NewYour coding agent can read the release notes before it upgrades.Set up the MCP server →
PyPI, #302 most downloaded on PyPI
The Jupyter Notebook format
Latest stable
5.11.1
Last release 3 months ago
17 Aug 2026
Ships unpredictably
gaps range from 1 weeks to 2.3 years
Nearly every release is documented
notes for 34 of 36 stable releases
3 versions withdrawn
withdrawn after publishing
11 years old
43 releases · first in 2015
Drag across the ruler to compare, or tap a release to open it.
disable pendingdeprecationWarnings for now #458 ( @Carreau )
The following people contributed discussions, new ideas, code and documentation contributions, and review.
See our definition of contributors.
Your coding agent can read these notes before it upgrades. Set up the MCP server →
…on creation) and are now emitting a PendingDeprecationWarning to indicate or wish to move to a context manager API and avoid resouce leaking. Again th…
NotebookNotary's stores are regularly note .close()d, so we introduce the ability to use it at a context manger once, (as .store is opened on creation) and are now emitting a PendingDeprecationWarning to indicate or wish to move to a context manager API and avoid resouce leaking. Again this is a PendingDeprecationWarning, please let us know if the current context manager API is problematic.
The following people contributed discussions, new ideas, code and documentation contributions, and review.
See our definition of contributors.
(GitHub contributors page for this release)
@blink1073 (activity) | @Carreau (activity) | @claude (activity) | @crypdick (activity) | @jasongrout (activity) | @LecrisUT (activity) | @rgbkrk (activity)
Maintenance and upkeep improvements
Revert PR #378 (Simplify NotebookNotary._data_dir_default) #397 ( @krassowski )
Maintenance and upkeep improvements
No merged PRs
No merged PRs
Update Release Scripts #394 (@blink1073)
(GitHub contributors page for this release)
@blink1073 | @Carreau | @cmd-ntrf | @dependabot | @pre-commit-ci | @pydanny
Fix version handling #372 ( @blink1073 )
Fix deprecation warning when importing from jsonschema #368 ( @eladkal )
Maintenance and upkeep improvements
Add strip_invalid_metadata argument to validator normalize method #355 (@jonabc)
(GitHub contributors page for this release)
@blink1073 | @dcsaba89 | @jonabc | @pre-commit-ci | @rschroll | @westurner
Undeprecate validate(nb, relax_add_props=True) #343 (@minrk)
Only require nbformat_minor for v4 #342 (@minrk)
Expose more attributes for typing #337 (@blink1073)
Fix deprecation warning suggestion by @martinRenou in https://github.com/jupyter/nbformat/pull/311
Release 5.7.0
Full Changelog: https://github.com/jupyter/nbformat/compare/5.6.1...5.7.0
Fix handling of __version__ on Python 3.7.
__version__ on Python 3.7.Fix docs and type annotations for validator.normalize.
validator.normalize.The biggest change in nbformat 5.5.0 is the deprecation of arguments to validate() that try to fix notebook errors during validation.
The biggest change in nbformat 5.5.0 is the deprecation of arguments
to validate() that try to fix notebook errors during validation.
validate() is a function that is core to the security model of
Jupyter. Callers rely on it not mutating its argument or trying to fix
the notebook passed to it.
Automatically fixing a notebook during validation can hide subtle bugs. The auto-fixing arguments are therefore deprecated and will be removed in a future release; validation will fail instead of silently modifying an invalid notebook.
Use normalize() explicitly when a notebook from an external source needs
to be repaired before validation. It returns a normalized copy and the
number of changes made:
from nbformat.validator import normalize
changes, normalized = normalize(notebook)
nbformat.validate(normalized)
The preferred approach is still to generate valid notebooks at the source; normalization is a compatibility tool, not a substitute for fixing the producer.
nbformat is now built with flit, and uses pyproject.tomlNothing published for this version
Add project URLs to setup.py by @tlinhart in https://github.com/jupyter/nbformat/pull/266
Release 5.4.0
Full Changelog: https://github.com/jupyter/nbformat/compare/5.3.0...5.4.0
setup.pynbformat.currentmypy and typings supportAdopt pre-commit and auto-formatters
fastjsonschema by defaultpre-commit and auto-formattersjsonschema to 2.6, handle warningsNothing published for this version
Return the notebook when no conversion is needed by @bryanwweber in https://github.com/jupyter/nbformat/pull/221
Release 5.2.0
Full Changelog: https://github.com/jupyter/nbformat/compare/5.1.3...5.2.0
Change id generation to be hash based to avoid problematic word combinations
- Fixed missing file in manifest
Changes convert.upgrade to upgrade minor 4.x versions to 4.5
Implemented CellIds from JEP-62
Nothing published for this version
Nothing published for this version
Add optional support for using [fastjsonschema]{.title-ref} as the JSON validation library. To enable fast validation, install [fastjsonschema]{.title
Fixed a bug where default values for validator.get_validator() failed with an import error
nbformat.read() function has a better duck-type interface and will raise more meaningful error messages if it can't parse a notebook document.
Allow notebook format 4.0 and 4.1 to have the arbitrary JSON mimebundles from format 4.2 for pragmatic purposes.
Fixed issue causing python 2 to pick up 5.0.x releases.
Removed debug print statements from project.
Added schema validation files for older versions. This was breaking notebook generation.
Explicitly state that metadata fields can be ignored.
metadata.jupyter).source_hidden and outputs_hidden as official front-end
metadata fields to indicate hiding source and outputs areas. NB:
These fields should not be used to hide elements in exported
formats.nbformat.validate can be set to not
raise a ValidationError if additional properties are included.A new pluggable SignatureStore class allows specifying different ways to record the signatures of trusted notebooks. The default is still an SQLite da
SignatureStore class allows specifying different
ways to record the signatures of trusted notebooks. The default is
still an SQLite database. See
pluggable_signature_store for more
information.nbformat.read and
nbformat.write accept file paths as
bytes as well as unicode.nbformat.validate
on an empty dictionary.Update nbformat spec version to 4.2, allowing JSON outputs to have any JSONable type, not just object, and mime-types of the form application/anything
object, and mime-types of the form
application/anything+json.authors in notebook metadata.
nb.metadata.authors shall be a list of objects with the property
name, a string of each author's full name.cat notebook | jupyter trustUpdate nbformat spec version to 4.1, adding support for attachments on markdown and raw cells.
:memory: if
the database cannot be opened.Nothing published for this version
The first release of nbformat as its own package.
The first release of nbformat as its own package.
Nothing published for this version
One column per quarter.