Module airbyte_agent_sdk.connectors.jira.types
Type definitions for jira connector.
Classes
AirbyteSearchParams(*args, **kwargs)
: Parameters for Airbyte cache search operations (generic, use entity-specific query types for better type hints).
Ancestors (in MRO)
- builtins.dict
Class variables
cursor: str
: The type of the None singleton.
fields: list[list[str]]
: The type of the None singleton.
limit: int
: The type of the None singleton.
query: dict[str, typing.Any]
: The type of the None singleton.
IssueCommentsAndCondition(*args, **kwargs)
: dict() -> new empty dictionary
dict(mapping) -> new dictionary initialized from a mapping object's
(key, value) pairs
dict(iterable) -> new dictionary initialized as if via:
d = {}
for k, v in iterable:
d[k] = v
dict(**kwargs) -> new dictionary initialized with the name=value pairs
in the keyword argument list. For example: dict(one=1, two=2)
Ancestors (in MRO)
- builtins.dict
Class variables
and: list[airbyte_agent_sdk.connectors.jira.types.IssueCommentsEqCondition | airbyte_agent_sdk.connectors.jira.types.IssueCommentsNeqCondition | airbyte_agent_sdk.connectors.jira.types.IssueCommentsGtCondition | airbyte_agent_sdk.connectors.jira.types.IssueCommentsGteCondition | airbyte_agent_sdk.connectors.jira.types.IssueCommentsLtCondition | airbyte_agent_sdk.connectors.jira.types.IssueCommentsLteCondition | airbyte_agent_sdk.connectors.jira.types.IssueCommentsInCondition | airbyte_agent_sdk.connectors.jira.types.IssueCommentsStartswithCondition | airbyte_agent_sdk.connectors.jira.types.IssueCommentsEndswithCondition | airbyte_agent_sdk.connectors.jira.types.IssueCommentsFuzzyCondition | airbyte_agent_sdk.connectors.jira.types.IssueCommentsKeywordCondition | airbyte_agent_sdk.connectors.jira.types.IssueCommentsContainsCondition | airbyte_agent_sdk.connectors.jira.types.IssueCommentsArrayContainsCondition | airbyte_agent_sdk.connectors.jira.types.IssueCommentsNotCondition | airbyte_agent_sdk.connectors.jira.types.IssueCommentsAndCondition | airbyte_agent_sdk.connectors.jira.types.IssueCommentsOrCondition | airbyte_agent_sdk.connectors.jira.types.IssueCommentsAnyCondition]
: The type of the None singleton.
IssueCommentsAnyCondition(*args, **kwargs)
: dict() -> new empty dictionary
dict(mapping) -> new dictionary initialized from a mapping object's
(key, value) pairs
dict(iterable) -> new dictionary initialized as if via:
d = {}
for k, v in iterable:
d[k] = v
dict(**kwargs) -> new dictionary initialized with the name=value pairs
in the keyword argument list. For example: dict(one=1, two=2)
Ancestors (in MRO)
- builtins.dict
Class variables
any: airbyte_agent_sdk.connectors.jira.types.IssueCommentsAnyValueFilter
: The type of the None singleton.
IssueCommentsAnyValueFilter(*args, **kwargs)
: Available fields with Any value type. Used for 'contains' and 'any' conditions.
Ancestors (in MRO)
- builtins.dict
Class variables
author: Any
: The ID of the user who created the comment
body: Any
: The comment text in Atlassian Document Format
created: Any
: The date and time at which the comment was created
id: Any
: The ID of the comment
issue_id: Any
: Id of the related issue
jsd_public: Any
: Whether the comment is visible in Jira Service Desk
properties: Any
: A list of comment properties
rendered_body: Any
: The rendered version of the comment
self: Any
: The URL of the comment
update_author: Any
: The ID of the user who updated the comment last
updated: Any
: The date and time at which the comment was updated last
visibility: Any
: The group or role to which this item is visible
IssueCommentsArrayContainsCondition(*args, **kwargs)
: Exact membership test on an array field. Example: {"array_contains": {"tags": "premium"}}
Ancestors (in MRO)
- builtins.dict
Class variables
array_contains: airbyte_agent_sdk.connectors.jira.types.IssueCommentsAnyValueFilter
: The type of the None singleton.
IssueCommentsContainsCondition(*args, **kwargs)
: Case-insensitive substring match on a scalar field. Example: {"contains": {"subject": "billing"}}
Ancestors (in MRO)
- builtins.dict
Class variables
contains: airbyte_agent_sdk.connectors.jira.types.IssueCommentsAnyValueFilter
: The type of the None singleton.
IssueCommentsCreateParams(*args, **kwargs)
: Parameters for issue_comments.create operation
Ancestors (in MRO)
- builtins.dict
Class variables
body: airbyte_agent_sdk.connectors.jira.types.IssueCommentsCreateParamsBody
: The type of the None singleton.
expand: str
: The type of the None singleton.
issue_id_or_key: str
: The type of the None singleton.
properties: list[dict[str, typing.Any]]
: The type of the None singleton.
visibility: airbyte_agent_sdk.connectors.jira.types.IssueCommentsCreateParamsVisibility
: The type of the None singleton.
IssueCommentsCreateParamsBody(*args, **kwargs)
: Comment content in Atlassian Document Format (ADF)
Ancestors (in MRO)
- builtins.dict
Class variables
content: list[airbyte_agent_sdk.connectors.jira.types.IssueCommentsCreateParamsBodyContentItem]
: The type of the None singleton.
type_: str
: The type of the None singleton.
version: int
: The type of the None singleton.
IssueCommentsCreateParamsBodyContentItem(*args, **kwargs)
: Nested schema for IssueCommentsCreateParamsBody.content_item
Ancestors (in MRO)
- builtins.dict
Class variables
content: list[airbyte_agent_sdk.connectors.jira.types.IssueCommentsCreateParamsBodyContentItemContentItem]
: The type of the None singleton.
type_: str
: The type of the None singleton.
IssueCommentsCreateParamsBodyContentItemContentItem(*args, **kwargs)
: Nested schema for IssueCommentsCreateParamsBodyContentItem.content_item
Ancestors (in MRO)
- builtins.dict
Class variables
text: str
: The type of the None singleton.
type_: str
: The type of the None singleton.
IssueCommentsCreateParamsVisibility(*args, **kwargs)
: Restrict comment visibility to a group or role
Ancestors (in MRO)
- builtins.dict
Class variables
identifier: str
: The type of the None singleton.
type_: str
: The type of the None singleton.
value: str
: The type of the None singleton.
IssueCommentsDeleteParams(*args, **kwargs)
: Parameters for issue_comments.delete operation
Ancestors (in MRO)
- builtins.dict
Class variables
comment_id: str
: The type of the None singleton.
issue_id_or_key: str
: The type of the None singleton.
IssueCommentsEndswithCondition(*args, **kwargs)
: Literal case-insensitive suffix match.
Ancestors (in MRO)
- builtins.dict
Class variables
endswith: airbyte_agent_sdk.connectors.jira.types.IssueCommentsStringFilter
: The type of the None singleton.
IssueCommentsEqCondition(*args, **kwargs)
: Equal to: field equals value.
Ancestors (in MRO)
- builtins.dict
Class variables
eq: airbyte_agent_sdk.connectors.jira.types.IssueCommentsSearchFilter
: The type of the None singleton.
IssueCommentsFuzzyCondition(*args, **kwargs)
: Ordered word text match (case-insensitive).
Ancestors (in MRO)
- builtins.dict
Class variables
fuzzy: airbyte_agent_sdk.connectors.jira.types.IssueCommentsStringFilter
: The type of the None singleton.
IssueCommentsGetParams(*args, **kwargs)
: Parameters for issue_comments.get operation
Ancestors (in MRO)
- builtins.dict
Class variables
comment_id: str
: The type of the None singleton.
expand: str
: The type of the None singleton.
issue_id_or_key: str
: The type of the None singleton.
IssueCommentsGtCondition(*args, **kwargs)
: Greater than: field > value.
Ancestors (in MRO)
- builtins.dict
Class variables
gt: airbyte_agent_sdk.connectors.jira.types.IssueCommentsSearchFilter
: The type of the None singleton.
IssueCommentsGteCondition(*args, **kwargs)
: Greater than or equal: field >= value.
Ancestors (in MRO)
- builtins.dict
Class variables
gte: airbyte_agent_sdk.connectors.jira.types.IssueCommentsSearchFilter
: The type of the None singleton.
IssueCommentsInCondition(*args, **kwargs)
: dict() -> new empty dictionary
dict(mapping) -> new dictionary initialized from a mapping object's
(key, value) pairs
dict(iterable) -> new dictionary initialized as if via:
d = {}
for k, v in iterable:
d[k] = v
dict(**kwargs) -> new dictionary initialized with the name=value pairs
in the keyword argument list. For example: dict(one=1, two=2)
Ancestors (in MRO)
- builtins.dict
Class variables
in: airbyte_agent_sdk.connectors.jira.types.IssueCommentsInFilter
: The type of the None singleton.
IssueCommentsInFilter(*args, **kwargs)
: Available fields for 'in' condition (values are lists).
Ancestors (in MRO)
- builtins.dict
Class variables
author: list[dict[str, typing.Any]]
: The ID of the user who created the comment
body: list[dict[str, typing.Any]]
: The comment text in Atlassian Document Format
created: list[str]
: The date and time at which the comment was created
id: list[str]
: The ID of the comment
issue_id: list[str]
: Id of the related issue
jsd_public: list[bool]
: Whether the comment is visible in Jira Service Desk
properties: list[list[typing.Any]]
: A list of comment properties
rendered_body: list[str]
: The rendered version of the comment
self: list[str]
: The URL of the comment
update_author: list[dict[str, typing.Any]]
: The ID of the user who updated the comment last
updated: list[str]
: The date and time at which the comment was updated last
visibility: list[dict[str, typing.Any]]
: The group or role to which this item is visible
IssueCommentsKeywordCondition(*args, **kwargs)
: Keyword text match (any word present).
Ancestors (in MRO)
- builtins.dict
Class variables
keyword: airbyte_agent_sdk.connectors.jira.types.IssueCommentsStringFilter
: The type of the None singleton.
IssueCommentsListParams(*args, **kwargs)
: Parameters for issue_comments.list operation
Ancestors (in MRO)
- builtins.dict
Class variables
expand: str
: The type of the None singleton.
issue_id_or_key: str
: The type of the None singleton.
max_results: int
: The type of the None singleton.
order_by: str
: The type of the None singleton.
start_at: int
: The type of the None singleton.
IssueCommentsLtCondition(*args, **kwargs)
: Less than: field < value.
Ancestors (in MRO)
- builtins.dict
Class variables
lt: airbyte_agent_sdk.connectors.jira.types.IssueCommentsSearchFilter
: The type of the None singleton.
IssueCommentsLteCondition(*args, **kwargs)
: Less than or equal: field <= value.
Ancestors (in MRO)
- builtins.dict
Class variables
lte: airbyte_agent_sdk.connectors.jira.types.IssueCommentsSearchFilter
: The type of the None singleton.
IssueCommentsNeqCondition(*args, **kwargs)
: Not equal to: field does not equal value.
Ancestors (in MRO)
- builtins.dict
Class variables
neq: airbyte_agent_sdk.connectors.jira.types.IssueCommentsSearchFilter
: The type of the None singleton.
IssueCommentsNotCondition(*args, **kwargs)
: dict() -> new empty dictionary
dict(mapping) -> new dictionary initialized from a mapping object's
(key, value) pairs
dict(iterable) -> new dictionary initialized as if via:
d = {}
for k, v in iterable:
d[k] = v
dict(**kwargs) -> new dictionary initialized with the name=value pairs
in the keyword argument list. For example: dict(one=1, two=2)
Ancestors (in MRO)
- builtins.dict
Class variables
not: airbyte_agent_sdk.connectors.jira.types.IssueCommentsEqCondition | airbyte_agent_sdk.connectors.jira.types.IssueCommentsNeqCondition | airbyte_agent_sdk.connectors.jira.types.IssueCommentsGtCondition | airbyte_agent_sdk.connectors.jira.types.IssueCommentsGteCondition | airbyte_agent_sdk.connectors.jira.types.IssueCommentsLtCondition | airbyte_agent_sdk.connectors.jira.types.IssueCommentsLteCondition | airbyte_agent_sdk.connectors.jira.types.IssueCommentsInCondition | airbyte_agent_sdk.connectors.jira.types.IssueCommentsStartswithCondition | airbyte_agent_sdk.connectors.jira.types.IssueCommentsEndswithCondition | airbyte_agent_sdk.connectors.jira.types.IssueCommentsFuzzyCondition | airbyte_agent_sdk.connectors.jira.types.IssueCommentsKeywordCondition | airbyte_agent_sdk.connectors.jira.types.IssueCommentsContainsCondition | airbyte_agent_sdk.connectors.jira.types.IssueCommentsArrayContainsCondition | airbyte_agent_sdk.connectors.jira.types.IssueCommentsNotCondition | airbyte_agent_sdk.connectors.jira.types.IssueCommentsAndCondition | airbyte_agent_sdk.connectors.jira.types.IssueCommentsOrCondition | airbyte_agent_sdk.connectors.jira.types.IssueCommentsAnyCondition
: The type of the None singleton.
