Update Python to >=3.14,<3.15

[renovate]

This PR contains the following updates:

Package	Type	Update	Change
python (source)	requires-python	minor	>=3.12,<3.13 → >=3.14,<3.15

---

Warning

Some dependencies could not be looked up. Check the Dependency Dashboard for more information.

---

Release Notes

python/cpython (python)

v3.14.7

Compare Source

v3.14.6

Compare Source

v3.14.5

Compare Source

v3.14.4

Compare Source

v3.14.3

Compare Source

v3.14.2

Compare Source

v3.14.1

Compare Source

v3.14.0

Compare Source

v3.13.15

Compare Source

v3.13.14

Compare Source

v3.13.13

Compare Source

v3.13.12

Compare Source

v3.13.11

Compare Source

v3.13.10

Compare Source

v3.13.9

Compare Source

v3.13.8

Compare Source

v3.13.7

Compare Source

v3.13.6

Compare Source

v3.13.5

Compare Source

v3.13.4

Compare Source

v3.13.3

Compare Source

v3.13.2

Compare Source

v3.13.1

Compare Source

v3.13.0

Compare Source

---

Configuration

📅 Schedule: (in timezone US/Eastern)

• Branch creation
  • "every weekend"
• Automerge
  • At any time (no schedule defined)

🚦 Automerge: Disabled by config. Please merge this manually once you are satisfied.

♻ Rebasing: Whenever PR is behind base branch, or you tick the rebase/retry checkbox.

🔕 Ignore: Close this PR and you won't be reminded about this update again.

---

• If you want to rebase/retry this PR, check this box

---

This PR was generated by Mend Renovate. View the repository job log.

[github-actions]

OpenAPI Changes

No changes detected

View full changelog

Unexpected changes? Ensure your branch is up-to-date with main (consider rebasing).

[mbertrand]

@shanbady, GensimEncoder is blocking the Python 3.14 upgrade. gensim 4.4.0 won't install on 3.14 because its C extension doesn't compile, and no gensim release has 3.14 wheels yet. The fix is in piskvorky/gensim#3662, which is still open.

I think we can either:

1. Wait for Ship a .pyx-only sdist so gensim builds on Python 3.13+ (fixes #3634) piskvorky/gensim#3662 to be merged and released, and hold this PR until then.
2. Switch the local default to a different lightweight model.

Do you have a preference, or a model in mind for option 2? Maybe modelvec?

[shanbady]

@mbertrand i think it might be worth switching to model2vec if the results look reasonable:

from model2vec import StaticModel

from vector_search.encoders.base import BaseEncoder


class Model2VecEncoder(BaseEncoder):
    """
    Static (distilled) embedding encoder - fast on CPU, no torch/GPU required
    """

    def __init__(self, model_name="minishlab/potion-retrieval-32M"):
        self.model_name = model_name
        self.model = StaticModel.from_pretrained(model_name)

    def embed_documents(self, documents):
        return self.model.encode(documents).tolist()

    def dim(self):
        return self.model.dim