
• Both RFCs compile with the installed MDX Markdown compiler and GFM parser: 30 tables, 310 rows and four Mermaid code blocks per language. This checks Markdown compilation, not a rendered Mermaid/browser or full website build.
• No runtime, parser, migration, seekdb or OceanBase execution is claimed. The full website build was not run.
• The long infographic was generated using the built-in image-generation tool and visually checked against the design.

AI usage statement

OpenAI Codex assisted with plugin-source inspection, architecture, bilingual documentation, API-contract consistency checks and the separate overview infographic. No implementation changes are included.