IssueCommentsOrCondition(*args, **kwargs)
: dict() -> new empty dictionary
dict(mapping) -> new dictionary initialized from a mapping object's
(key, value) pairs
dict(iterable) -> new dictionary initialized as if via:
d = {}
for k, v in iterable:
d[k] = v
dict(**kwargs) -> new dictionary initialized with the name=value pairs
in the keyword argument list. For example: dict(one=1, two=2)
Ancestors (in MRO)
- builtins.dict
Class variables
or: list[airbyte_agent_sdk.connectors.jira.types.IssueCommentsEqCondition | airbyte_agent_sdk.connectors.jira.types.IssueCommentsNeqCondition | airbyte_agent_sdk.connectors.jira.types.IssueCommentsGtCondition | airbyte_agent_sdk.connectors.jira.types.IssueCommentsGteCondition | airbyte_agent_sdk.connectors.jira.types.IssueCommentsLtCondition | airbyte_agent_sdk.connectors.jira.types.IssueCommentsLteCondition | airbyte_agent_sdk.connectors.jira.types.IssueCommentsInCondition | airbyte_agent_sdk.connectors.jira.types.IssueCommentsStartswithCondition | airbyte_agent_sdk.connectors.jira.types.IssueCommentsEndswithCondition | airbyte_agent_sdk.connectors.jira.types.IssueCommentsFuzzyCondition | airbyte_agent_sdk.connectors.jira.types.IssueCommentsKeywordCondition | airbyte_agent_sdk.connectors.jira.types.IssueCommentsContainsCondition | airbyte_agent_sdk.connectors.jira.types.IssueCommentsArrayContainsCondition | airbyte_agent_sdk.connectors.jira.types.IssueCommentsNotCondition | airbyte_agent_sdk.connectors.jira.types.IssueCommentsAndCondition | airbyte_agent_sdk.connectors.jira.types.IssueCommentsOrCondition | airbyte_agent_sdk.connectors.jira.types.IssueCommentsAnyCondition]
: The type of the None singleton.
IssueCommentsSearchFilter(*args, **kwargs)
: Available fields for filtering issue_comments search queries.
Ancestors (in MRO)
- builtins.dict
Class variables
author: dict[str, typing.Any] | None
: The ID of the user who created the comment
body: dict[str, typing.Any]
: The comment text in Atlassian Document Format
created: str
: The date and time at which the comment was created
id: str
: The ID of the comment
issue_id: str | None
: Id of the related issue
jsd_public: bool
: Whether the comment is visible in Jira Service Desk
properties: list[typing.Any]
: A list of comment properties
rendered_body: str | None
: The rendered version of the comment
self: str
: The URL of the comment
update_author: dict[str, typing.Any] | None
: The ID of the user who updated the comment last
updated: str
: The date and time at which the comment was updated last
visibility: dict[str, typing.Any] | None
: The group or role to which this item is visible
IssueCommentsSearchQuery(*args, **kwargs)
: Search query for issue_comments entity.
Ancestors (in MRO)
- builtins.dict
Class variables
filter: airbyte_agent_sdk.connectors.jira.types.IssueCommentsEqCondition | airbyte_agent_sdk.connectors.jira.types.IssueCommentsNeqCondition | airbyte_agent_sdk.connectors.jira.types.IssueCommentsGtCondition | airbyte_agent_sdk.connectors.jira.types.IssueCommentsGteCondition | airbyte_agent_sdk.connectors.jira.types.IssueCommentsLtCondition | airbyte_agent_sdk.connectors.jira.types.IssueCommentsLteCondition | airbyte_agent_sdk.connectors.jira.types.IssueCommentsInCondition | airbyte_agent_sdk.connectors.jira.types.IssueCommentsStartswithCondition | airbyte_agent_sdk.connectors.jira.types.IssueCommentsEndswithCondition | airbyte_agent_sdk.connectors.jira.types.IssueCommentsFuzzyCondition | airbyte_agent_sdk.connectors.jira.types.IssueCommentsKeywordCondition | airbyte_agent_sdk.connectors.jira.types.IssueCommentsContainsCondition | airbyte_agent_sdk.connectors.jira.types.IssueCommentsArrayContainsCondition | airbyte_agent_sdk.connectors.jira.types.IssueCommentsNotCondition | airbyte_agent_sdk.connectors.jira.types.IssueCommentsAndCondition | airbyte_agent_sdk.connectors.jira.types.IssueCommentsOrCondition | airbyte_agent_sdk.connectors.jira.types.IssueCommentsAnyCondition
: The type of the None singleton.
sort: list[airbyte_agent_sdk.connectors.jira.types.IssueCommentsSortFilter]
: The type of the None singleton.
IssueCommentsSortFilter(*args, **kwargs)
: Available fields for sorting issue_comments search results.
Ancestors (in MRO)
- builtins.dict
Class variables
author: Literal['asc', 'desc']
: The ID of the user who created the comment
body: Literal['asc', 'desc']
: The comment text in Atlassian Document Format
created: Literal['asc', 'desc']
: The date and time at which the comment was created
id: Literal['asc', 'desc']
: The ID of the comment
issue_id: Literal['asc', 'desc']
: Id of the related issue
jsd_public: Literal['asc', 'desc']
: Whether the comment is visible in Jira Service Desk
properties: Literal['asc', 'desc']
: A list of comment properties
rendered_body: Literal['asc', 'desc']
: The rendered version of the comment
self: Literal['asc', 'desc']
: The URL of the comment
update_author: Literal['asc', 'desc']
: The ID of the user who updated the comment last
updated: Literal['asc', 'desc']
: The date and time at which the comment was updated last
visibility: Literal['asc', 'desc']
: The group or role to which this item is visible
IssueCommentsStartswithCondition(*args, **kwargs)
: Literal case-insensitive prefix match.
Ancestors (in MRO)
- builtins.dict
Class variables
startswith: airbyte_agent_sdk.connectors.jira.types.IssueCommentsStringFilter
: The type of the None singleton.
IssueCommentsStringFilter(*args, **kwargs)
: String fields for text search conditions (startswith, endswith, fuzzy, keyword).
Ancestors (in MRO)
- builtins.dict
Class variables
author: str
: The ID of the user who created the comment
body: str
: The comment text in Atlassian Document Format
created: str
: The date and time at which the comment was created
id: str
: The ID of the comment
issue_id: str
: Id of the related issue
jsd_public: str
: Whether the comment is visible in Jira Service Desk
properties: str
: A list of comment properties
rendered_body: str
: The rendered version of the comment
self: str
: The URL of the comment
update_author: str
: The ID of the user who updated the comment last
updated: str
: The date and time at which the comment was updated last
visibility: str
: The group or role to which this item is visible
IssueCommentsUpdateParams(*args, **kwargs)
: Parameters for issue_comments.update operation
Ancestors (in MRO)
- builtins.dict
Class variables
body: airbyte_agent_sdk.connectors.jira.types.IssueCommentsUpdateParamsBody
: The type of the None singleton.
comment_id: str
: The type of the None singleton.
expand: str
: The type of the None singleton.
issue_id_or_key: str
: The type of the None singleton.
notify_users: bool
: The type of the None singleton.
visibility: airbyte_agent_sdk.connectors.jira.types.IssueCommentsUpdateParamsVisibility
: The type of the None singleton.
IssueCommentsUpdateParamsBody(*args, **kwargs)
: Updated comment content in Atlassian Document Format (ADF)
Ancestors (in MRO)
- builtins.dict
Class variables
content: list[airbyte_agent_sdk.connectors.jira.types.IssueCommentsUpdateParamsBodyContentItem]
: The type of the None singleton.
type_: str
: The type of the None singleton.
version: int
: The type of the None singleton.
IssueCommentsUpdateParamsBodyContentItem(*args, **kwargs)
: Nested schema for IssueCommentsUpdateParamsBody.content_item
Ancestors (in MRO)
- builtins.dict
Class variables
content: list[airbyte_agent_sdk.connectors.jira.types.IssueCommentsUpdateParamsBodyContentItemContentItem]
: The type of the None singleton.
type_: str
: The type of the None singleton.
IssueCommentsUpdateParamsBodyContentItemContentItem(*args, **kwargs)
: Nested schema for IssueCommentsUpdateParamsBodyContentItem.content_item
Ancestors (in MRO)
- builtins.dict
Class variables
text: str
: The type of the None singleton.
type_: str
: The type of the None singleton.
IssueCommentsUpdateParamsVisibility(*args, **kwargs)
: Restrict comment visibility to a group or role
Ancestors (in MRO)
- builtins.dict
Class variables
identifier: str
: The type of the None singleton.
type_: str
: The type of the None singleton.
value: str
: The type of the None singleton.
IssueFieldsAndCondition(*args, **kwargs)
: dict() -> new empty dictionary
dict(mapping) -> new dictionary initialized from a mapping object's
(key, value) pairs
dict(iterable) -> new dictionary initialized as if via:
d = {}
for k, v in iterable:
d[k] = v
dict(**kwargs) -> new dictionary initialized with the name=value pairs
in the keyword argument list. For example: dict(one=1, two=2)
Ancestors (in MRO)
- builtins.dict
Class variables
and: list[airbyte_agent_sdk.connectors.jira.types.IssueFieldsEqCondition | airbyte_agent_sdk.connectors.jira.types.IssueFieldsNeqCondition | airbyte_agent_sdk.connectors.jira.types.IssueFieldsGtCondition | airbyte_agent_sdk.connectors.jira.types.IssueFieldsGteCondition | airbyte_agent_sdk.connectors.jira.types.IssueFieldsLtCondition | airbyte_agent_sdk.connectors.jira.types.IssueFieldsLteCondition | airbyte_agent_sdk.connectors.jira.types.IssueFieldsInCondition | airbyte_agent_sdk.connectors.jira.types.IssueFieldsStartswithCondition | airbyte_agent_sdk.connectors.jira.types.IssueFieldsEndswithCondition | airbyte_agent_sdk.connectors.jira.types.IssueFieldsFuzzyCondition | airbyte_agent_sdk.connectors.jira.types.IssueFieldsKeywordCondition | airbyte_agent_sdk.connectors.jira.types.IssueFieldsContainsCondition | airbyte_agent_sdk.connectors.jira.types.IssueFieldsArrayContainsCondition | airbyte_agent_sdk.connectors.jira.types.IssueFieldsNotCondition | airbyte_agent_sdk.connectors.jira.types.IssueFieldsAndCondition | airbyte_agent_sdk.connectors.jira.types.IssueFieldsOrCondition | airbyte_agent_sdk.connectors.jira.types.IssueFieldsAnyCondition]
: The type of the None singleton.
IssueFieldsAnyCondition(*args, **kwargs)
: dict() -> new empty dictionary
dict(mapping) -> new dictionary initialized from a mapping object's
(key, value) pairs
dict(iterable) -> new dictionary initialized as if via:
d = {}
for k, v in iterable:
d[k] = v
dict(**kwargs) -> new dictionary initialized with the name=value pairs
in the keyword argument list. For example: dict(one=1, two=2)
Ancestors (in MRO)
- builtins.dict
Class variables
any: airbyte_agent_sdk.connectors.jira.types.IssueFieldsAnyValueFilter
: The type of the None singleton.
IssueFieldsAnyValueFilter(*args, **kwargs)
: Available fields with Any value type. Used for 'contains' and 'any' conditions.
Ancestors (in MRO)
- builtins.dict
Class variables
clause_names: Any
: The names that can be used to reference the field in an advanced search
custom: Any
: Whether the field is a custom field
id: Any
: The ID of the field
key: Any
: The key of the field
name: Any
: The name of the field
navigable: Any
: Whether the field can be used as a column on the issue navigator
orderable: Any
: Whether the content of the field can be used to order lists
schema_: Any
: The data schema for the field
scope: Any
: The scope of the field
searchable: Any
: Whether the content of the field can be searched
untranslated_name: Any
: The untranslated name of the field
IssueFieldsArrayContainsCondition(*args, **kwargs)
: Exact membership test on an array field. Example: {"array_contains": {"tags": "premium"}}
Ancestors (in MRO)
- builtins.dict
Class variables
array_contains: airbyte_agent_sdk.connectors.jira.types.IssueFieldsAnyValueFilter
: The type of the None singleton.
IssueFieldsContainsCondition(*args, **kwargs)
: Case-insensitive substring match on a scalar field. Example: {"contains": {"subject": "billing"}}
Ancestors (in MRO)
- builtins.dict
Class variables
contains: airbyte_agent_sdk.connectors.jira.types.IssueFieldsAnyValueFilter
: The type of the None singleton.
IssueFieldsEndswithCondition(*args, **kwargs)
: Literal case-insensitive suffix match.
Ancestors (in MRO)
- builtins.dict
Class variables
endswith: airbyte_agent_sdk.connectors.jira.types.IssueFieldsStringFilter
: The type of the None singleton.
