Datasets:

Epulo-ai
/

open-kitchen-protocol

Error code:   StreamingRowsError
Exception:    CastError
Message:      Couldn't cast
description: string
sites: list<item: struct<site_id: string, format: string, permanence: string, country: string, valid_from:  (... 38 chars omitted)
  child 0, item: struct<site_id: string, format: string, permanence: string, country: string, valid_from: timestamp[s (... 26 chars omitted)
      child 0, site_id: string
      child 1, format: string
      child 2, permanence: string
      child 3, country: string
      child 4, valid_from: timestamp[s]
      child 5, valid_to: timestamp[s]
service_context: struct<ref: string, type: string, covers: int64>
  child 0, ref: string
  child 1, type: string
  child 2, covers: int64
events: list<item: struct<event_id: string, site_id: string, station_id: string, actor_kind: string, actor_r (... 383 chars omitted)
  child 0, item: struct<event_id: string, site_id: string, station_id: string, actor_kind: string, actor_ref: string, (... 371 chars omitted)
      child 0, event_id: string
      child 1, site_id: string
      child 2, station_id: string
      child 3, actor_kind: string
      child 4, actor_ref: string
      child 5, verb: string
      child 6, object_refs: list<item: string>
          child 0, item: string
      child 7, t_start: timestamp[s]
      child 8, t_end: timestamp[s]
      child 9, outcome: string
      child 10, measures: struct<weight_g: int64, duration_s: int64, distance_m: int64, temp_c: double, count: int64>
          child 0, weight_g: int64
          child 1, duration_s: int64
          child 2, distance_m: int64
          child 3, temp_c: double
          child 4, count: int64
      child 11, source: string
      child 12, confidence: double
      child 13, privacy_tier: string
      child 14, session_ref: string
      child 15, service_context_ref: string
      child 16, variant: string
      child 17, quality_flags: list<item: string>
          child 0, item: string
note: string
records_are_synthetic: bool
dataset_repository: string
files: list<item: struct<path: string, github_path: string, sha256: string, generated: bool>>
  child 0, item: struct<path: string, github_path: string, sha256: string, generated: bool>
      child 0, path: string
      child 1, github_path: string
      child 2, sha256: string
      child 3, generated: bool
protocol_status: string
github_commit: string
github_repository: string
to
{'github_repository': Value('string'), 'github_commit': Value('string'), 'protocol_status': Value('string'), 'dataset_repository': Value('string'), 'records_are_synthetic': Value('bool'), 'files': List({'path': Value('string'), 'github_path': Value('string'), 'sha256': Value('string'), 'generated': Value('bool')})}
because column names don't match
Traceback:    Traceback (most recent call last):
                File "/src/services/worker/src/worker/utils.py", line 147, in get_rows_or_raise
                  return get_rows(
                      dataset=dataset,
                  ...<4 lines>...
                      column_names=column_names,
                  )
                File "/src/libs/libcommon/src/libcommon/utils.py", line 272, in decorator
                  return func(*args, **kwargs)
                File "/src/services/worker/src/worker/utils.py", line 127, in get_rows
                  rows_plus_one = list(itertools.islice(safe_iter(ds, dataset=dataset), rows_max_number + 1))
                File "/src/services/worker/src/worker/utils.py", line 483, in safe_iter
                  yield from ds.decode(False) if ds.features else ds
                File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 2840, in __iter__
                  for key, example in ex_iterable:
                                      ^^^^^^^^^^^
                File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 2373, in __iter__
                  for key, pa_table in self._iter_arrow():
                                       ~~~~~~~~~~~~~~~~^^
                File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 2398, in _iter_arrow
                  for key, pa_table in self.ex_iterable._iter_arrow():
                                       ~~~~~~~~~~~~~~~~~~~~~~~~~~~~^^
                File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 536, in _iter_arrow
                  for key, pa_table in iterator:
                                       ^^^^^^^^
                File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 419, in _iter_arrow
