Visitar URL original
Document the vector SQL options and add IndexType.LSM_SPARSE_VECTOR by tae898 · Pull Request #384 · humemai/arcadedb-embedded-python · GitHub
Skip to content

Document the vector SQL options and add IndexType.LSM_SPARSE_VECTOR - #384

Merged
tae898 merged 2 commits into
mainfrom
vector-api-382
Oct 21, 2026
Merged

tae898 merged 2 commits into
mainfrom
vector-api-382

Conversation

@tae898

@tae898 tae898 commented Oct 21, 2026

Copy link
Copy Markdown

Closes #382.

Following the bindings' DSL-first design (docs/api-access-methods.md), the vector features the engine has and our docs did not show are documented as SQL, with tests that run the documented statements and assert the answers; no new wrapper API.

  • docs/guide/vectors.md: graph build settings in CREATE INDEX ... LSM_VECTOR METADATA (neighborOverflowFactor, alphaDiversityRelaxation, addHierarchy), the search options of vector.neighbors (efSearch, filter, where from 26.11.1, groupBy, groupSize, maxDistance), and the sparse index (LSM_SPARSE_VECTOR, vector.sparseNeighbors, minScore); api/vector.md, api/schema.md, api/database.md and java-api-coverage.md follow.
  • IndexType.LSM_SPARSE_VECTOR: the enum lacked the engine's sparse index type, so schema.create_index(..., index_type=...) could not name it.
  • create_vector_index docstring: its add_hierarchy=True default is a deliberate choice of these bindings, while the engine (and plain SQL) defaults to a flat graph.
  • tests/test_vector_sql.py: 13 tests (search options, metadata reaching the index, the two addHierarchy defaults, sparse minScore, the enum member); the two where tests skip before 26.11.1 and a third asserts the engine's "Unknown option" error there.

Checked locally against the 26.10.2 engine: the vector SQL, parameter and sparse test files (61 passed, 2 skipped) and the schema tests (59 passed), pre-commit on the changed files, the strict docs build. The where tests run once the bindings carry 26.11.1.

🤖 Generated with Claude Code

tae898 and others added 2 commits October 21, 2026 12:54
…382)

The vector guide and API pages now show the vector.neighbors option map (efSearch, filter, where since 26.11.1, groupBy, groupSize, maxDistance), the LSM_VECTOR build settings neighborOverflowFactor, alphaDiversityRelaxation and addHierarchy, and the sparse index options (weightQuantization, minScore). The create_vector_index docstring and docs say that add_hierarchy=True is a deliberate default while the engine defaults to false. IndexType gains the missing LSM_SPARSE_VECTOR member. New tests run the documented SQL and assert the answers; the where tests skip on engines older than 26.11.1.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…nosec on test-owned SQL literals (#382)

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
@tae898
tae898 merged commit 2ce5297 into main Oct 21, 2026
42 of 44 checks passed
@tae898
tae898 deleted the vector-api-382 branch October 21, 2026 06:54
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.

Vector API: find_nearest filters and groups, build options, sparse index helper

1 participant