IssueFieldsEqCondition(*args, **kwargs)
: Equal to: field equals value.
Ancestors (in MRO)
- builtins.dict
Class variables
eq: airbyte_agent_sdk.connectors.jira.types.IssueFieldsSearchFilter
: The type of the None singleton.
IssueFieldsFuzzyCondition(*args, **kwargs)
: Ordered word text match (case-insensitive).
Ancestors (in MRO)
- builtins.dict
Class variables
fuzzy: airbyte_agent_sdk.connectors.jira.types.IssueFieldsStringFilter
: The type of the None singleton.
IssueFieldsGtCondition(*args, **kwargs)
: Greater than: field > value.
Ancestors (in MRO)
- builtins.dict
Class variables
gt: airbyte_agent_sdk.connectors.jira.types.IssueFieldsSearchFilter
: The type of the None singleton.
IssueFieldsGteCondition(*args, **kwargs)
: Greater than or equal: field >= value.
Ancestors (in MRO)
- builtins.dict
Class variables
gte: airbyte_agent_sdk.connectors.jira.types.IssueFieldsSearchFilter
: The type of the None singleton.
IssueFieldsInCondition(*args, **kwargs)
: dict() -> new empty dictionary
dict(mapping) -> new dictionary initialized from a mapping object's
(key, value) pairs
dict(iterable) -> new dictionary initialized as if via:
d = {}
for k, v in iterable:
d[k] = v
dict(**kwargs) -> new dictionary initialized with the name=value pairs
in the keyword argument list. For example: dict(one=1, two=2)
Ancestors (in MRO)
- builtins.dict
Class variables
in: airbyte_agent_sdk.connectors.jira.types.IssueFieldsInFilter
: The type of the None singleton.
IssueFieldsInFilter(*args, **kwargs)
: Available fields for 'in' condition (values are lists).
Ancestors (in MRO)
- builtins.dict
Class variables
clause_names: list[list[typing.Any]]
: The names that can be used to reference the field in an advanced search
custom: list[bool]
: Whether the field is a custom field
id: list[str]
: The ID of the field
key: list[str]
: The key of the field
name: list[str]
: The name of the field
navigable: list[bool]
: Whether the field can be used as a column on the issue navigator
orderable: list[bool]
: Whether the content of the field can be used to order lists
schema_: list[dict[str, typing.Any]]
: The data schema for the field
scope: list[dict[str, typing.Any]]
: The scope of the field
searchable: list[bool]
: Whether the content of the field can be searched
untranslated_name: list[str]
: The untranslated name of the field
IssueFieldsKeywordCondition(*args, **kwargs)
: Keyword text match (any word present).
Ancestors (in MRO)
- builtins.dict
Class variables
keyword: airbyte_agent_sdk.connectors.jira.types.IssueFieldsStringFilter
: The type of the None singleton.
IssueFieldsListParams(*args, **kwargs)
: Parameters for issue_fields.list operation
Ancestors (in MRO)
- builtins.dict
IssueFieldsLtCondition(*args, **kwargs)
: Less than: field < value.
Ancestors (in MRO)
- builtins.dict
Class variables
lt: airbyte_agent_sdk.connectors.jira.types.IssueFieldsSearchFilter
: The type of the None singleton.
IssueFieldsLteCondition(*args, **kwargs)
: Less than or equal: field <= value.
Ancestors (in MRO)
- builtins.dict
Class variables
lte: airbyte_agent_sdk.connectors.jira.types.IssueFieldsSearchFilter
: The type of the None singleton.
IssueFieldsNeqCondition(*args, **kwargs)
: Not equal to: field does not equal value.
Ancestors (in MRO)
- builtins.dict
Class variables
neq: airbyte_agent_sdk.connectors.jira.types.IssueFieldsSearchFilter
: The type of the None singleton.
IssueFieldsNotCondition(*args, **kwargs)
: dict() -> new empty dictionary
dict(mapping) -> new dictionary initialized from a mapping object's
(key, value) pairs
dict(iterable) -> new dictionary initialized as if via:
d = {}
for k, v in iterable:
d[k] = v
dict(**kwargs) -> new dictionary initialized with the name=value pairs
in the keyword argument list. For example: dict(one=1, two=2)
Ancestors (in MRO)
- builtins.dict
Class variables
not: airbyte_agent_sdk.connectors.jira.types.IssueFieldsEqCondition | airbyte_agent_sdk.connectors.jira.types.IssueFieldsNeqCondition | airbyte_agent_sdk.connectors.jira.types.IssueFieldsGtCondition | airbyte_agent_sdk.connectors.jira.types.IssueFieldsGteCondition | airbyte_agent_sdk.connectors.jira.types.IssueFieldsLtCondition | airbyte_agent_sdk.connectors.jira.types.IssueFieldsLteCondition | airbyte_agent_sdk.connectors.jira.types.IssueFieldsInCondition | airbyte_agent_sdk.connectors.jira.types.IssueFieldsStartswithCondition | airbyte_agent_sdk.connectors.jira.types.IssueFieldsEndswithCondition | airbyte_agent_sdk.connectors.jira.types.IssueFieldsFuzzyCondition | airbyte_agent_sdk.connectors.jira.types.IssueFieldsKeywordCondition | airbyte_agent_sdk.connectors.jira.types.IssueFieldsContainsCondition | airbyte_agent_sdk.connectors.jira.types.IssueFieldsArrayContainsCondition | airbyte_agent_sdk.connectors.jira.types.IssueFieldsNotCondition | airbyte_agent_sdk.connectors.jira.types.IssueFieldsAndCondition | airbyte_agent_sdk.connectors.jira.types.IssueFieldsOrCondition | airbyte_agent_sdk.connectors.jira.types.IssueFieldsAnyCondition
: The type of the None singleton.
IssueFieldsOrCondition(*args, **kwargs)
: dict() -> new empty dictionary
dict(mapping) -> new dictionary initialized from a mapping object's
(key, value) pairs
dict(iterable) -> new dictionary initialized as if via:
d = {}
for k, v in iterable:
d[k] = v
dict(**kwargs) -> new dictionary initialized with the name=value pairs
in the keyword argument list. For example: dict(one=1, two=2)
Ancestors (in MRO)
- builtins.dict
Class variables
or: list[airbyte_agent_sdk.connectors.jira.types.IssueFieldsEqCondition | airbyte_agent_sdk.connectors.jira.types.IssueFieldsNeqCondition | airbyte_agent_sdk.connectors.jira.types.IssueFieldsGtCondition | airbyte_agent_sdk.connectors.jira.types.IssueFieldsGteCondition | airbyte_agent_sdk.connectors.jira.types.IssueFieldsLtCondition | airbyte_agent_sdk.connectors.jira.types.IssueFieldsLteCondition | airbyte_agent_sdk.connectors.jira.types.IssueFieldsInCondition | airbyte_agent_sdk.connectors.jira.types.IssueFieldsStartswithCondition | airbyte_agent_sdk.connectors.jira.types.IssueFieldsEndswithCondition | airbyte_agent_sdk.connectors.jira.types.IssueFieldsFuzzyCondition | airbyte_agent_sdk.connectors.jira.types.IssueFieldsKeywordCondition | airbyte_agent_sdk.connectors.jira.types.IssueFieldsContainsCondition | airbyte_agent_sdk.connectors.jira.types.IssueFieldsArrayContainsCondition | airbyte_agent_sdk.connectors.jira.types.IssueFieldsNotCondition | airbyte_agent_sdk.connectors.jira.types.IssueFieldsAndCondition | airbyte_agent_sdk.connectors.jira.types.IssueFieldsOrCondition | airbyte_agent_sdk.connectors.jira.types.IssueFieldsAnyCondition]
: The type of the None singleton.
IssueFieldsSearchFilter(*args, **kwargs)
: Available fields for filtering issue_fields search queries.
Ancestors (in MRO)
- builtins.dict
Class variables
clause_names: list[typing.Any]
: The names that can be used to reference the field in an advanced search
custom: bool
: Whether the field is a custom field
id: str
: The ID of the field
key: str | None
: The key of the field
name: str
: The name of the field
navigable: bool
: Whether the field can be used as a column on the issue navigator
orderable: bool
: Whether the content of the field can be used to order lists
schema_: dict[str, typing.Any] | None
: The data schema for the field
scope: dict[str, typing.Any] | None
: The scope of the field
searchable: bool
: Whether the content of the field can be searched
untranslated_name: str | None
: The untranslated name of the field
IssueFieldsSearchParams(*args, **kwargs)
: Parameters for issue_fields.search operation
Ancestors (in MRO)
- builtins.dict
Class variables
expand: str
: The type of the None singleton.
id: list[str]
: The type of the None singleton.
max_results: int
: The type of the None singleton.
order_by: str
: The type of the None singleton.
query: str
: The type of the None singleton.
start_at: int
: The type of the None singleton.
type: list[str]
: The type of the None singleton.
IssueFieldsSearchQuery(*args, **kwargs)
: Search query for issue_fields entity.
Ancestors (in MRO)
- builtins.dict
Class variables
filter: airbyte_agent_sdk.connectors.jira.types.IssueFieldsEqCondition | airbyte_agent_sdk.connectors.jira.types.IssueFieldsNeqCondition | airbyte_agent_sdk.connectors.jira.types.IssueFieldsGtCondition | airbyte_agent_sdk.connectors.jira.types.IssueFieldsGteCondition | airbyte_agent_sdk.connectors.jira.types.IssueFieldsLtCondition | airbyte_agent_sdk.connectors.jira.types.IssueFieldsLteCondition | airbyte_agent_sdk.connectors.jira.types.IssueFieldsInCondition | airbyte_agent_sdk.connectors.jira.types.IssueFieldsStartswithCondition | airbyte_agent_sdk.connectors.jira.types.IssueFieldsEndswithCondition | airbyte_agent_sdk.connectors.jira.types.IssueFieldsFuzzyCondition | airbyte_agent_sdk.connectors.jira.types.IssueFieldsKeywordCondition | airbyte_agent_sdk.connectors.jira.types.IssueFieldsContainsCondition | airbyte_agent_sdk.connectors.jira.types.IssueFieldsArrayContainsCondition | airbyte_agent_sdk.connectors.jira.types.IssueFieldsNotCondition | airbyte_agent_sdk.connectors.jira.types.IssueFieldsAndCondition | airbyte_agent_sdk.connectors.jira.types.IssueFieldsOrCondition | airbyte_agent_sdk.connectors.jira.types.IssueFieldsAnyCondition
: The type of the None singleton.
sort: list[airbyte_agent_sdk.connectors.jira.types.IssueFieldsSortFilter]
: The type of the None singleton.
IssueFieldsSortFilter(*args, **kwargs)
: Available fields for sorting issue_fields search results.
Ancestors (in MRO)
- builtins.dict
Class variables
clause_names: Literal['asc', 'desc']
: The names that can be used to reference the field in an advanced search
custom: Literal['asc', 'desc']
: Whether the field is a custom field
id: Literal['asc', 'desc']
: The ID of the field
key: Literal['asc', 'desc']
: The key of the field
name: Literal['asc', 'desc']
: The name of the field
navigable: Literal['asc', 'desc']
: Whether the field can be used as a column on the issue navigator
orderable: Literal['asc', 'desc']
: Whether the content of the field can be used to order lists
schema_: Literal['asc', 'desc']
: The data schema for the field
scope: Literal['asc', 'desc']
: The scope of the field
searchable: Literal['asc', 'desc']
: Whether the content of the field can be searched
untranslated_name: Literal['asc', 'desc']
: The untranslated name of the field
IssueFieldsStartswithCondition(*args, **kwargs)
: Literal case-insensitive prefix match.
Ancestors (in MRO)
- builtins.dict
Class variables
startswith: airbyte_agent_sdk.connectors.jira.types.IssueFieldsStringFilter
: The type of the None singleton.
IssueFieldsStringFilter(*args, **kwargs)
: String fields for text search conditions (startswith, endswith, fuzzy, keyword).
Ancestors (in MRO)
- builtins.dict
Class variables
clause_names: str
: The names that can be used to reference the field in an advanced search
custom: str
: Whether the field is a custom field
id: str
: The ID of the field
key: str
: The key of the field
name: str
: The name of the field
navigable: str
: Whether the field can be used as a column on the issue navigator
orderable: str
: Whether the content of the field can be used to order lists
schema_: str
: The data schema for the field
scope: str
: The scope of the field
searchable: str
: Whether the content of the field can be searched
untranslated_name: str
: The untranslated name of the field
IssueLinksCreateParams(*args, **kwargs)
: Parameters for issue_links.create operation
Ancestors (in MRO)
- builtins.dict
Class variables
comment: airbyte_agent_sdk.connectors.jira.types.IssueLinksCreateParamsComment
: The type of the None singleton.
inward_issue: airbyte_agent_sdk.connectors.jira.types.IssueLinksCreateParamsInwardissue
: The type of the None singleton.
outward_issue: airbyte_agent_sdk.connectors.jira.types.IssueLinksCreateParamsOutwardissue
: The type of the None singleton.
type: airbyte_agent_sdk.connectors.jira.types.IssueLinksCreateParamsType
: The type of the None singleton.
