feat: support deferred embedding via update/RecordPatch (#88)#93
Merged
beinan merged 1 commit intoJun 15, 2026
Merged
Conversation
) Append raw records first and enrich them with embeddings later. Adds an `embedding` field to the record patch path so a vector can be attached or replaced by id or external_id after the original write. - core: RecordPatch.embedding, applied in update_visible_record - api: RecordPatchDto.embedding; copied in core + server patch_from_dto - python: embedding param on PyO3 update and Context.update() - Rust client inherits the field via the UpdateRecordRequest JSON body A record without an embedding is durably stored but excluded from vector search until enriched, since search skips null-embedding records. Tests: Rust core round-trip (by id + external_id) and Python e2e (raw-first -> enrich, incl. add_many bulk). README documents the pattern. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
beinan
approved these changes
Jun 15, 2026
beinan
pushed a commit
that referenced
this pull request
Jun 15, 2026
…) (#92) ## Summary Brings the REST `search` and `list` endpoints to parity with the core and Python APIs by adding `filters`, `include_expired`, and `include_retired`, reusing the existing `RecordFilters` JSON semantics and `LifecycleQueryOptions` — mirroring the plumbing already used by REST `retrieve`. No new query language, no SQL, and no change to default visibility for expired/retired/superseded records. Closes #89. > ###⚠️ Stacked on #93 — review/merge that first > This branch is stacked on top of #93 (deferred embedding, #88). Until #93 lands on `main`, the diff below shows **both** commits (`b336bdc` is #93's work; `1aae34a` is this PR's #89 work). Once #93 merges I'll rebase this branch onto `main` so the diff is #89-only. Please review the second commit for #89. ## What changed (#89) - **`lance-context-api`** — `SearchRequest` gains `filters` / `include_expired` / `include_retired` (legacy `{query, limit}` payloads still deserialize). The `ContextStoreApi` trait now takes `search(&SearchRequest)` and `list(limit, offset, filters, include_expired, include_retired)` (breaking change to the trait; acceptable pre-1.0). - **`lance-context-core`** — trait `search`/`list` route through the existing `search_filtered_with_options` / `list_filtered_with_options`. - **`lance-context-server`** — `search` applies filters + lifecycle; `list_records` accepts `filters` (URL JSON string) plus `include_expired` / `include_retired` query params, matching the existing `related_records` style. - **`lance-context` (unified) + `lance-context-client`** — dispatch and `RemoteContextStore` updated; the low-level client `list_records` sends the new query params. Adds a `serde_json` dependency to the `lance-context` crate (the trait now exposes `serde_json::Value`). ## Tests - 8 new server route tests — `search` and `list`, each covering: metadata filter, built-in field filter, `include_expired`, `include_retired` (superseded), and `include_relationships` toggling. - A `list` invalid-filters-JSON rejection test. - 3 `lance-context-api` serde tests, including backward compatibility for the pre-#89 `SearchRequest` payload. ## Verification - `cargo fmt --all -- --check` - `cargo clippy --workspace --all-targets -- -D warnings` - `cargo test`: core (44), server (14, incl. 8 new), api (3) — all green. - Python untouched (already had filter/lifecycle parity). 🤖 Generated with [Claude Code](https://claude.com/claude-code) --------- Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
Lets callers append raw records first and enrich them with embeddings later — the common pattern for bulk ingestion where source chunks must be persisted immediately but embeddings are computed asynchronously (large documents, rate-limited or remote embedding providers).
Adds an
embeddingfield to the record patch path so a vector can be attached or replaced byidorexternal_idafter the original write. A record without an embedding is durably stored but excluded from vector search until enriched (searchskips null-embedding records).Closes #88.
What changed
RecordPatch.embedding, applied inupdate_visible_record.RecordPatchDto.embedding; copied through in both the core and serverpatch_from_dto.embeddingparameter on the PyO3updateandContext.update().UpdateRecordRequestJSON body.Tests
idand byexternal_id).add_manybulk ingestion.🤖 Generated with Claude Code