Tiktoken

- Support for `o1` and `o3` models
- Better error messages when loading invalid vocabulary files
- Support for encoding to numpy arrays
- Delayed imports when not strictly necessary

- Support for `o1-` and `chatgpt-4o-` models
- Build wheels for Python 3.13
- Add possessive quantifiers to limit backtracking in regular expressions, thanks to l0rinc!
- Provide a better error message and type for invalid token decode
- Permit tuples in type hints
- Better error message for passing invalid input to `get_encoding`
- Better error messages during plugin loading
- Add a `__version__` attribute
- Update versions of `pyo3`, `regex`, `fancy-regex`
- Drop support for Python 3.8

- Support for `gpt-4o`
- Performance improvements

- Optimise regular expressions for a 20% performance improvement, thanks to paplorinc!
- Add `text-embedding-3-*` models to `encoding_for_model`
- Check content hash for downloaded files
- Allow pickling `Encoding` objects. Registered `Encoding` will be pickled by reference
- Workaround PyO3 bug for frozenset conversion

Thank you to paplorinc, mdwelsh, Praneet460!

- Build wheels for Python 3.12
- Update version of PyO3 to allow multiple imports
- Avoid permission errors when using default cache logic

- Add `encoding_name_for_model`, undo some renames to variables that are implementation details