IssueLinksCreateParamsComment(*args, **kwargs)
: A comment about the link in Atlassian Document Format (ADF)
Ancestors (in MRO)
- builtins.dict
Class variables
body: airbyte_agent_sdk.connectors.jira.types.IssueLinksCreateParamsCommentBody
: The type of the None singleton.
IssueLinksCreateParamsCommentBody(*args, **kwargs)
: Nested schema for IssueLinksCreateParamsComment.body
Ancestors (in MRO)
- builtins.dict
Class variables
content: list[airbyte_agent_sdk.connectors.jira.types.IssueLinksCreateParamsCommentBodyContentItem]
: The type of the None singleton.
type_: str
: The type of the None singleton.
version: int
: The type of the None singleton.
IssueLinksCreateParamsCommentBodyContentItem(*args, **kwargs)
: Nested schema for IssueLinksCreateParamsCommentBody.content_item
Ancestors (in MRO)
- builtins.dict
Class variables
content: list[airbyte_agent_sdk.connectors.jira.types.IssueLinksCreateParamsCommentBodyContentItemContentItem]
: The type of the None singleton.
type_: str
: The type of the None singleton.
IssueLinksCreateParamsCommentBodyContentItemContentItem(*args, **kwargs)
: Nested schema for IssueLinksCreateParamsCommentBodyContentItem.content_item
Ancestors (in MRO)
- builtins.dict
Class variables
text: str
: The type of the None singleton.
type_: str
: The type of the None singleton.
IssueLinksCreateParamsInwardissue(*args, **kwargs)
: The inward issue (the issue that is affected by the link)
Ancestors (in MRO)
- builtins.dict
Class variables
id: str
: The type of the None singleton.
key: str
: The type of the None singleton.
IssueLinksCreateParamsOutwardissue(*args, **kwargs)
: The outward issue (the issue that causes the link)
Ancestors (in MRO)
- builtins.dict
Class variables
id: str
: The type of the None singleton.
key: str
: The type of the None singleton.
IssueLinksCreateParamsType(*args, **kwargs)
: The type of link (e.g., Blocks, Duplicate, Relates)
Ancestors (in MRO)
- builtins.dict
Class variables
id: str
: The type of the None singleton.
inward: str
: The type of the None singleton.
name: str
: The type of the None singleton.
outward: str
: The type of the None singleton.
IssueTransitionsCreateParams(*args, **kwargs)
: Parameters for issue_transitions.create operation
Ancestors (in MRO)
- builtins.dict
Class variables
fields: dict[str, typing.Any]
: The type of the None singleton.
history_metadata: dict[str, typing.Any]
: The type of the None singleton.
issue_id_or_key: str
: The type of the None singleton.
transition: airbyte_agent_sdk.connectors.jira.types.IssueTransitionsCreateParamsTransition
: The type of the None singleton.
Methods
update(...) ‑> dict[str, typing.Any]
: D.update([E, ]**F) -> None. Update D from mapping/iterable E and F.
If E is present and has a .keys() method, then does: for k in E.keys(): D[k] = E[k]
If E is present and lacks a .keys() method, then does: for k, v in E: D[k] = v
In either case, this is followed by: for k in F: D[k] = F[k]
IssueTransitionsCreateParamsTransition(*args, **kwargs)
: The transition to perform
Ancestors (in MRO)
- builtins.dict
Class variables
id: str
: The type of the None singleton.
IssueTransitionsListParams(*args, **kwargs)
: Parameters for issue_transitions.list operation
Ancestors (in MRO)
- builtins.dict
Class variables
expand: str
: The type of the None singleton.
include_unavailable_transitions: bool
: The type of the None singleton.
issue_id_or_key: str
: The type of the None singleton.
skip_remote_only_condition: bool
: The type of the None singleton.
sort_by_ops_bar_and_status: bool
: The type of the None singleton.
transition_id: str
: The type of the None singleton.
IssueWorklogsAndCondition(*args, **kwargs)
: dict() -> new empty dictionary
dict(mapping) -> new dictionary initialized from a mapping object's
(key, value) pairs
dict(iterable) -> new dictionary initialized as if via:
d = {}
for k, v in iterable:
d[k] = v
dict(**kwargs) -> new dictionary initialized with the name=value pairs
in the keyword argument list. For example: dict(one=1, two=2)
Ancestors (in MRO)
- builtins.dict
Class variables
and: list[airbyte_agent_sdk.connectors.jira.types.IssueWorklogsEqCondition | airbyte_agent_sdk.connectors.jira.types.IssueWorklogsNeqCondition | airbyte_agent_sdk.connectors.jira.types.IssueWorklogsGtCondition | airbyte_agent_sdk.connectors.jira.types.IssueWorklogsGteCondition | airbyte_agent_sdk.connectors.jira.types.IssueWorklogsLtCondition | airbyte_agent_sdk.connectors.jira.types.IssueWorklogsLteCondition | airbyte_agent_sdk.connectors.jira.types.IssueWorklogsInCondition | airbyte_agent_sdk.connectors.jira.types.IssueWorklogsStartswithCondition | airbyte_agent_sdk.connectors.jira.types.IssueWorklogsEndswithCondition | airbyte_agent_sdk.connectors.jira.types.IssueWorklogsFuzzyCondition | airbyte_agent_sdk.connectors.jira.types.IssueWorklogsKeywordCondition | airbyte_agent_sdk.connectors.jira.types.IssueWorklogsContainsCondition | airbyte_agent_sdk.connectors.jira.types.IssueWorklogsArrayContainsCondition | airbyte_agent_sdk.connectors.jira.types.IssueWorklogsNotCondition | airbyte_agent_sdk.connectors.jira.types.IssueWorklogsAndCondition | airbyte_agent_sdk.connectors.jira.types.IssueWorklogsOrCondition | airbyte_agent_sdk.connectors.jira.types.IssueWorklogsAnyCondition]
: The type of the None singleton.
IssueWorklogsAnyCondition(*args, **kwargs)
: dict() -> new empty dictionary
dict(mapping) -> new dictionary initialized from a mapping object's
(key, value) pairs
dict(iterable) -> new dictionary initialized as if via:
d = {}
for k, v in iterable:
d[k] = v
dict(**kwargs) -> new dictionary initialized with the name=value pairs
in the keyword argument list. For example: dict(one=1, two=2)
Ancestors (in MRO)
- builtins.dict
Class variables
any: airbyte_agent_sdk.connectors.jira.types.IssueWorklogsAnyValueFilter
: The type of the None singleton.
IssueWorklogsAnyValueFilter(*args, **kwargs)
: Available fields with Any value type. Used for 'contains' and 'any' conditions.
Ancestors (in MRO)
- builtins.dict
Class variables
author: Any
: Details of the user who created the worklog
comment: Any
: A comment about the worklog in Atlassian Document Format
created: Any
: The datetime on which the worklog was created
id: Any
: The ID of the worklog record
issue_id: Any
: The ID of the issue this worklog is for
properties: Any
: Details of properties for the worklog
self: Any
: The URL of the worklog item
started: Any
: The datetime on which the worklog effort was started
time_spent: Any
: The time spent working on the issue as days, hours, or minutes
time_spent_seconds: Any
: The time in seconds spent working on the issue
update_author: Any
: Details of the user who last updated the worklog
updated: Any
: The datetime on which the worklog was last updated
visibility: Any
: Details about any restrictions in the visibility of the worklog
IssueWorklogsArrayContainsCondition(*args, **kwargs)
: Exact membership test on an array field. Example: {"array_contains": {"tags": "premium"}}
Ancestors (in MRO)
- builtins.dict
Class variables
array_contains: airbyte_agent_sdk.connectors.jira.types.IssueWorklogsAnyValueFilter
: The type of the None singleton.
IssueWorklogsContainsCondition(*args, **kwargs)
: Case-insensitive substring match on a scalar field. Example: {"contains": {"subject": "billing"}}
Ancestors (in MRO)
- builtins.dict
Class variables
contains: airbyte_agent_sdk.connectors.jira.types.IssueWorklogsAnyValueFilter
: The type of the None singleton.
IssueWorklogsCreateParams(*args, **kwargs)
: Parameters for issue_worklogs.create operation
Ancestors (in MRO)
- builtins.dict
Class variables
adjust_estimate: str
: The type of the None singleton.
comment: airbyte_agent_sdk.connectors.jira.types.IssueWorklogsCreateParamsComment
: The type of the None singleton.
issue_id_or_key: str
: The type of the None singleton.
notify_users: bool
: The type of the None singleton.
started: str
: The type of the None singleton.
time_spent: str
: The type of the None singleton.
time_spent_seconds: int
: The type of the None singleton.
visibility: airbyte_agent_sdk.connectors.jira.types.IssueWorklogsCreateParamsVisibility
: The type of the None singleton.
IssueWorklogsCreateParamsComment(*args, **kwargs)
: A comment about the work done in Atlassian Document Format (ADF)
Ancestors (in MRO)
- builtins.dict
Class variables
content: list[airbyte_agent_sdk.connectors.jira.types.IssueWorklogsCreateParamsCommentContentItem]
: The type of the None singleton.
type_: str
: The type of the None singleton.
version: int
: The type of the None singleton.
IssueWorklogsCreateParamsCommentContentItem(*args, **kwargs)
: Nested schema for IssueWorklogsCreateParamsComment.content_item
Ancestors (in MRO)
- builtins.dict
Class variables
content: list[airbyte_agent_sdk.connectors.jira.types.IssueWorklogsCreateParamsCommentContentItemContentItem]
: The type of the None singleton.
type_: str
: The type of the None singleton.
IssueWorklogsCreateParamsCommentContentItemContentItem(*args, **kwargs)
: Nested schema for IssueWorklogsCreateParamsCommentContentItem.content_item
Ancestors (in MRO)
- builtins.dict
Class variables
text: str
: The type of the None singleton.
type_: str
: The type of the None singleton.
IssueWorklogsCreateParamsVisibility(*args, **kwargs)
: Restrict worklog visibility to a group or role
Ancestors (in MRO)
- builtins.dict
Class variables
identifier: str
: The type of the None singleton.
type_: str
: The type of the None singleton.
value: str
: The type of the None singleton.
IssueWorklogsEndswithCondition(*args, **kwargs)
: Literal case-insensitive suffix match.
Ancestors (in MRO)
- builtins.dict
Class variables
endswith: airbyte_agent_sdk.connectors.jira.types.IssueWorklogsStringFilter
: The type of the None singleton.
IssueWorklogsEqCondition(*args, **kwargs)
: Equal to: field equals value.
Ancestors (in MRO)
- builtins.dict
Class variables
eq: airbyte_agent_sdk.connectors.jira.types.IssueWorklogsSearchFilter
: The type of the None singleton.
IssueWorklogsFuzzyCondition(*args, **kwargs)
: Ordered word text match (case-insensitive).
Ancestors (in MRO)
- builtins.dict
Class variables
fuzzy: airbyte_agent_sdk.connectors.jira.types.IssueWorklogsStringFilter
: The type of the None singleton.
IssueWorklogsGetParams(*args, **kwargs)
: Parameters for issue_worklogs.get operation
Ancestors (in MRO)
- builtins.dict
Class variables
expand: str
: The type of the None singleton.
issue_id_or_key: str
: The type of the None singleton.
worklog_id: str
: The type of the None singleton.
IssueWorklogsGtCondition(*args, **kwargs)
: Greater than: field > value.
Ancestors (in MRO)
- builtins.dict
Class variables
gt: airbyte_agent_sdk.connectors.jira.types.IssueWorklogsSearchFilter
: The type of the None singleton.
IssueWorklogsGteCondition(*args, **kwargs)
: Greater than or equal: field >= value.
Ancestors (in MRO)
- builtins.dict
Class variables
gte: airbyte_agent_sdk.connectors.jira.types.IssueWorklogsSearchFilter
: The type of the None singleton.
IssueWorklogsInCondition(*args, **kwargs)
: dict() -> new empty dictionary
dict(mapping) -> new dictionary initialized from a mapping object's
(key, value) pairs
dict(iterable) -> new dictionary initialized as if via:
d = {}
for k, v in iterable:
d[k] = v
dict(**kwargs) -> new dictionary initialized with the name=value pairs
in the keyword argument list. For example: dict(one=1, two=2)
Ancestors (in MRO)
- builtins.dict
Class variables
in: airbyte_agent_sdk.connectors.jira.types.IssueWorklogsInFilter
: The type of the None singleton.
IssueWorklogsInFilter(*args, **kwargs)
: Available fields for 'in' condition (values are lists).
