feat(ollama): add dimensions parameter to OllamaDocumentEmbedder and OllamaTextEmbedder by ShubhamGond105 · Pull Request #3322 · deepset-ai/haystack-core-integrations

ShubhamGond105 · 2026-05-15T19:59:14Z

Related Issues

fixes Add dimensions parameter to OllamaDocumentEmbedder and OllamaTextEmbedder #3295

Proposed Changes:

Added dimensions: int | None = None parameter to both OllamaDocumentEmbedder
and OllamaTextEmbedder.

The Ollama SDK (>= 0.6.2) supports a dimensions parameter on client.embed()
that enables server-side embedding truncation via Matryoshka Representation Learning (MRL).
This parameter is a top-level argument of the request payload and cannot be passed
via generation_kwargs / options.

With this change, users can now do:

embedder = OllamaDocumentEmbedder(model="nomic-embed-text", dimensions=512)

instead of manually truncating and re-normalizing vectors client-side.

How did you test it?

Added unit tests for both OllamaDocumentEmbedder and OllamaTextEmbedder covering:
- dimensions defaults to None
- dimensions is stored on the instance
- dimensions is forwarded to the sync client
- dimensions is forwarded to the async client
- dimensions survives serialization via default_to_dict / default_from_dict
All 20 unit tests pass locally

Notes for the reviewer

dimensions=None preserves existing behaviour — fully backward compatible
Both sync (run) and async (run_async) paths updated in both embedders
Also updated text_embedder.py to use the newer client.embed() API
instead of the deprecated client.embeddings() for consistency

Checklist

I have read the contributors guidelines and the code of conduct
I have updated the related issue with new insights and changes
I added unit tests and updated the docstrings
I've used one of the conventional commit types for my PR title: feat:

ShubhamGond105 · 2026-05-16T15:29:52Z

The chroma test failures are caused by an extra commit (63e66be)
that was accidentally included when I synced my fork.
That commit is not part of my changes.

The ollama test failures are integration tests that require
a running Ollama server which is not available in CI.

My actual changes only touch the ollama embedder files and their tests.

…xtEmbedder

bogdankostic

Hi @ShubhamGond105, thank your for this PR! The linter check is currently failing, please make sure to run the linter as described in our contributing guidelines.

github-actions · 2026-05-19T17:06:09Z

Coverage report (ollama)

Click to see where and how coverage changed

File	Statements	Missing	Coverage	Coverage (new stmts)	Lines missing
integrations/ollama/src/haystack_integrations/components/embedders/ollama
document_embedder.py
text_embedder.py
Project Total

_{This report was generated by python-coverage-comment-action}

ShubhamGond105 · 2026-05-19T17:07:40Z

Hi @bogdankostic ,

I have fixed the lint errors — all imports have been moved to the top level as required. All checks are now passing.

Please let me know if there is anything else to address.

Thank you!

bogdankostic

Thanks for fixing the lint errors! The PR is almost good to go, I just added a couple of minor comments.

bogdankostic · 2026-05-20T11:43:06Z

+            The desired number of dimensions in the embedding output. Only supported by models
+            that implement Matryoshka Representation Learning (MRL), such as nomic-embed-text-v1.5,
+            mxbai-embed-large, and qwen3-embedding. If None (default), the full vector is returned.
+            Requires ollama-python >= 0.6.2.


Let's remove this line from the docstring and instead bump the version of ollama-python in pyproject.toml

bogdankostic · 2026-05-20T11:44:12Z

+        Computes the embeddings of a string using embedding models compatible with the Ollama Library.

-    It uses embedding models compatible with the Ollama Library.
-
-    Usage example:
+        Usage example:
    ```python
-    from haystack_integrations.components.embedders.ollama import OllamaTextEmbedder
+        from haystack_integrations.components.embedders.ollama import OllamaTextEmbedder

-    embedder = OllamaTextEmbedder()
-    result = embedder.run(text="What do llamas say once you have thanked them? No probllama!")
-    print(result['embedding'])
+        embedder = OllamaTextEmbedder()
+        result = embedder.run(text="What do llamas say once you have thanked them? No probllama!")
+        print(result['embedding'])


Let's revert the extra indent introduced here.

bogdankostic · 2026-05-20T11:44:23Z

+            The desired number of dimensions in the embedding output. Only supported by models
+            that implement Matryoshka Representation Learning (MRL), such as nomic-embed-text-v1.5,
+            mxbai-embed-large, and qwen3-embedding. If None (default), the full vector is returned.
+            Requires ollama-python >= 0.6.2.


Let's remove this line from the docstring and instead bump the version of ollama-python in pyproject.toml

bogdankostic · 2026-05-20T11:49:15Z

+        assert embedder.dimensions == 512
+
+    def test_dimensions_passed_to_embed_client(self):
+


Let's remove these empty lines in tests between the test name and test code.

bogdankostic · 2026-05-20T11:51:15Z

+        mock_response = {"embeddings": [[0.1, 0.2, 0.3]]}
+        embedder._async_client.embed = AsyncMock(return_value=mock_response)
+
+        asyncio.run(embedder._embed_batch_async(["hello"], batch_size=32))


For async tests, let's use the @pytest.mark.async decorator, change def to async def and call async methods using await.

