Skip to content

chore(release): v0.2.0b26 — OCI embedding output_dimensions + input_type fix #294

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
fede-kamel merged 1 commit into from
May 30, 2026
Merged

chore(release): v0.2.0b26 — OCI embedding output_dimensions + input_type fix #294

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
31 changes: 30 additions & 1 deletion CHANGELOG.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -8,6 +8,34 @@ policy.

## [Unreleased]

## [0.2.0b26] - 2026-05-30

### Fixed — OCI embedding dimension now deterministic (`output_dimensions`)

`OCIEmbeddings` left the embedding dimension to OCI's implicit server-side
default. For `cohere.embed-v4.0` — which is Matryoshka and returns exactly
the requested `output_dimensions` (256/512/1024/1536), defaulting to 1536 —
a vector column indexed at one dimension (e.g. by an ingestion pipeline or an
earlier default) and queried at another raised `ORA-51803` ("Vector dimension
count must match …") at search time. Verified against live OCI: `input_type`
(`SEARCH_QUERY` vs `SEARCH_DOCUMENT`) does **not** change the output dimension
— it's `output_dimensions` (closes #292).

- Added `OCIEmbeddingConfig.output_dimensions` (default `None`), forwarded to
`EmbedTextDetails` on every embed path via a shared `_build_embed_details`
helper and **only when set**, so non-Matryoshka models (e.g. embed-v3.0,
fixed 1024) never receive the field.
- `config.dimension` now prefers `output_dimensions`, so the vector store
sizes its column to match the embeddings being produced.

### Fixed — `OCIEmbeddings` honors configured `input_type`

`embed_query()` hardcoded `input_type="SEARCH_QUERY"` (ignoring config) and
`embed_documents()` hardcoded `SEARCH_DOCUMENT`. Added
`OCIEmbeddingConfig.query_input_type` (default `SEARCH_QUERY`) and made both
methods read from config; removed a dead `original_type` local and an
incorrect "frozen config" comment in `embed_query`.

## [0.2.0b25] - 2026-05-30

### Fixed — explicit `wallet_password=""` dropped for auto-login wallets
Expand Down Expand Up @@ -1617,7 +1645,8 @@ First internal-review version. Core shape established:
- Observability: OpenTelemetry spans and metrics, structured logging.
- Streaming: `AsyncIterator[LocusEvent]`, SSE, console handler.

[Unreleased]: https://github.com/oracle-samples/locus/compare/v0.2.0b25...main
[Unreleased]: https://github.com/oracle-samples/locus/compare/v0.2.0b26...main
[0.2.0b26]: https://github.com/oracle-samples/locus/compare/v0.2.0b25...v0.2.0b26
[0.2.0b25]: https://github.com/oracle-samples/locus/compare/v0.2.0b24...v0.2.0b25
[0.2.0b10]: https://github.com/oracle-samples/locus/compare/v0.2.0b9...v0.2.0b10
[0.2.0b9]: https://github.com/oracle-samples/locus/compare/v0.2.0b7...v0.2.0b9
Expand Down
2 changes: 1 addition & 1 deletion pyproject.toml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -4,7 +4,7 @@ build-backend = "hatchling.build"

[project]
name = "locus-sdk"
version = "0.2.0b25"
version = "0.2.0b26"
description = "Multi-agent workflows for Python — stream them, branch them, pause for a human, resume next week. Built on Oracle Generative AI."
readme = "README.md"
license = "UPL-1.0"
Expand Down