Ancestors (in MRO)
- builtins.dict
Class variables
author: list[dict[str, typing.Any]]
: Details of the user who created the worklog
comment: list[dict[str, typing.Any]]
: A comment about the worklog in Atlassian Document Format
created: list[str]
: The datetime on which the worklog was created
id: list[str]
: The ID of the worklog record
issue_id: list[str]
: The ID of the issue this worklog is for
properties: list[list[typing.Any]]
: Details of properties for the worklog
self: list[str]
: The URL of the worklog item
started: list[str]
: The datetime on which the worklog effort was started
time_spent: list[str]
: The time spent working on the issue as days, hours, or minutes
time_spent_seconds: list[int]
: The time in seconds spent working on the issue
update_author: list[dict[str, typing.Any]]
: Details of the user who last updated the worklog
updated: list[str]
: The datetime on which the worklog was last updated
visibility: list[dict[str, typing.Any]]
: Details about any restrictions in the visibility of the worklog
IssueWorklogsKeywordCondition(*args, **kwargs)
: Keyword text match (any word present).
Ancestors (in MRO)
- builtins.dict
Class variables
keyword: airbyte_agent_sdk.connectors.jira.types.IssueWorklogsStringFilter
: The type of the None singleton.
IssueWorklogsListParams(*args, **kwargs)
: Parameters for issue_worklogs.list operation
Ancestors (in MRO)
- builtins.dict
Class variables
expand: str
: The type of the None singleton.
issue_id_or_key: str
: The type of the None singleton.
max_results: int
: The type of the None singleton.
start_at: int
: The type of the None singleton.
IssueWorklogsLtCondition(*args, **kwargs)
: Less than: field < value.
Ancestors (in MRO)
- builtins.dict
Class variables
lt: airbyte_agent_sdk.connectors.jira.types.IssueWorklogsSearchFilter
: The type of the None singleton.
IssueWorklogsLteCondition(*args, **kwargs)
: Less than or equal: field <= value.
Ancestors (in MRO)
- builtins.dict
Class variables
lte: airbyte_agent_sdk.connectors.jira.types.IssueWorklogsSearchFilter
: The type of the None singleton.
IssueWorklogsNeqCondition(*args, **kwargs)
: Not equal to: field does not equal value.
Ancestors (in MRO)
- builtins.dict
Class variables
neq: airbyte_agent_sdk.connectors.jira.types.IssueWorklogsSearchFilter
: The type of the None singleton.
IssueWorklogsNotCondition(*args, **kwargs)
: dict() -> new empty dictionary
dict(mapping) -> new dictionary initialized from a mapping object's
(key, value) pairs
dict(iterable) -> new dictionary initialized as if via:
d = {}
for k, v in iterable:
d[k] = v
dict(**kwargs) -> new dictionary initialized with the name=value pairs
in the keyword argument list. For example: dict(one=1, two=2)
Ancestors (in MRO)
- builtins.dict
Class variables
not: airbyte_agent_sdk.connectors.jira.types.IssueWorklogsEqCondition | airbyte_agent_sdk.connectors.jira.types.IssueWorklogsNeqCondition | airbyte_agent_sdk.connectors.jira.types.IssueWorklogsGtCondition | airbyte_agent_sdk.connectors.jira.types.IssueWorklogsGteCondition | airbyte_agent_sdk.connectors.jira.types.IssueWorklogsLtCondition | airbyte_agent_sdk.connectors.jira.types.IssueWorklogsLteCondition | airbyte_agent_sdk.connectors.jira.types.IssueWorklogsInCondition | airbyte_agent_sdk.connectors.jira.types.IssueWorklogsStartswithCondition | airbyte_agent_sdk.connectors.jira.types.IssueWorklogsEndswithCondition | airbyte_agent_sdk.connectors.jira.types.IssueWorklogsFuzzyCondition | airbyte_agent_sdk.connectors.jira.types.IssueWorklogsKeywordCondition | airbyte_agent_sdk.connectors.jira.types.IssueWorklogsContainsCondition | airbyte_agent_sdk.connectors.jira.types.IssueWorklogsArrayContainsCondition | airbyte_agent_sdk.connectors.jira.types.IssueWorklogsNotCondition | airbyte_agent_sdk.connectors.jira.types.IssueWorklogsAndCondition | airbyte_agent_sdk.connectors.jira.types.IssueWorklogsOrCondition | airbyte_agent_sdk.connectors.jira.types.IssueWorklogsAnyCondition
: The type of the None singleton.
IssueWorklogsOrCondition(*args, **kwargs)
: dict() -> new empty dictionary
dict(mapping) -> new dictionary initialized from a mapping object's
(key, value) pairs
dict(iterable) -> new dictionary initialized as if via:
d = {}
for k, v in iterable:
d[k] = v
dict(**kwargs) -> new dictionary initialized with the name=value pairs
in the keyword argument list. For example: dict(one=1, two=2)
Ancestors (in MRO)
- builtins.dict
Class variables
or: list[airbyte_agent_sdk.connectors.jira.types.IssueWorklogsEqCondition | airbyte_agent_sdk.connectors.jira.types.IssueWorklogsNeqCondition | airbyte_agent_sdk.connectors.jira.types.IssueWorklogsGtCondition | airbyte_agent_sdk.connectors.jira.types.IssueWorklogsGteCondition | airbyte_agent_sdk.connectors.jira.types.IssueWorklogsLtCondition | airbyte_agent_sdk.connectors.jira.types.IssueWorklogsLteCondition | airbyte_agent_sdk.connectors.jira.types.IssueWorklogsInCondition | airbyte_agent_sdk.connectors.jira.types.IssueWorklogsStartswithCondition | airbyte_agent_sdk.connectors.jira.types.IssueWorklogsEndswithCondition | airbyte_agent_sdk.connectors.jira.types.IssueWorklogsFuzzyCondition | airbyte_agent_sdk.connectors.jira.types.IssueWorklogsKeywordCondition | airbyte_agent_sdk.connectors.jira.types.IssueWorklogsContainsCondition | airbyte_agent_sdk.connectors.jira.types.IssueWorklogsArrayContainsCondition | airbyte_agent_sdk.connectors.jira.types.IssueWorklogsNotCondition | airbyte_agent_sdk.connectors.jira.types.IssueWorklogsAndCondition | airbyte_agent_sdk.connectors.jira.types.IssueWorklogsOrCondition | airbyte_agent_sdk.connectors.jira.types.IssueWorklogsAnyCondition]
: The type of the None singleton.
IssueWorklogsSearchFilter(*args, **kwargs)
: Available fields for filtering issue_worklogs search queries.
Ancestors (in MRO)
- builtins.dict
Class variables
author: dict[str, typing.Any]
: Details of the user who created the worklog
comment: dict[str, typing.Any] | None
: A comment about the worklog in Atlassian Document Format
created: str
: The datetime on which the worklog was created
id: str
: The ID of the worklog record
issue_id: str
: The ID of the issue this worklog is for
properties: list[typing.Any]
: Details of properties for the worklog
self: str
: The URL of the worklog item
started: str
: The datetime on which the worklog effort was started
time_spent: str | None
: The time spent working on the issue as days, hours, or minutes
time_spent_seconds: int
: The time in seconds spent working on the issue
update_author: dict[str, typing.Any] | None
: Details of the user who last updated the worklog
updated: str
: The datetime on which the worklog was last updated
visibility: dict[str, typing.Any] | None
: Details about any restrictions in the visibility of the worklog
IssueWorklogsSearchQuery(*args, **kwargs)
: Search query for issue_worklogs entity.
Ancestors (in MRO)
- builtins.dict
Class variables
filter: airbyte_agent_sdk.connectors.jira.types.IssueWorklogsEqCondition | airbyte_agent_sdk.connectors.jira.types.IssueWorklogsNeqCondition | airbyte_agent_sdk.connectors.jira.types.IssueWorklogsGtCondition | airbyte_agent_sdk.connectors.jira.types.IssueWorklogsGteCondition | airbyte_agent_sdk.connectors.jira.types.IssueWorklogsLtCondition | airbyte_agent_sdk.connectors.jira.types.IssueWorklogsLteCondition | airbyte_agent_sdk.connectors.jira.types.IssueWorklogsInCondition | airbyte_agent_sdk.connectors.jira.types.IssueWorklogsStartswithCondition | airbyte_agent_sdk.connectors.jira.types.IssueWorklogsEndswithCondition | airbyte_agent_sdk.connectors.jira.types.IssueWorklogsFuzzyCondition | airbyte_agent_sdk.connectors.jira.types.IssueWorklogsKeywordCondition | airbyte_agent_sdk.connectors.jira.types.IssueWorklogsContainsCondition | airbyte_agent_sdk.connectors.jira.types.IssueWorklogsArrayContainsCondition | airbyte_agent_sdk.connectors.jira.types.IssueWorklogsNotCondition | airbyte_agent_sdk.connectors.jira.types.IssueWorklogsAndCondition | airbyte_agent_sdk.connectors.jira.types.IssueWorklogsOrCondition | airbyte_agent_sdk.connectors.jira.types.IssueWorklogsAnyCondition
: The type of the None singleton.
sort: list[airbyte_agent_sdk.connectors.jira.types.IssueWorklogsSortFilter]
: The type of the None singleton.
IssueWorklogsSortFilter(*args, **kwargs)
: Available fields for sorting issue_worklogs search results.
Ancestors (in MRO)
- builtins.dict
Class variables
author: Literal['asc', 'desc']
: Details of the user who created the worklog
comment: Literal['asc', 'desc']
: A comment about the worklog in Atlassian Document Format
created: Literal['asc', 'desc']
: The datetime on which the worklog was created
id: Literal['asc', 'desc']
: The ID of the worklog record
issue_id: Literal['asc', 'desc']
: The ID of the issue this worklog is for
properties: Literal['asc', 'desc']
: Details of properties for the worklog
self: Literal['asc', 'desc']
: The URL of the worklog item
started: Literal['asc', 'desc']
: The datetime on which the worklog effort was started
time_spent: Literal['asc', 'desc']
: The time spent working on the issue as days, hours, or minutes
time_spent_seconds: Literal['asc', 'desc']
: The time in seconds spent working on the issue
update_author: Literal['asc', 'desc']
: Details of the user who last updated the worklog
updated: Literal['asc', 'desc']
: The datetime on which the worklog was last updated
visibility: Literal['asc', 'desc']
: Details about any restrictions in the visibility of the worklog
IssueWorklogsStartswithCondition(*args, **kwargs)
: Literal case-insensitive prefix match.
Ancestors (in MRO)
- builtins.dict
Class variables
startswith: airbyte_agent_sdk.connectors.jira.types.IssueWorklogsStringFilter
: The type of the None singleton.
IssueWorklogsStringFilter(*args, **kwargs)
: String fields for text search conditions (startswith, endswith, fuzzy, keyword).
Ancestors (in MRO)
- builtins.dict
Class variables
author: str
: Details of the user who created the worklog
comment: str
: A comment about the worklog in Atlassian Document Format
created: str
: The datetime on which the worklog was created
id: str
: The ID of the worklog record
issue_id: str
: The ID of the issue this worklog is for
properties: str
: Details of properties for the worklog
self: str
: The URL of the worklog item
started: str
: The datetime on which the worklog effort was started
time_spent: str
: The time spent working on the issue as days, hours, or minutes
time_spent_seconds: str
: The time in seconds spent working on the issue
update_author: str
: Details of the user who last updated the worklog
updated: str
: The datetime on which the worklog was last updated
visibility: str
: Details about any restrictions in the visibility of the worklog
IssuesAndCondition(*args, **kwargs)
: dict() -> new empty dictionary
dict(mapping) -> new dictionary initialized from a mapping object's
(key, value) pairs
dict(iterable) -> new dictionary initialized as if via:
d = {}
for k, v in iterable:
d[k] = v
dict(**kwargs) -> new dictionary initialized with the name=value pairs
in the keyword argument list. For example: dict(one=1, two=2)
Ancestors (in MRO)
- builtins.dict
Class variables
and: list[airbyte_agent_sdk.connectors.jira.types.IssuesEqCondition | airbyte_agent_sdk.connectors.jira.types.IssuesNeqCondition | airbyte_agent_sdk.connectors.jira.types.IssuesGtCondition | airbyte_agent_sdk.connectors.jira.types.IssuesGteCondition | airbyte_agent_sdk.connectors.jira.types.IssuesLtCondition | airbyte_agent_sdk.connectors.jira.types.IssuesLteCondition | airbyte_agent_sdk.connectors.jira.types.IssuesInCondition | airbyte_agent_sdk.connectors.jira.types.IssuesStartswithCondition | airbyte_agent_sdk.connectors.jira.types.IssuesEndswithCondition | airbyte_agent_sdk.connectors.jira.types.IssuesFuzzyCondition | airbyte_agent_sdk.connectors.jira.types.IssuesKeywordCondition | airbyte_agent_sdk.connectors.jira.types.IssuesContainsCondition | airbyte_agent_sdk.connectors.jira.types.IssuesArrayContainsCondition | airbyte_agent_sdk.connectors.jira.types.IssuesNotCondition | airbyte_agent_sdk.connectors.jira.types.IssuesAndCondition | airbyte_agent_sdk.connectors.jira.types.IssuesOrCondition | airbyte_agent_sdk.connectors.jira.types.IssuesAnyCondition]
: The type of the None singleton.