                  for key, pa_table in self.generate_tables_fn(**gen_kwags):
                                       ~~~~~~~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^
                File "/usr/local/lib/python3.14/site-packages/datasets/packaged_modules/json/json.py", line 343, in _generate_tables
                  self._cast_table(pa_table, json_field_paths=json_field_paths),
                  ~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
                File "/usr/local/lib/python3.14/site-packages/datasets/packaged_modules/json/json.py", line 132, in _cast_table
                  pa_table = table_cast(pa_table, self.info.features.arrow_schema)
                File "/usr/local/lib/python3.14/site-packages/datasets/table.py", line 2378, in table_cast
                  return cast_table_to_schema(table, schema)
                File "/usr/local/lib/python3.14/site-packages/datasets/table.py", line 2306, in cast_table_to_schema
                  raise CastError(
                  ...<3 lines>...
                  )
              datasets.table.CastError: Couldn't cast
              description: string
              sites: list<item: struct<site_id: string, format: string, permanence: string, country: string, valid_from:  (... 38 chars omitted)
                child 0, item: struct<site_id: string, format: string, permanence: string, country: string, valid_from: timestamp[s (... 26 chars omitted)
                    child 0, site_id: string
                    child 1, format: string
                    child 2, permanence: string
                    child 3, country: string
                    child 4, valid_from: timestamp[s]
                    child 5, valid_to: timestamp[s]
              service_context: struct<ref: string, type: string, covers: int64>
                child 0, ref: string
                child 1, type: string
                child 2, covers: int64
              events: list<item: struct<event_id: string, site_id: string, station_id: string, actor_kind: string, actor_r (... 383 chars omitted)
                child 0, item: struct<event_id: string, site_id: string, station_id: string, actor_kind: string, actor_ref: string, (... 371 chars omitted)
                    child 0, event_id: string
                    child 1, site_id: string
                    child 2, station_id: string
                    child 3, actor_kind: string
                    child 4, actor_ref: string
                    child 5, verb: string
                    child 6, object_refs: list<item: string>
                        child 0, item: string
                    child 7, t_start: timestamp[s]
                    child 8, t_end: timestamp[s]
                    child 9, outcome: string
                    child 10, measures: struct<weight_g: int64, duration_s: int64, distance_m: int64, temp_c: double, count: int64>
                        child 0, weight_g: int64
                        child 1, duration_s: int64
                        child 2, distance_m: int64
                        child 3, temp_c: double
                        child 4, count: int64
                    child 11, source: string
                    child 12, confidence: double
                    child 13, privacy_tier: string
                    child 14, session_ref: string
                    child 15, service_context_ref: string
                    child 16, variant: string
                    child 17, quality_flags: list<item: string>
                        child 0, item: string
              note: string
              records_are_synthetic: bool
              dataset_repository: string
              files: list<item: struct<path: string, github_path: string, sha256: string, generated: bool>>
                child 0, item: struct<path: string, github_path: string, sha256: string, generated: bool>
                    child 0, path: string
                    child 1, github_path: string
                    child 2, sha256: string
                    child 3, generated: bool
              protocol_status: string
              github_commit: string
              github_repository: string
              to
              {'github_repository': Value('string'), 'github_commit': Value('string'), 'protocol_status': Value('string'), 'dataset_repository': Value('string'), 'records_are_synthetic': Value('bool'), 'files': List({'path': Value('string'), 'github_path': Value('string'), 'sha256': Value('string'), 'generated': Value('bool')})}
              because column names don't match

