[v3.0.0bx] Python 3.13 support

[JasonWeill]

Problem

Per https://devguide.python.org/versions/, on 2024-10-07, Python 3.13 was officially released, and Python 3.8 is now end of life (EOL). The documentation mentions 3.8 but does not mention 3.12.

Proposed Solution

Test this extension with Python 3.13. Add support for 3.13 and remove 3.8 as a supported Python version.

[krassowski]

I've tried bumping the version on CI to 3.13 in #1029 but it looks like the dependencies are not ready:

@jupyter-ai/magics: Building wheels for collected packages: jupyter_ai_magics, jsonpath-ng, numpy, pyarrow, tokenizers, fastavro, sentencepiece
@jupyter-ai/magics:   Building editable for jupyter_ai_magics (pyproject.toml): started
@jupyter-ai/magics:   Building editable for jupyter_ai_magics (pyproject.toml): finished with status 'done'
@jupyter-ai/magics:   Created wheel for jupyter_ai_magics: filename=jupyter_ai_magics-2.25.0-py3-none-any.whl size=4192 sha256=fefeb6e4a7611aad581f112533849048ba27cfbf13e3c088611dd1f67d0713ee
@jupyter-ai/magics:   Stored in directory: /tmp/pip-ephem-wheel-cache-w0mjej1b/wheels/31/1c/38/c11594954714690c0c46346d59bc385ef752ef76903e01896c
@jupyter-ai/magics:   Building wheel for jsonpath-ng (pyproject.toml): started
@jupyter-ai/magics:   Building wheel for jsonpath-ng (pyproject.toml): finished with status 'done'
@jupyter-ai/magics:   Created wheel for jsonpath-ng: filename=jsonpath_ng-1.7.0-py3-none-any.whl size=30100 sha256=f6d7687a64b09213fa71b546d62e1c9ab5db7107f1e441e90111e899947adc6a
@jupyter-ai/magics:   Stored in directory: /home/runner/.cache/pip/wheels/e9/25/cb/994c9e117083ff481c85008cdbc9f3abf5027a8645ca6b6b51
@jupyter-ai/magics:   Building wheel for numpy (pyproject.toml): started
@jupyter-ai/magics:   Building wheel for numpy (pyproject.toml): finished with status 'done'
@jupyter-ai/magics:   Created wheel for numpy: filename=numpy-1.26.4-cp313-cp313-linux_x86_64.whl size=6917291 sha256=35bba69d126b4acc4d4649652463d17a8b189a415174f88a70ce22fd78b3b15b
@jupyter-ai/magics:   Stored in directory: /home/runner/.cache/pip/wheels/8b/2d/9f/b6b46373f328e2ef50388915d351ccacbedac929459b5459bf
@jupyter-ai/magics:   Building wheel for pyarrow (pyproject.toml): started
@jupyter-ai/magics:   Building wheel for pyarrow (pyproject.toml): finished with status 'error'
@jupyter-ai/magics:   error: subprocess-exited-with-error
@jupyter-ai/magics:   
@jupyter-ai/magics:   × Building wheel for pyarrow (pyproject.toml) did not run successfully.
@jupyter-ai/magics:   │ exit code: 1
@jupyter-ai/magics:   ╰─> [395 lines of output]
@jupyter-ai/magics:       <string>:34: DeprecationWarning: pkg_resources is deprecated as an API. See https://setuptools.pypa.io/en/latest/pkg_resources.html
@jupyter-ai/magics:       running bdist_wheel
@jupyter-ai/magics:       running build
@jupyter-ai/magics:       running build_py
@jupyter-ai/magics:       creating build/lib.linux-x86_64-cpython-313/pyarrow
@jupyter-ai/magics:       copying pyarrow/flight.py -> build/lib.linux-x86_64-cpython-313/pyarrow
@jupyter-ai/magics:       copying pyarrow/benchmark.py -> build/lib.linux-x86_64-cpython-313/pyarrow
@jupyter-ai/magics:       copying pyarrow/_compute_docstrings.py -> build/lib.linux-x86_64-cpython-313/pyarrow
@jupyter-ai/magics:       copying pyarrow/acero.py -> build/lib.linux-x86_64-cpython-313/pyarrow
@jupyter-ai/magics:       copying pyarrow/json.py -> build/lib.linux-x86_64-cpython-313/pyarrow
@jupyter-ai/magics:       copying pyarrow/feather.py -> build/lib.linux-x86_64-cpython-313/pyarrow