IssuesAnyCondition(*args, **kwargs)
: dict() -> new empty dictionary
dict(mapping) -> new dictionary initialized from a mapping object's
(key, value) pairs
dict(iterable) -> new dictionary initialized as if via:
d = {}
for k, v in iterable:
d[k] = v
dict(**kwargs) -> new dictionary initialized with the name=value pairs
in the keyword argument list. For example: dict(one=1, two=2)
Ancestors (in MRO)
- builtins.dict
Class variables
any: airbyte_agent_sdk.connectors.jira.types.IssuesAnyValueFilter
: The type of the None singleton.
IssuesAnyValueFilter(*args, **kwargs)
: Available fields with Any value type. Used for 'contains' and 'any' conditions.
Ancestors (in MRO)
- builtins.dict
Class variables
changelog: Any
: Details of changelogs associated with the issue
created: Any
: The timestamp when the issue was created
editmeta: Any
: The metadata for the fields on the issue that can be amended
expand: Any
: Expand options that include additional issue details in the response
fields: Any
: Details of various fields associated with the issue
fields_to_include: Any
: Specify the fields to include in the fetched issues data
id: Any
: The unique ID of the issue
key: Any
: The unique key of the issue
names: Any
: The ID and name of each field present on the issue
operations: Any
: The operations that can be performed on the issue
project_id: Any
: The ID of the project containing the issue
project_key: Any
: The key of the project containing the issue
properties: Any
: Details of the issue properties identified in the request
rendered_fields: Any
: The rendered value of each field present on the issue
schema_: Any
: The schema describing each field present on the issue
self: Any
: The URL of the issue details
transitions: Any
: The transitions that can be performed on the issue
updated: Any
: The timestamp when the issue was last updated
versioned_representations: Any
: The versions of each field on the issue
IssuesArrayContainsCondition(*args, **kwargs)
: Exact membership test on an array field. Example: {"array_contains": {"tags": "premium"}}
Ancestors (in MRO)
- builtins.dict
Class variables
array_contains: airbyte_agent_sdk.connectors.jira.types.IssuesAnyValueFilter
: The type of the None singleton.
IssuesAssigneeUpdateParams(*args, **kwargs)
: Parameters for issues_assignee.update operation
Ancestors (in MRO)
- builtins.dict
Class variables
account_id: str
: The type of the None singleton.
issue_id_or_key: str
: The type of the None singleton.
IssuesContainsCondition(*args, **kwargs)
: Case-insensitive substring match on a scalar field. Example: {"contains": {"subject": "billing"}}
Ancestors (in MRO)
- builtins.dict
Class variables
contains: airbyte_agent_sdk.connectors.jira.types.IssuesAnyValueFilter
: The type of the None singleton.
IssuesCreateParams(*args, **kwargs)
: Parameters for issues.create operation
Ancestors (in MRO)
- builtins.dict
Class variables
fields: airbyte_agent_sdk.connectors.jira.types.IssuesCreateParamsFields
: The type of the None singleton.
update_history: bool
: The type of the None singleton.
Methods
update(...) ‑> dict[str, typing.Any]
: D.update([E, ]**F) -> None. Update D from mapping/iterable E and F.
If E is present and has a .keys() method, then does: for k in E.keys(): D[k] = E[k]
If E is present and lacks a .keys() method, then does: for k, v in E: D[k] = v
In either case, this is followed by: for k in F: D[k] = F[k]
IssuesCreateParamsFields(*args, **kwargs)
: The issue fields to set
Ancestors (in MRO)
- builtins.dict
Class variables
assignee: airbyte_agent_sdk.connectors.jira.types.IssuesCreateParamsFieldsAssignee
: The type of the None singleton.
description: airbyte_agent_sdk.connectors.jira.types.IssuesCreateParamsFieldsDescription
: The type of the None singleton.
issuetype: airbyte_agent_sdk.connectors.jira.types.IssuesCreateParamsFieldsIssuetype
: The type of the None singleton.
labels: list[str]
: The type of the None singleton.
parent: airbyte_agent_sdk.connectors.jira.types.IssuesCreateParamsFieldsParent
: The type of the None singleton.
priority: airbyte_agent_sdk.connectors.jira.types.IssuesCreateParamsFieldsPriority
: The type of the None singleton.
project: airbyte_agent_sdk.connectors.jira.types.IssuesCreateParamsFieldsProject
: The type of the None singleton.
summary: str
: The type of the None singleton.
IssuesCreateParamsFieldsAssignee(*args, **kwargs)
: The user to assign the issue to
Ancestors (in MRO)
- builtins.dict
Class variables
account_id: str
: The type of the None singleton.
IssuesCreateParamsFieldsDescription(*args, **kwargs)
: Issue description in Atlassian Document Format (ADF)
Ancestors (in MRO)
- builtins.dict
Class variables
content: list[airbyte_agent_sdk.connectors.jira.types.IssuesCreateParamsFieldsDescriptionContentItem]
: The type of the None singleton.
type_: str
: The type of the None singleton.
version: int
: The type of the None singleton.
IssuesCreateParamsFieldsDescriptionContentItem(*args, **kwargs)
: Nested schema for IssuesCreateParamsFieldsDescription.content_item
Ancestors (in MRO)
- builtins.dict
Class variables
content: list[airbyte_agent_sdk.connectors.jira.types.IssuesCreateParamsFieldsDescriptionContentItemContentItem]
: The type of the None singleton.
type_: str
: The type of the None singleton.
IssuesCreateParamsFieldsDescriptionContentItemContentItem(*args, **kwargs)
: Nested schema for IssuesCreateParamsFieldsDescriptionContentItem.content_item
Ancestors (in MRO)
- builtins.dict
Class variables
text: str
: The type of the None singleton.
type_: str
: The type of the None singleton.
IssuesCreateParamsFieldsIssuetype(*args, **kwargs)
: The type of issue (e.g., Bug, Task, Story)
Ancestors (in MRO)
- builtins.dict
Class variables
id: str
: The type of the None singleton.
name: str
: The type of the None singleton.
IssuesCreateParamsFieldsParent(*args, **kwargs)
: Parent issue for subtasks
Ancestors (in MRO)
- builtins.dict
Class variables
key: str
: The type of the None singleton.
IssuesCreateParamsFieldsPriority(*args, **kwargs)
: Issue priority
Ancestors (in MRO)
- builtins.dict
Class variables
id: str
: The type of the None singleton.
name: str
: The type of the None singleton.
IssuesCreateParamsFieldsProject(*args, **kwargs)
: The project to create the issue in
Ancestors (in MRO)
- builtins.dict
Class variables
id: str
: The type of the None singleton.
key: str
: The type of the None singleton.
IssuesDeleteParams(*args, **kwargs)
: Parameters for issues.delete operation
Ancestors (in MRO)
- builtins.dict
Class variables
delete_subtasks: bool
: The type of the None singleton.
issue_id_or_key: str
: The type of the None singleton.
IssuesEndswithCondition(*args, **kwargs)
: Literal case-insensitive suffix match.
Ancestors (in MRO)
- builtins.dict
Class variables
endswith: airbyte_agent_sdk.connectors.jira.types.IssuesStringFilter
: The type of the None singleton.
IssuesEqCondition(*args, **kwargs)
: Equal to: field equals value.
Ancestors (in MRO)
- builtins.dict
Class variables
eq: airbyte_agent_sdk.connectors.jira.types.IssuesSearchFilter
: The type of the None singleton.
IssuesFuzzyCondition(*args, **kwargs)
: Ordered word text match (case-insensitive).
Ancestors (in MRO)
- builtins.dict
Class variables
fuzzy: airbyte_agent_sdk.connectors.jira.types.IssuesStringFilter
: The type of the None singleton.
IssuesGetParams(*args, **kwargs)
: Parameters for issues.get operation
Ancestors (in MRO)
- builtins.dict
Class variables
expand: str
: The type of the None singleton.
fail_fast: bool
: The type of the None singleton.
fields: str
: The type of the None singleton.
fields_by_keys: bool
: The type of the None singleton.
issue_id_or_key: str
: The type of the None singleton.
properties: str
: The type of the None singleton.
update_history: bool
: The type of the None singleton.
IssuesGtCondition(*args, **kwargs)
: Greater than: field > value.
Ancestors (in MRO)
- builtins.dict
Class variables
gt: airbyte_agent_sdk.connectors.jira.types.IssuesSearchFilter
: The type of the None singleton.
IssuesGteCondition(*args, **kwargs)
: Greater than or equal: field >= value.
Ancestors (in MRO)
- builtins.dict
Class variables
gte: airbyte_agent_sdk.connectors.jira.types.IssuesSearchFilter
: The type of the None singleton.
IssuesInCondition(*args, **kwargs)
: dict() -> new empty dictionary
dict(mapping) -> new dictionary initialized from a mapping object's
(key, value) pairs
dict(iterable) -> new dictionary initialized as if via:
d = {}
for k, v in iterable:
d[k] = v
dict(**kwargs) -> new dictionary initialized with the name=value pairs
in the keyword argument list. For example: dict(one=1, two=2)
Ancestors (in MRO)
- builtins.dict
Class variables
in: airbyte_agent_sdk.connectors.jira.types.IssuesInFilter
: The type of the None singleton.
IssuesInFilter(*args, **kwargs)
: Available fields for 'in' condition (values are lists).
Ancestors (in MRO)
- builtins.dict
Class variables
changelog: list[dict[str, typing.Any]]
: Details of changelogs associated with the issue
created: list[str]
: The timestamp when the issue was created
editmeta: list[dict[str, typing.Any]]
: The metadata for the fields on the issue that can be amended
expand: list[str]
: Expand options that include additional issue details in the response
fields: list[dict[str, typing.Any]]
: Details of various fields associated with the issue
fields_to_include: list[dict[str, typing.Any]]
: Specify the fields to include in the fetched issues data
id: list[str]
: The unique ID of the issue
key: list[str]
: The unique key of the issue
names: list[dict[str, typing.Any]]
: The ID and name of each field present on the issue
operations: list[dict[str, typing.Any]]
: The operations that can be performed on the issue
project_id: list[str]
: The ID of the project containing the issue
project_key: list[str]
: The key of the project containing the issue
properties: list[dict[str, typing.Any]]
: Details of the issue properties identified in the request
rendered_fields: list[dict[str, typing.Any]]
: The rendered value of each field present on the issue
schema_: list[dict[str, typing.Any]]
: The schema describing each field present on the issue
self: list[str]
: The URL of the issue details
transitions: list[list[typing.Any]]
: The transitions that can be performed on the issue
updated: list[str]
: The timestamp when the issue was last updated
versioned_representations: list[dict[str, typing.Any]]
: The versions of each field on the issue
IssuesKeywordCondition(*args, **kwargs)
: Keyword text match (any word present).
Ancestors (in MRO)
- builtins.dict
Class variables
keyword: airbyte_agent_sdk.connectors.jira.types.IssuesStringFilter
: The type of the None singleton.
IssuesLtCondition(*args, **kwargs)
: Less than: field < value.
Ancestors (in MRO)
- builtins.dict
Class variables
lt: airbyte_agent_sdk.connectors.jira.types.IssuesSearchFilter
: The type of the None singleton.
IssuesLteCondition(*args, **kwargs)
: Less than or equal: field <= value.
Ancestors (in MRO)
- builtins.dict
Class variables
lte: airbyte_agent_sdk.connectors.jira.types.IssuesSearchFilter
: The type of the None singleton.
IssuesNeqCondition(*args, **kwargs)
: Not equal to: field does not equal value.
Ancestors (in MRO)
- builtins.dict
Class variables
neq: airbyte_agent_sdk.connectors.jira.types.IssuesSearchFilter
: The type of the None singleton.
IssuesNotCondition(*args, **kwargs)
: dict() -> new empty dictionary
dict(mapping) -> new dictionary initialized from a mapping object's
(key, value) pairs
dict(iterable) -> new dictionary initialized as if via:
d = {}
for k, v in iterable:
d[k] = v
dict(**kwargs) -> new dictionary initialized with the name=value pairs
in the keyword argument list. For example: dict(one=1, two=2)
Ancestors (in MRO)
- builtins.dict
Class variables
not: airbyte_agent_sdk.connectors.jira.types.IssuesEqCondition | airbyte_agent_sdk.connectors.jira.types.IssuesNeqCondition | airbyte_agent_sdk.connectors.jira.types.IssuesGtCondition | airbyte_agent_sdk.connectors.jira.types.IssuesGteCondition | airbyte_agent_sdk.connectors.jira.types.IssuesLtCondition | airbyte_agent_sdk.connectors.jira.types.IssuesLteCondition | airbyte_agent_sdk.connectors.jira.types.IssuesInCondition | airbyte_agent_sdk.connectors.jira.types.IssuesStartswithCondition | airbyte_agent_sdk.connectors.jira.types.IssuesEndswithCondition | airbyte_agent_sdk.connectors.jira.types.IssuesFuzzyCondition | airbyte_agent_sdk.connectors.jira.types.IssuesKeywordCondition | airbyte_agent_sdk.connectors.jira.types.IssuesContainsCondition | airbyte_agent_sdk.connectors.jira.types.IssuesArrayContainsCondition | airbyte_agent_sdk.connectors.jira.types.IssuesNotCondition | airbyte_agent_sdk.connectors.jira.types.IssuesAndCondition | airbyte_agent_sdk.connectors.jira.types.IssuesOrCondition | airbyte_agent_sdk.connectors.jira.types.IssuesAnyCondition
: The type of the None singleton.