Need help to make the dataset viewer work? Make sure to review how to configure the dataset viewer, and open a discussion for direct support.

Open Kitchen Protocol: kitchen work, described together

A public working draft for recording kitchen operations across people, robots and software agents. Start with one workflow, inspect its Events, and show us the meaning your implementation needs.

OKP v0.1 draft | 3 synthetic scenarios | 13 Events | Apache-2.0

OKP can be used without an Epulo account, a CKB subscription or a proprietary service. Epulo stewards the project.

What GitHub and Hugging Face provide

GitHub is the protocol's source of truth: specifications, validator, tests and contribution process. This Hugging Face dataset provides synthetic fixtures and the corresponding protocol snapshot for inspection and evaluation.

This snapshot corresponds to reviewed GitHub commit 93403595edf05f9da382dbc9ad291f34443ca878 It is a development snapshot, not a new numbered release. alignment.json records source paths and SHA-256 checksums.

Explore one kitchen scenario

File in this dataset	Scenario	Events
breakfast-rush.example.json	Synthetic morning production with human, robot and agent Events	4
banqueting.example.json	Synthetic preparation, service and handoffs across sites	5
inflight.example.json	Synthetic catering production and service context	4

Each JSON document contains an events array and contextual information. Events describe actions, timestamps, actor categories, references, outcomes, measurements and stated observation sources. Values and outcomes in these files are illustrative, not measurements from deployed kitchens.

Other files:

ONTOLOGY.md: conceptual entities, actions and draft tier policy.
kitchen-event.schema.json: the Event schema.
docs/CONFORMANCE.md: current strict validation profile.
docs/TIER_ACTOR_CHECK.md: the optional actor-reference check.
LICENSE: Apache-2.0 license text.

The schema and example files sit at the root of this Hugging Face repository. In GitHub they sit in schema/ and examples/ respectively. Validator commands in the documentation are run from the GitHub checkout.

Reproduce validation

Git and Python 3.8 or newer are sufficient. No paid service is required.

git clone https://github.com/Epulo-ai/open-kitchen-protocol.git
cd open-kitchen-protocol
git checkout --detach 93403595edf05f9da382dbc9ad291f34443ca878
python3 -m unittest discover -s tests -v
python3 tools/validate_okp.py --strict examples/*.json

Publication requires the regression tests and synthetic examples to pass the current strict profile. It checks Event structure and selected consistency rules. It does not resolve all contextual references or prove an entire workflow.

The ontology intends T2 records to omit actor references and T3 to represent aggregates. The synthetic fixtures still retain actor_ref on T2/T3 Events. The optional experimental check makes that known mismatch visible:

python3 tools/validate_okp.py --strict --check-tier-actors examples/*.json

Expected result: exit code 1, with 13 actor-reference errors, comprising 10 T2 and 3 T3 Events. This does not change the existing strict profile or rewrite the files. It is one field check, not full tier enforcement.

Intended use and limits

Use these examples to explore the vocabulary, build parsers, test mappings, or report concrete implementation gaps. They are too small and too synthetic to establish kitchen coverage, a representative benchmark, performance, commercial savings or production reliability.

The files contain no camera streams, sensor sequences or motor trajectories. The LeRobot mapping is conceptual. A tested runnable converter is not included, and these files are not presented as a ready-to-train LeRobot dataset.

An Event records an operational observation. CKB is a separately developed culinary knowledge layer; its private contents are not distributed here. Cloud planning, local sensing and skills, robot controllers and independent safety systems have separate responsibilities. These fixtures do not execute or certify a robot.

Pseudonymous actor tokens and tier labels do not establish anonymity, consent, publication rights or legal compliance. The Event validator requires a session reference for human Events but does not verify the referenced agreement or retention information. Contextual fields can still link records.

Real customer-kitchen data is not included. Publishing OKP does not grant Epulo or anyone else access to a contributor's operational data.

Contribute one workflow

Operators: describe five steps and one exception, quality check or handoff in plain language. No coding is required.

Builders: create a synthetic Event or report one concrete meaning the draft cannot preserve. Include the exact revision, command, result and first point of confusion. An independently returned file is evidence of evaluation; sustained use is stronger evidence of adoption.

Use the GitHub issue tracker or submit a pull request. Only contribute synthetic or explicitly rights-cleared material suitable for public distribution. Keep private customer, employee, recipe and equipment material out of public submissions.

The public protocol and example fixtures are offered under Apache-2.0. The standardization goal remains open for community implementation and review.

Downloads last month: 45

Total file size:

105 kB