Skip to content

docs(rfc): define artifact search projections and join-free retrieval - #1803

Open
frf12 wants to merge 1 commit into
oceanbase:masterfrom
frf12:codex/memory-artifact-rfc
Open

frf12 wants to merge 1 commit into
oceanbase:masterfrom
frf12:codex/memory-artifact-rfc

Conversation

@frf12

@frf12 frf12 commented Sep 29, 2026 •

Copy link
Copy Markdown
Member

Which issue or RFC does this PR close?

RFC #1803 proposal; no issue is closed by this documentation PR.

Amends RFCs 0014, 0051, 0080, and 1417. Related contracts include RFCs 1396, 1467, 1549, and 1652. Migration follows #1771.

Rationale for this change

In OceanBase, joins between full-text retrieval and other business tables degrade the full-text index path. Distance-ordered vector scans participating in merge joins also cause nearest-neighbor loss. Artifact search needs an explicit boundary that keeps those business associations out of candidate retrieval.

Families differ in search granularity, payload size, vector configuration, and tag maintenance. A common retrieval contract allows search wide tables or separate current projections without requiring one physical table for every Family.

What changes are included in this PR?

  • English and Chinese RFCs covering the problem, proposed query boundary, core contracts, tradeoffs, alternatives, and open questions.
  • No business-table joins in full-text or vector retrieval statements for covered backends, including correlated business lookups and outer joins in the same retrieval statement.
  • Mandatory compliance for OceanBase; whether other backends must comply remains open, with benefits and costs of both options documented.
  • Backend tradeoffs compare a single shared-model change and fewer maintenance branches with backend-specific optimization, considering the embedded local use of SQLite and seekDB. Architecture for divergent storage models remains a deferred question, not a prerequisite.
  • A common standard with independently designed and delivered Family adaptations; simultaneous completion is not required.
  • Recommended Family-owned wide tables or separate current projections, with independent choices for full text, vectors, topics, and chunks.
  • Eligibility-aware candidate selection that preserves Scope, tag, lifecycle, and access semantics.
  • Separate batch content reads by exact reference after candidate selection, under the same consistent state and without deferring eligibility filters to content completion.
  • Atomic projection maintenance and preserved history; schema changes and projection rebuilding follow the unified migration process in docs(rfc): define unified versioned database migrations #1771.

Are there any user-facing changes?

Documentation only. The proposal changes future search storage and upgrade requirements while preserving public result content, exact references, filters, authorization, and scoring contracts. Detailed table schemas, SQL, and migration procedures are outside this PR.

How was this change tested?

  • Checked relative Markdown links, the nine RFC sections, UTF-8 text, and whitespace.
  • Reviewed English and Chinese content for consistent retrieval, filtering, and content-read boundaries.
  • git diff --check and repository commit hooks.
  • Checked that the PR contains only the two RFC documents.

No application tests or database benchmarks were run for this documentation-only change.

AI usage statement

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

@frf12 frf12 changed the title docs(rfc): propose individual memory artifacts and current search projections docs(rfc): define per-family artifact search wide tables Sep 30, 2026
@frf12
frf12 force-pushed the codex/memory-artifact-rfc branch 3 times, most recently from 7684a26 to 70f80f3 Compare October 3, 2026 18:46
@frf12 frf12 changed the title docs(rfc): define per-family artifact search wide tables docs(rfc): define artifact search projections and join-free retrieval Oct 3, 2026
@frf12
frf12 force-pushed the codex/memory-artifact-rfc branch 2 times, most recently from 9082206 to 80e399f Compare October 4, 2026 16:11
@frf12
frf12 force-pushed the codex/memory-artifact-rfc branch from 80e399f to e2fcc87 Compare October 4, 2026 16:48

This branch has not been deployed

No deployments
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant