feat: add Ollama text vectorizer

[s-agbede]

Summary

Adds OllamaTextVectorizer support for local Ollama embedding models.

This PR includes Ollama optional dependency wiring, a sync/async OllamaTextVectorizer implementation, vectorizer registry/from-dict support for type: "ollama", mocked unit tests, and opt-in real Ollama integration tests guarded by REDISVL_TEST_OLLAMA=1.

Usage

from redisvl.utils.vectorize import OllamaTextVectorizer

vectorizer = OllamaTextVectorizer(model="nomic-embed-text")
embedding = vectorizer.embed("hello world")

To run the real Ollama integration tests locally:

ollama pull nomic-embed-text
REDISVL_TEST_OLLAMA=1 uv run pytest tests/integration/test_ollama_vectorizer_integration.py

Testing:

uv run pytest tests/unit/test_ollama_vectorizer.py tests/integration/test_ollama_vectorizer_integration.py -q
REDISVL_TEST_OLLAMA=1 uv run pytest tests/integration/test_ollama_vectorizer_integration.py -q
uv run black --check ./redisvl ./tests
uv run isort ./redisvl ./tests --check-only --profile black
uv lock --check
uv run mypy redisvl
make test

Full repo test result:

---

Note

Low Risk
Additive feature behind an optional extra; no changes to existing vectorizers or core Redis paths. Main operational risk is init-time calls to a local Ollama server when users instantiate the vectorizer.

Overview
Adds local Ollama as a first-class text embedding provider in RedisVL.

A new OllamaTextVectorizer talks to a running Ollama server (sync/async embed / embed_many, optional host, dimension probing on init, retries, and input validation). It is exported from redisvl.utils.vectorize, registered as Vectorizers.ollama, and constructible via vectorizer_from_dict with type: "ollama". Optional install is wired as redisvl[ollama] / all (ollama>=0.5.4); the vectorizers user guide documents setup and usage.

Tests: broad unit coverage with a fake ollama client; opt-in integration tests when REDISVL_TEST_OLLAMA=1 and a pulled model are available.

^{Reviewed by Cursor Bugbot for commit 6d6873f. Bugbot is set up for automated code reviews on this repo. Configure here.}

[jit-ci]

Hi, I’m Jit, a friendly security platform designed to help developers build secure applications from day zero with an MVS (Minimal viable security) mindset.

In case there are security findings, they will be communicated to you as a comment inside the PR.

Hope you’ll enjoy using Jit.

Questions? Comments? Want to learn more? Get in touch with us.

[nkanu17]

Yess!! Let's go! Ollama finally

[Copilot]

Pull request overview

Adds an Ollama-backed text embedding provider to RedisVL’s vectorizer system, enabling local embedding generation via an Ollama daemon while fitting into the existing vectorizer registry/from-dict plumbing and test suite.

Changes:

• Introduces OllamaTextVectorizer with sync/async embedding APIs and automatic dimension discovery.
• Wires type: "ollama" into the Vectorizers enum and vectorizer_from_dict, and exports the new vectorizer from redisvl.utils.vectorize.
• Adds an ollama optional extra (also included in all), updates uv.lock, and adds mocked unit tests plus opt-in integration tests gated by REDISVL_TEST_OLLAMA=1.

Reviewed changes

Copilot reviewed 6 out of 7 changed files in this pull request and generated 3 comments.

Show a summary per file

File	Description
redisvl/utils/vectorize/text/ollama.py	New Ollama vectorizer implementation (sync/async) and model-dimension initialization.
redisvl/utils/vectorize/base.py	Adds ollama to the Vectorizers enum.
redisvl/utils/vectorize/__init__.py	Exposes OllamaTextVectorizer and adds type: "ollama" support in vectorizer_from_dict.
pyproject.toml	Adds ollama optional extra and includes it in all.
uv.lock	Locks the new ollama dependency and updates extras metadata.
tests/unit/test_ollama_vectorizer.py	Mocked unit tests for init, sync/async embedding, and registry/from-dict integration.
tests/integration/test_ollama_vectorizer_integration.py	Opt-in real Ollama integration tests gated by env var.

Comments suppressed due to low confidence (1)

redisvl/utils/vectorize/text/ollama.py:298

• Same validation issue as _embed_many: the error message iterates contents to compute element types even when contents is not a list. Passing a non-iterable will raise during message formatting and obscure the actual input-type error; guard before iterating or simplify the message for non-list inputs.

        if not isinstance(contents, list) or not all(
            isinstance(c, str) for c in contents
        ):
            raise TypeError(
                f"Input contents must be a list of strings to embed, got {type(contents)} with elements of types {[type(c) for c in contents]}"

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[nkanu17]

@s-agbede This looks great!
Can you add an example under /docs/user_guide/04_vectorizers.ipynb

[s-agbede]

This looks great!
Can you add an example under /docs/user_guide/04_vectorizers.ipynb

@s-agbede This looks great! Can you add an example under /docs/user_guide/04_vectorizers.ipynb

Thanks @nkanu17. pushed the example changes now

[cursor]

Cursor Bugbot has reviewed your changes and found 1 potential issue.

^{Reviewed by Cursor Bugbot for commit 6d6873f. Configure here.}

[cursor]

Unreachable exception handler in _set_model_dims

Low Severity

The except (KeyError, IndexError) handler in _set_model_dims is dead code. The _embed method already catches all exceptions except ConnectionError and TypeError and wraps them as ValueError (lines 200–203). Any KeyError or IndexError from the Ollama API response (e.g., missing "embeddings" key or empty list) gets converted to ValueError inside _embed before it can propagate to _set_model_dims. Additionally, len() on the returned list cannot raise these either. The intended "Unexpected response" message will never be surfaced.

Additional Locations (1)

• redisvl/utils/vectorize/text/ollama.py#L199-L203

 

^{Reviewed by Cursor Bugbot for commit 6d6873f. Configure here.}