IssuesOrCondition(*args, **kwargs)
: dict() -> new empty dictionary
dict(mapping) -> new dictionary initialized from a mapping object's
(key, value) pairs
dict(iterable) -> new dictionary initialized as if via:
d = {}
for k, v in iterable:
d[k] = v
dict(**kwargs) -> new dictionary initialized with the name=value pairs
in the keyword argument list. For example: dict(one=1, two=2)
Ancestors (in MRO)
- builtins.dict
Class variables
or: list[airbyte_agent_sdk.connectors.jira.types.IssuesEqCondition | airbyte_agent_sdk.connectors.jira.types.IssuesNeqCondition | airbyte_agent_sdk.connectors.jira.types.IssuesGtCondition | airbyte_agent_sdk.connectors.jira.types.IssuesGteCondition | airbyte_agent_sdk.connectors.jira.types.IssuesLtCondition | airbyte_agent_sdk.connectors.jira.types.IssuesLteCondition | airbyte_agent_sdk.connectors.jira.types.IssuesInCondition | airbyte_agent_sdk.connectors.jira.types.IssuesStartswithCondition | airbyte_agent_sdk.connectors.jira.types.IssuesEndswithCondition | airbyte_agent_sdk.connectors.jira.types.IssuesFuzzyCondition | airbyte_agent_sdk.connectors.jira.types.IssuesKeywordCondition | airbyte_agent_sdk.connectors.jira.types.IssuesContainsCondition | airbyte_agent_sdk.connectors.jira.types.IssuesArrayContainsCondition | airbyte_agent_sdk.connectors.jira.types.IssuesNotCondition | airbyte_agent_sdk.connectors.jira.types.IssuesAndCondition | airbyte_agent_sdk.connectors.jira.types.IssuesOrCondition | airbyte_agent_sdk.connectors.jira.types.IssuesAnyCondition]
: The type of the None singleton.
IssuesSearchFilter(*args, **kwargs)
: Available fields for filtering issues search queries.
Ancestors (in MRO)
- builtins.dict
Class variables
changelog: dict[str, typing.Any] | None
: Details of changelogs associated with the issue
created: str | None
: The timestamp when the issue was created
editmeta: dict[str, typing.Any] | None
: The metadata for the fields on the issue that can be amended
expand: str
: Expand options that include additional issue details in the response
fields: dict[str, typing.Any]
: Details of various fields associated with the issue
fields_to_include: dict[str, typing.Any]
: Specify the fields to include in the fetched issues data
id: str
: The unique ID of the issue
key: str
: The unique key of the issue
names: dict[str, typing.Any]
: The ID and name of each field present on the issue
operations: dict[str, typing.Any] | None
: The operations that can be performed on the issue
project_id: str
: The ID of the project containing the issue
project_key: str
: The key of the project containing the issue
properties: dict[str, typing.Any]
: Details of the issue properties identified in the request
rendered_fields: dict[str, typing.Any]
: The rendered value of each field present on the issue
schema_: dict[str, typing.Any]
: The schema describing each field present on the issue
self: str
: The URL of the issue details
transitions: list[typing.Any]
: The transitions that can be performed on the issue
updated: str | None
: The timestamp when the issue was last updated
versioned_representations: dict[str, typing.Any]
: The versions of each field on the issue
IssuesSearchParams(*args, **kwargs)
: Parameters for issues.search operation
Ancestors (in MRO)
- builtins.dict
Class variables
expand: str
: The type of the None singleton.
fail_fast: bool
: The type of the None singleton.
fields: str
: The type of the None singleton.
fields_by_keys: bool
: The type of the None singleton.
jql: str
: The type of the None singleton.
max_results: int
: The type of the None singleton.
next_page_token: str
: The type of the None singleton.
properties: str
: The type of the None singleton.
IssuesSearchQuery(*args, **kwargs)
: Search query for issues entity.
Ancestors (in MRO)
- builtins.dict
Class variables
filter: airbyte_agent_sdk.connectors.jira.types.IssuesEqCondition | airbyte_agent_sdk.connectors.jira.types.IssuesNeqCondition | airbyte_agent_sdk.connectors.jira.types.IssuesGtCondition | airbyte_agent_sdk.connectors.jira.types.IssuesGteCondition | airbyte_agent_sdk.connectors.jira.types.IssuesLtCondition | airbyte_agent_sdk.connectors.jira.types.IssuesLteCondition | airbyte_agent_sdk.connectors.jira.types.IssuesInCondition | airbyte_agent_sdk.connectors.jira.types.IssuesStartswithCondition | airbyte_agent_sdk.connectors.jira.types.IssuesEndswithCondition | airbyte_agent_sdk.connectors.jira.types.IssuesFuzzyCondition | airbyte_agent_sdk.connectors.jira.types.IssuesKeywordCondition | airbyte_agent_sdk.connectors.jira.types.IssuesContainsCondition | airbyte_agent_sdk.connectors.jira.types.IssuesArrayContainsCondition | airbyte_agent_sdk.connectors.jira.types.IssuesNotCondition | airbyte_agent_sdk.connectors.jira.types.IssuesAndCondition | airbyte_agent_sdk.connectors.jira.types.IssuesOrCondition | airbyte_agent_sdk.connectors.jira.types.IssuesAnyCondition
: The type of the None singleton.
sort: list[airbyte_agent_sdk.connectors.jira.types.IssuesSortFilter]
: The type of the None singleton.
IssuesSortFilter(*args, **kwargs)
: Available fields for sorting issues search results.
Ancestors (in MRO)
- builtins.dict
Class variables
changelog: Literal['asc', 'desc']
: Details of changelogs associated with the issue
created: Literal['asc', 'desc']
: The timestamp when the issue was created
editmeta: Literal['asc', 'desc']
: The metadata for the fields on the issue that can be amended
expand: Literal['asc', 'desc']
: Expand options that include additional issue details in the response
fields: Literal['asc', 'desc']
: Details of various fields associated with the issue
fields_to_include: Literal['asc', 'desc']
: Specify the fields to include in the fetched issues data
id: Literal['asc', 'desc']
: The unique ID of the issue
key: Literal['asc', 'desc']
: The unique key of the issue
names: Literal['asc', 'desc']
: The ID and name of each field present on the issue
operations: Literal['asc', 'desc']
: The operations that can be performed on the issue
project_id: Literal['asc', 'desc']
: The ID of the project containing the issue
project_key: Literal['asc', 'desc']
: The key of the project containing the issue
properties: Literal['asc', 'desc']
: Details of the issue properties identified in the request
rendered_fields: Literal['asc', 'desc']
: The rendered value of each field present on the issue
schema_: Literal['asc', 'desc']
: The schema describing each field present on the issue
self: Literal['asc', 'desc']
: The URL of the issue details
transitions: Literal['asc', 'desc']
: The transitions that can be performed on the issue
updated: Literal['asc', 'desc']
: The timestamp when the issue was last updated
versioned_representations: Literal['asc', 'desc']
: The versions of each field on the issue
IssuesStartswithCondition(*args, **kwargs)
: Literal case-insensitive prefix match.
Ancestors (in MRO)
- builtins.dict
Class variables
startswith: airbyte_agent_sdk.connectors.jira.types.IssuesStringFilter
: The type of the None singleton.
IssuesStringFilter(*args, **kwargs)
: String fields for text search conditions (startswith, endswith, fuzzy, keyword).
Ancestors (in MRO)
- builtins.dict
Class variables
changelog: str
: Details of changelogs associated with the issue
created: str
: The timestamp when the issue was created
editmeta: str
: The metadata for the fields on the issue that can be amended
expand: str
: Expand options that include additional issue details in the response
fields: str
: Details of various fields associated with the issue
fields_to_include: str
: Specify the fields to include in the fetched issues data
id: str
: The unique ID of the issue
key: str
: The unique key of the issue
names: str
: The ID and name of each field present on the issue
operations: str
: The operations that can be performed on the issue
project_id: str
: The ID of the project containing the issue
project_key: str
: The key of the project containing the issue
properties: str
: Details of the issue properties identified in the request
rendered_fields: str
: The rendered value of each field present on the issue
schema_: str
: The schema describing each field present on the issue
self: str
: The URL of the issue details
transitions: str
: The transitions that can be performed on the issue
updated: str
: The timestamp when the issue was last updated
versioned_representations: str
: The versions of each field on the issue
IssuesUpdateParams(*args, **kwargs)
: Parameters for issues.update operation
Ancestors (in MRO)
- builtins.dict
Class variables
expand: str
: The type of the None singleton.
fields: airbyte_agent_sdk.connectors.jira.types.IssuesUpdateParamsFields
: The type of the None singleton.
issue_id_or_key: str
: The type of the None singleton.
notify_users: bool
: The type of the None singleton.
override_editable_flag: bool
: The type of the None singleton.
override_screen_security: bool
: The type of the None singleton.
return_issue: bool
: The type of the None singleton.
transition: airbyte_agent_sdk.connectors.jira.types.IssuesUpdateParamsTransition
: The type of the None singleton.
Methods
update(...) ‑> dict[str, typing.Any]
: D.update([E, ]**F) -> None. Update D from mapping/iterable E and F.
If E is present and has a .keys() method, then does: for k in E.keys(): D[k] = E[k]
If E is present and lacks a .keys() method, then does: for k, v in E: D[k] = v
In either case, this is followed by: for k in F: D[k] = F[k]
IssuesUpdateParamsFields(*args, **kwargs)
: The issue fields to update
Ancestors (in MRO)
- builtins.dict
Class variables
assignee: airbyte_agent_sdk.connectors.jira.types.IssuesUpdateParamsFieldsAssignee
: The type of the None singleton.
description: airbyte_agent_sdk.connectors.jira.types.IssuesUpdateParamsFieldsDescription
: The type of the None singleton.
labels: list[str]
: The type of the None singleton.
priority: airbyte_agent_sdk.connectors.jira.types.IssuesUpdateParamsFieldsPriority
: The type of the None singleton.
summary: str
: The type of the None singleton.
IssuesUpdateParamsFieldsAssignee(*args, **kwargs)
: The user to assign the issue to
Ancestors (in MRO)
- builtins.dict
Class variables
account_id: str
: The type of the None singleton.
IssuesUpdateParamsFieldsDescription(*args, **kwargs)
: Issue description in Atlassian Document Format (ADF)
Ancestors (in MRO)
- builtins.dict
Class variables
content: list[airbyte_agent_sdk.connectors.jira.types.IssuesUpdateParamsFieldsDescriptionContentItem]
: The type of the None singleton.
type_: str
: The type of the None singleton.
version: int
: The type of the None singleton.
IssuesUpdateParamsFieldsDescriptionContentItem(*args, **kwargs)
: Nested schema for IssuesUpdateParamsFieldsDescription.content_item
Ancestors (in MRO)
- builtins.dict
Class variables
content: list[airbyte_agent_sdk.connectors.jira.types.IssuesUpdateParamsFieldsDescriptionContentItemContentItem]
: The type of the None singleton.
type_: str
: The type of the None singleton.
IssuesUpdateParamsFieldsDescriptionContentItemContentItem(*args, **kwargs)
: Nested schema for IssuesUpdateParamsFieldsDescriptionContentItem.content_item
Ancestors (in MRO)
- builtins.dict
Class variables
text: str
: The type of the None singleton.
type_: str
: The type of the None singleton.
IssuesUpdateParamsFieldsPriority(*args, **kwargs)
: Issue priority
Ancestors (in MRO)
- builtins.dict
Class variables
id: str
: The type of the None singleton.
name: str
: The type of the None singleton.
IssuesUpdateParamsTransition(*args, **kwargs)
: Transition the issue to a new status
Ancestors (in MRO)
- builtins.dict
Class variables
id: str
: The type of the None singleton.
ProjectsAndCondition(*args, **kwargs)
: dict() -> new empty dictionary
dict(mapping) -> new dictionary initialized from a mapping object's
(key, value) pairs
dict(iterable) -> new dictionary initialized as if via:
d = {}
for k, v in iterable:
d[k] = v
dict(**kwargs) -> new dictionary initialized with the name=value pairs
in the keyword argument list. For example: dict(one=1, two=2)
Ancestors (in MRO)
- builtins.dict
Class variables
and: list[airbyte_agent_sdk.connectors.jira.types.ProjectsEqCondition | airbyte_agent_sdk.connectors.jira.types.ProjectsNeqCondition | airbyte_agent_sdk.connectors.jira.types.ProjectsGtCondition | airbyte_agent_sdk.connectors.jira.types.ProjectsGteCondition | airbyte_agent_sdk.connectors.jira.types.ProjectsLtCondition | airbyte_agent_sdk.connectors.jira.types.ProjectsLteCondition | airbyte_agent_sdk.connectors.jira.types.ProjectsInCondition | airbyte_agent_sdk.connectors.jira.types.ProjectsStartswithCondition | airbyte_agent_sdk.connectors.jira.types.ProjectsEndswithCondition | airbyte_agent_sdk.connectors.jira.types.ProjectsFuzzyCondition | airbyte_agent_sdk.connectors.jira.types.ProjectsKeywordCondition | airbyte_agent_sdk.connectors.jira.types.ProjectsContainsCondition | airbyte_agent_sdk.connectors.jira.types.ProjectsArrayContainsCondition | airbyte_agent_sdk.connectors.jira.types.ProjectsNotCondition | airbyte_agent_sdk.connectors.jira.types.ProjectsAndCondition | airbyte_agent_sdk.connectors.jira.types.ProjectsOrCondition | airbyte_agent_sdk.connectors.jira.types.ProjectsAnyCondition]
: The type of the None singleton.