… version note, bump ollama version

ShubhamGond105 · 2026-05-20T17:41:59Z

Hi @bogdankostic,

I have addressed all the feedback points:

Removed Requires ollama-python >= 0.6.2. from both docstrings
Bumped ollama>=0.5.0 to ollama>=0.6.2 in pyproject.toml
Reverted the extra indentation in text_embedder.py class docstring
Removed empty lines between test name and test code
Fixed async test to use @pytest.mark.asyncio with async def and await

All checks are now passing. Please take a look when you get a chance!

Thank you!

bogdankostic

Thanks for addressing the feedback @ShubhamGond105! :)

ShubhamGond105 requested a review from a team as a code owner May 15, 2026 19:59

ShubhamGond105 requested review from bogdankostic and removed request for a team May 15, 2026 19:59

github-actions Bot added integration:chroma integration:ollama type:documentation Improvements or additions to documentation labels May 15, 2026

ShubhamGond105 mentioned this pull request May 15, 2026

Add dimensions parameter to OllamaDocumentEmbedder and OllamaTextEmbedder #3295

Closed

ShubhamGond105 force-pushed the feature/ollama-embedder-dimensions branch 2 times, most recently from 5a3b834 to d50a95f Compare May 16, 2026 15:43

feat: add dimensions parameter to OllamaDocumentEmbedder and OllamaTe…

4b0367b

…xtEmbedder

ShubhamGond105 force-pushed the feature/ollama-embedder-dimensions branch from d50a95f to 4b0367b Compare May 16, 2026 18:24

bogdankostic requested changes May 19, 2026

View reviewed changes

bogdankostic self-assigned this May 19, 2026

fix(ollama): move imports to top level to fix lint errors

a961aeb

ShubhamGond105 requested a review from bogdankostic May 20, 2026 06:59

bogdankostic requested changes May 20, 2026

View reviewed changes

fix(ollama): address reviewer feedback - fix docstring indent, remove…

75b0f74

… version note, bump ollama version

ShubhamGond105 requested a review from bogdankostic May 20, 2026 17:57

Bump ollama version to 0.5.4

82c5871

bogdankostic changed the title ~~feat: add dimensions parameter to OllamaDocumentEmbedder and OllamaTextEmbedder~~ feat(ollama): add dimensions parameter to OllamaDocumentEmbedder and OllamaTextEmbedder May 21, 2026

Merge branch 'main' into feature/ollama-embedder-dimensions

a2eeea8

bogdankostic approved these changes May 21, 2026

View reviewed changes

bogdankostic merged commit dbb62df into deepset-ai:main May 21, 2026
13 checks passed

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

feat(ollama): add dimensions parameter to OllamaDocumentEmbedder and OllamaTextEmbedder#3322

feat(ollama): add dimensions parameter to OllamaDocumentEmbedder and OllamaTextEmbedder#3322
bogdankostic merged 5 commits into
deepset-ai:mainfrom
ShubhamGond105:feature/ollama-embedder-dimensions

ShubhamGond105 commented May 15, 2026

Uh oh!

ShubhamGond105 commented May 16, 2026

Uh oh!

bogdankostic left a comment

Uh oh!

github-actions Bot commented May 19, 2026 •

edited

Loading

Uh oh!

ShubhamGond105 commented May 19, 2026

Uh oh!

bogdankostic left a comment

Uh oh!

bogdankostic May 20, 2026

Uh oh!

bogdankostic May 20, 2026

Uh oh!

bogdankostic May 20, 2026

Uh oh!

bogdankostic May 20, 2026

Uh oh!

bogdankostic May 20, 2026

Uh oh!

ShubhamGond105 commented May 20, 2026

Uh oh!

bogdankostic left a comment

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

		assert embedder.dimensions == 512

		def test_dimensions_passed_to_embed_client(self):

Conversation

ShubhamGond105 commented May 15, 2026

Related Issues

Proposed Changes:

How did you test it?

Notes for the reviewer

Checklist

Uh oh!

ShubhamGond105 commented May 16, 2026

Uh oh!

bogdankostic left a comment

Choose a reason for hiding this comment

Uh oh!

github-actions Bot commented May 19, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Coverage report (ollama)

Uh oh!

ShubhamGond105 commented May 19, 2026

Uh oh!

bogdankostic left a comment

Choose a reason for hiding this comment

Uh oh!

bogdankostic May 20, 2026

Choose a reason for hiding this comment

Uh oh!

bogdankostic May 20, 2026

Choose a reason for hiding this comment

Uh oh!

bogdankostic May 20, 2026

Choose a reason for hiding this comment

Uh oh!

bogdankostic May 20, 2026

Choose a reason for hiding this comment

Uh oh!

bogdankostic May 20, 2026

Choose a reason for hiding this comment

Uh oh!

ShubhamGond105 commented May 20, 2026

Uh oh!

bogdankostic left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

github-actions Bot commented May 19, 2026 •

edited

Loading