@jupyter-ai/magics:       copying pyarrow/compute.py -> build/lib.linux-x86_64-cpython-313/pyarrow
@jupyter-ai/magics:       copying pyarrow/jvm.py -> build/lib.linux-x86_64-cpython-313/pyarrow
@jupyter-ai/magics:       copying pyarrow/util.py -> build/lib.linux-x86_64-cpython-313/pyarrow
@jupyter-ai/magics:       copying pyarrow/conftest.py -> build/lib.linux-x86_64-cpython-313/pyarrow
@jupyter-ai/magics:       copying pyarrow/__init__.py -> build/lib.linux-x86_64-cpython-313/pyarrow
@jupyter-ai/magics:       copying pyarrow/substrait.py -> build/lib.linux-x86_64-cpython-313/pyarrow
@jupyter-ai/magics:       copying pyarrow/csv.py -> build/lib.linux-x86_64-cpython-313/pyarrow
@jupyter-ai/magics:       copying pyarrow/types.py -> build/lib.linux-x86_64-cpython-313/pyarrow
@jupyter-ai/magics:       copying pyarrow/ipc.py -> build/lib.linux-x86_64-cpython-313/pyarrow
@jupyter-ai/magics:       copying pyarrow/_generated_version.py -> build/lib.linux-x86_64-cpython-313/pyarrow
@jupyter-ai/magics:       copying pyarrow/cffi.py -> build/lib.linux-x86_64-cpython-313/pyarrow
@jupyter-ai/magics:       copying pyarrow/dataset.py -> build/lib.linux-x86_64-cpython-313/pyarrow
@jupyter-ai/magics:       copying pyarrow/fs.py -> build/lib.linux-x86_64-cpython-313/pyarrow
@jupyter-ai/magics:       copying pyarrow/orc.py -> build/lib.linux-x86_64-cpython-313/pyarrow
@jupyter-ai/magics:       copying pyarrow/pandas_compat.py -> build/lib.linux-x86_64-cpython-313/pyarrow
@jupyter-ai/magics:       copying pyarrow/cuda.py -> build/lib.linux-x86_64-cpython-313/pyarrow
@jupyter-ai/magics:       creating build/lib.linux-x86_64-cpython-313/benchmarks
@jupyter-ai/magics:       copying benchmarks/convert_pandas.py -> build/lib.linux-x86_64-cpython-313/benchmarks
@jupyter-ai/magics:       copying benchmarks/array_ops.py -> build/lib.linux-x86_64-cpython-313/benchmarks
@jupyter-ai/magics:       copying benchmarks/streaming.py -> build/lib.linux-x86_64-cpython-313/benchmarks
@jupyter-ai/magics:       copying benchmarks/parquet.py -> build/lib.linux-x86_64-cpython-313/benchmarks
@jupyter-ai/magics:       copying benchmarks/io.py -> build/lib.linux-x86_64-cpython-313/benchmarks
@jupyter-ai/magics:       copying benchmarks/common.py -> build/lib.linux-x86_64-cpython-313/benchmarks
@jupyter-ai/magics:       copying benchmarks/__init__.py -> build/lib.linux-x86_64-cpython-313/benchmarks
@jupyter-ai/magics:       copying benchmarks/microbenchmarks.py -> build/lib.linux-x86_64-cpython-313/benchmarks
@jupyter-ai/magics:          Compiling regex-syntax v0.8.5
@jupyter-ai/magics:          Compiling byteorder v1.5.0
@jupyter-ai/magics:          Compiling pkg-config v0.3.31
@jupyter-ai/magics:          Compiling rayon-core v1.12.1
@jupyter-ai/magics:          Compiling darling_core v0.20.10
@jupyter-ai/magics:          Compiling serde_derive v1.0.210
@jupyter-ai/magics:          Compiling darling_macro v0.20.10
@jupyter-ai/magics:          Compiling zerocopy-derive v0.7.35
@jupyter-ai/magics:          Compiling either v1.13.0
@jupyter-ai/magics:          Compiling parking_lot_core v0.9.10
@jupyter-ai/magics:          Compiling zerocopy v0.7.35
@jupyter-ai/magics:          Compiling darling v0.20.10
@jupyter-ai/magics:          Compiling regex-automata v0.4.8
@jupyter-ai/magics:          Compiling onig_sys v69.8.1
@jupyter-ai/magics:       error: failed to run custom build command for `pyo3-ffi v0.21.2`
@jupyter-ai/magics:       
@jupyter-ai/magics:       Caused by:
@jupyter-ai/magics:         process didn't exit successfully: `/tmp/pip-install-9s6smof8/tokenizers_c63090b8c4fe4e4db2bf4856b6fb172a/bindings/python/target/release/build/pyo3-ffi-a4e2687e45f3d2c6/build-script-build` (exit status: 1)
@jupyter-ai/magics:         --- stdout
@jupyter-ai/magics:         cargo:rerun-if-env-changed=PYO3_CROSS
@jupyter-ai/magics:         cargo:rerun-if-env-changed=PYO3_CROSS_LIB_DIR
@jupyter-ai/magics:         cargo:rerun-if-env-changed=PYO3_CROSS_PYTHON_VERSION
@jupyter-ai/magics:         cargo:rerun-if-env-changed=PYO3_CROSS_PYTHON_IMPLEMENTATION
@jupyter-ai/magics:         cargo:rerun-if-env-changed=PYO3_PRINT_CONFIG
@jupyter-ai/magics:         cargo:rerun-if-env-changed=PYO3_USE_ABI3_FORWARD_COMPATIBILITY
@jupyter-ai/magics:       
@jupyter-ai/magics:         --- stderr
@jupyter-ai/magics:         error: the configured Python interpreter version (3.13) is newer than PyO3's maximum supported version (3.12)
@jupyter-ai/magics:         = help: please check if an updated version of PyO3 is available. Current version: 0.21.2
@jupyter-ai/magics:         = help: set PYO3_USE_ABI3_FORWARD_COMPATIBILITY=1 to suppress this check and build anyway using the stable ABI
@jupyter-ai/magics:       warning: build failed, waiting for other jobs to finish...
@jupyter-ai/magics:       💥 maturin failed
@jupyter-ai/magics:         Caused by: Failed to build a native library through cargo
@jupyter-ai/magics:         Caused by: Cargo build finished with "exit status: 101": `env -u CARGO PYO3_ENVIRONMENT_SIGNATURE="cpython-3.13-64bit" PYO3_PYTHON="/opt/hostedtoolcache/Python/3.13.0/x64/bin/python" PYTHON_SYS_EXECUTABLE="/opt/hostedtoolcache/Python/3.13.0/x64/bin/python" "cargo" "rustc" "--features" "pyo3/extension-module" "--message-format" "json-render-diagnostics" "--manifest-path" "/tmp/pip-install-9s6smof8/tokenizers_c63090b8c4fe4e4db2bf4856b6fb172a/bindings/python/Cargo.toml" "--release" "--lib"`
@jupyter-ai/magics:       Error: command ['maturin', 'pep517', 'build-wheel', '-i', '/opt/hostedtoolcache/Python/3.13.0/x64/bin/python', '--compatibility', 'off'] returned non-zero exit status 1
@jupyter-ai/magics:       [end of output]
@jupyter-ai/magics:   
@jupyter-ai/magics:   note: This error originates from a subprocess, and is likely not a problem with pip.
@jupyter-ai/magics:   ERROR: Failed building wheel for tokenizers
@jupyter-ai/magics:   Building wheel for fastavro (pyproject.toml): started
@jupyter-ai/magics:   Building wheel for fastavro (pyproject.toml): finished with status 'done'
@jupyter-ai/magics:   Created wheel for fastavro: filename=fastavro-1.9.7-cp313-cp313-linux_x86_64.whl size=3286831 sha256=74a36638f5721809cc59b7b8f14136c882d96613ac8be48219cb2699e4727849
@jupyter-ai/magics:   Stored in directory: /home/runner/.cache/pip/wheels/9e/fe/0b/429cdc095e0d5ded719efce8d3147f639b9f8ae7e7d9a89d96
@jupyter-ai/magics:   Building wheel for sentencepiece (pyproject.toml): started
@jupyter-ai/magics:   Building wheel for sentencepiece (pyproject.toml): finished with status 'done'
@jupyter-ai/magics:   Created wheel for sentencepiece: filename=sentencepiece-0.2.0-cp313-cp313-linux_x86_64.whl size=1291226 sha256=30d6d6eaae02e49d2a4f9a7f8f92fa1d0c95edc0654143f02799f38e63af7e83
@jupyter-ai/magics:   Stored in directory: /home/runner/.cache/pip/wheels/02/3d/f0/8dbb11925fbd58340b128ae9334ae7028332d7a5751738f616
@jupyter-ai/magics: Successfully built jupyter_ai_magics jsonpath-ng numpy fastavro sentencepiece
@jupyter-ai/magics: Failed to build pyarrow tokenizers
@jupyter-ai/magics: ERROR: ERROR: Failed to build installable wheels for some pyproject.toml based projects (pyarrow, tokenizers)

[krassowski]

Blocked on:

• [Python][Packaging] Support Python 3.13 and upload wheels apache/arrow#43519
• Python 3.13 support huggingface/tokenizers#1639

[dlqqq]

@krassowski Thank you for documenting your research here! In summary, we need to wait for pyarrow and tokenizers to have Python 3.13 distributions before we can declare Python 3.13 support.

[krassowski]

Yes, though I would personally also wait until the CI passes with tests running on Python 3.13 before declaring support.

[cecilyen]

pyarrow has added support for Python 3.13. https://arrow.apache.org/docs/python/install.html#:~:text=3.13

[cecilyen]

tokenizers just added Python 3.13 support today. huggingface/tokenizers#1665

[krassowski]

tokenizers just added Python 3.13 support today. huggingface/tokenizers#1665

It looks like PR was merged but there has not been a new release yet?

[fperez]

Hi folks, checking in here... I just created a python 3.13 mamba environment to check, and at least using conda-forge, it's still not possible to run jupyter-ai on 3.13.

(base) longs[~]> mamba activate py313
(py313) longs[~]> mamba install jupyter-ai
conda-forge/noarch                                          Using cache
conda-forge/osx-64                                          Using cache
nodefaults/noarch                                           Using cache
nodefaults/osx-64                                           Using cache

Pinned packages:

  - python=3.13

error    libmamba Could not solve for environment specs
    The following packages are incompatible
    ├─ jupyter-ai =* * is installable with the potential options
    │  ├─ jupyter-ai [2.10.0|2.11.0|...|2.19.0] would require

[ loads more... ]

    │  └─ jupyter-ai [2.28.1|2.28.2|...|2.31.2] would require
    │     └─ faiss-cpu >=1.8.0,<2 *, which can be installed (as previously explained);
    └─ pin on python 3.13.* =* * is not installable because it requires
       └─ python =3.13 *, which conflicts with any installable versions previously reported.
critical libmamba Could not solve for environment specs

The problem seems to stem from faiss-cpu, which does list 3.13 as a supported version on PyPI, but for some reason won't resolve despite the package being on conda-forge (at least v1.9.0).

I don't have the bandwidth to track the dependency rabbit hole but wanted to ping here in case anyone has ideas of how to get this unstuck... I'd like to teach this fall using jupyter-ai and py3.13 in a conda/mamba-managed container setup, so I'd be grateful for any ideas on how to proceed further.

Thanks all!