ProjectsAnyCondition(*args, **kwargs)
: dict() -> new empty dictionary
dict(mapping) -> new dictionary initialized from a mapping object's
(key, value) pairs
dict(iterable) -> new dictionary initialized as if via:
d = {}
for k, v in iterable:
d[k] = v
dict(**kwargs) -> new dictionary initialized with the name=value pairs
in the keyword argument list. For example: dict(one=1, two=2)
Ancestors (in MRO)
- builtins.dict
Class variables
any: airbyte_agent_sdk.connectors.jira.types.ProjectsAnyValueFilter
: The type of the None singleton.
ProjectsAnyValueFilter(*args, **kwargs)
: Available fields with Any value type. Used for 'contains' and 'any' conditions.
Ancestors (in MRO)
- builtins.dict
Class variables
archived: Any
: Whether the project is archived
archived_by: Any
: The user who archived the project
archived_date: Any
: The date when the project was archived
assignee_type: Any
: The default assignee when creating issues for this project
avatar_urls: Any
: The URLs of the project's avatars
components: Any
: List of the components contained in the project
deleted: Any
: Whether the project is marked as deleted
deleted_by: Any
: The user who marked the project as deleted
deleted_date: Any
: The date when the project was marked as deleted
description: Any
: A brief description of the project
email: Any
: An email address associated with the project
entity_id: Any
: The unique identifier of the project entity
expand: Any
: Expand options that include additional project details in the response
favourite: Any
: Whether the project is selected as a favorite
id: Any
: The ID of the project
insight: Any
: Insights about the project
is_private: Any
: Whether the project is private
issue_type_hierarchy: Any
: The issue type hierarchy for the project
issue_types: Any
: List of the issue types available in the project
key: Any
: The key of the project
lead: Any
: The username of the project lead
name: Any
: The name of the project
permissions: Any
: User permissions on the project
project_category: Any
: The category the project belongs to
project_type_key: Any
: The project type of the project
properties: Any
: Map of project properties
retention_till_date: Any
: The date when the project is deleted permanently
roles: Any
: The name and self URL for each role defined in the project
self: Any
: The URL of the project details
simplified: Any
: Whether the project is simplified
style: Any
: The type of the project
url: Any
: A link to information about this project
uuid: Any
: Unique ID for next-gen projects
versions: Any
: The versions defined in the project
ProjectsArrayContainsCondition(*args, **kwargs)
: Exact membership test on an array field. Example: {"array_contains": {"tags": "premium"}}
Ancestors (in MRO)
- builtins.dict
Class variables
array_contains: airbyte_agent_sdk.connectors.jira.types.ProjectsAnyValueFilter
: The type of the None singleton.
ProjectsContainsCondition(*args, **kwargs)
: Case-insensitive substring match on a scalar field. Example: {"contains": {"subject": "billing"}}
Ancestors (in MRO)
- builtins.dict
Class variables
contains: airbyte_agent_sdk.connectors.jira.types.ProjectsAnyValueFilter
: The type of the None singleton.
ProjectsEndswithCondition(*args, **kwargs)
: Literal case-insensitive suffix match.
Ancestors (in MRO)
- builtins.dict
Class variables
endswith: airbyte_agent_sdk.connectors.jira.types.ProjectsStringFilter
: The type of the None singleton.
ProjectsEqCondition(*args, **kwargs)
: Equal to: field equals value.
Ancestors (in MRO)
- builtins.dict
Class variables
eq: airbyte_agent_sdk.connectors.jira.types.ProjectsSearchFilter
: The type of the None singleton.
ProjectsFuzzyCondition(*args, **kwargs)
: Ordered word text match (case-insensitive).
Ancestors (in MRO)
- builtins.dict
Class variables
fuzzy: airbyte_agent_sdk.connectors.jira.types.ProjectsStringFilter
: The type of the None singleton.
ProjectsGetParams(*args, **kwargs)
: Parameters for projects.get operation
Ancestors (in MRO)
- builtins.dict
Class variables
expand: str
: The type of the None singleton.
project_id_or_key: str
: The type of the None singleton.
properties: str
: The type of the None singleton.
ProjectsGtCondition(*args, **kwargs)
: Greater than: field > value.
Ancestors (in MRO)
- builtins.dict
Class variables
gt: airbyte_agent_sdk.connectors.jira.types.ProjectsSearchFilter
: The type of the None singleton.
ProjectsGteCondition(*args, **kwargs)
: Greater than or equal: field >= value.
Ancestors (in MRO)
- builtins.dict
Class variables
gte: airbyte_agent_sdk.connectors.jira.types.ProjectsSearchFilter
: The type of the None singleton.
ProjectsInCondition(*args, **kwargs)
: dict() -> new empty dictionary
dict(mapping) -> new dictionary initialized from a mapping object's
(key, value) pairs
dict(iterable) -> new dictionary initialized as if via:
d = {}
for k, v in iterable:
d[k] = v
dict(**kwargs) -> new dictionary initialized with the name=value pairs
in the keyword argument list. For example: dict(one=1, two=2)
Ancestors (in MRO)
- builtins.dict
Class variables
in: airbyte_agent_sdk.connectors.jira.types.ProjectsInFilter
: The type of the None singleton.
ProjectsInFilter(*args, **kwargs)
: Available fields for 'in' condition (values are lists).
Ancestors (in MRO)
- builtins.dict
Class variables
archived: list[bool]
: Whether the project is archived
archived_by: list[dict[str, typing.Any]]
: The user who archived the project
archived_date: list[str]
: The date when the project was archived
assignee_type: list[str]
: The default assignee when creating issues for this project
avatar_urls: list[dict[str, typing.Any]]
: The URLs of the project's avatars
components: list[list[typing.Any]]
: List of the components contained in the project
deleted: list[bool]
: Whether the project is marked as deleted
deleted_by: list[dict[str, typing.Any]]
: The user who marked the project as deleted
deleted_date: list[str]
: The date when the project was marked as deleted
description: list[str]
: A brief description of the project
email: list[str]
: An email address associated with the project
entity_id: list[str]
: The unique identifier of the project entity
expand: list[str]
: Expand options that include additional project details in the response
favourite: list[bool]
: Whether the project is selected as a favorite
id: list[str]
: The ID of the project
insight: list[dict[str, typing.Any]]
: Insights about the project
is_private: list[bool]
: Whether the project is private
issue_type_hierarchy: list[dict[str, typing.Any]]
: The issue type hierarchy for the project
issue_types: list[list[typing.Any]]
: List of the issue types available in the project
key: list[str]
: The key of the project
lead: list[dict[str, typing.Any]]
: The username of the project lead
name: list[str]
: The name of the project
permissions: list[dict[str, typing.Any]]
: User permissions on the project
project_category: list[dict[str, typing.Any]]
: The category the project belongs to
project_type_key: list[str]
: The project type of the project
properties: list[dict[str, typing.Any]]
: Map of project properties
retention_till_date: list[str]
: The date when the project is deleted permanently
roles: list[dict[str, typing.Any]]
: The name and self URL for each role defined in the project
self: list[str]
: The URL of the project details
simplified: list[bool]
: Whether the project is simplified
style: list[str]
: The type of the project
url: list[str]
: A link to information about this project
uuid: list[str]
: Unique ID for next-gen projects
versions: list[list[typing.Any]]
: The versions defined in the project
ProjectsKeywordCondition(*args, **kwargs)
: Keyword text match (any word present).
Ancestors (in MRO)
- builtins.dict
Class variables
keyword: airbyte_agent_sdk.connectors.jira.types.ProjectsStringFilter
: The type of the None singleton.
ProjectsLtCondition(*args, **kwargs)
: Less than: field < value.
Ancestors (in MRO)
- builtins.dict
Class variables
lt: airbyte_agent_sdk.connectors.jira.types.ProjectsSearchFilter
: The type of the None singleton.
ProjectsLteCondition(*args, **kwargs)
: Less than or equal: field <= value.
Ancestors (in MRO)
- builtins.dict
Class variables
lte: airbyte_agent_sdk.connectors.jira.types.ProjectsSearchFilter
: The type of the None singleton.
ProjectsNeqCondition(*args, **kwargs)
: Not equal to: field does not equal value.
Ancestors (in MRO)
- builtins.dict
Class variables
neq: airbyte_agent_sdk.connectors.jira.types.ProjectsSearchFilter
: The type of the None singleton.
ProjectsNotCondition(*args, **kwargs)
: dict() -> new empty dictionary
dict(mapping) -> new dictionary initialized from a mapping object's
(key, value) pairs
dict(iterable) -> new dictionary initialized as if via:
d = {}
for k, v in iterable:
d[k] = v
dict(**kwargs) -> new dictionary initialized with the name=value pairs
in the keyword argument list. For example: dict(one=1, two=2)
Ancestors (in MRO)
- builtins.dict
Class variables
not: airbyte_agent_sdk.connectors.jira.types.ProjectsEqCondition | airbyte_agent_sdk.connectors.jira.types.ProjectsNeqCondition | airbyte_agent_sdk.connectors.jira.types.ProjectsGtCondition | airbyte_agent_sdk.connectors.jira.types.ProjectsGteCondition | airbyte_agent_sdk.connectors.jira.types.ProjectsLtCondition | airbyte_agent_sdk.connectors.jira.types.ProjectsLteCondition | airbyte_agent_sdk.connectors.jira.types.ProjectsInCondition | airbyte_agent_sdk.connectors.jira.types.ProjectsStartswithCondition | airbyte_agent_sdk.connectors.jira.types.ProjectsEndswithCondition | airbyte_agent_sdk.connectors.jira.types.ProjectsFuzzyCondition | airbyte_agent_sdk.connectors.jira.types.ProjectsKeywordCondition | airbyte_agent_sdk.connectors.jira.types.ProjectsContainsCondition | airbyte_agent_sdk.connectors.jira.types.ProjectsArrayContainsCondition | airbyte_agent_sdk.connectors.jira.types.ProjectsNotCondition | airbyte_agent_sdk.connectors.jira.types.ProjectsAndCondition | airbyte_agent_sdk.connectors.jira.types.ProjectsOrCondition | airbyte_agent_sdk.connectors.jira.types.ProjectsAnyCondition
: The type of the None singleton.
ProjectsOrCondition(*args, **kwargs)
: dict() -> new empty dictionary
dict(mapping) -> new dictionary initialized from a mapping object's
(key, value) pairs
dict(iterable) -> new dictionary initialized as if via:
d = {}
for k, v in iterable:
d[k] = v
dict(**kwargs) -> new dictionary initialized with the name=value pairs
in the keyword argument list. For example: dict(one=1, two=2)
Ancestors (in MRO)
- builtins.dict
Class variables
or: list[airbyte_agent_sdk.connectors.jira.types.ProjectsEqCondition | airbyte_agent_sdk.connectors.jira.types.ProjectsNeqCondition | airbyte_agent_sdk.connectors.jira.types.ProjectsGtCondition | airbyte_agent_sdk.connectors.jira.types.ProjectsGteCondition | airbyte_agent_sdk.connectors.jira.types.ProjectsLtCondition | airbyte_agent_sdk.connectors.jira.types.ProjectsLteCondition | airbyte_agent_sdk.connectors.jira.types.ProjectsInCondition | airbyte_agent_sdk.connectors.jira.types.ProjectsStartswithCondition | airbyte_agent_sdk.connectors.jira.types.ProjectsEndswithCondition | airbyte_agent_sdk.connectors.jira.types.ProjectsFuzzyCondition | airbyte_agent_sdk.connectors.jira.types.ProjectsKeywordCondition | airbyte_agent_sdk.connectors.jira.types.ProjectsContainsCondition | airbyte_agent_sdk.connectors.jira.types.ProjectsArrayContainsCondition | airbyte_agent_sdk.connectors.jira.types.ProjectsNotCondition | airbyte_agent_sdk.connectors.jira.types.ProjectsAndCondition | airbyte_agent_sdk.connectors.jira.types.ProjectsOrCondition | airbyte_agent_sdk.connectors.jira.types.ProjectsAnyCondition]
: The type of the None singleton.
ProjectsSearchFilter(*args, **kwargs)
: Available fields for filtering projects search queries.
Ancestors (in MRO)
- builtins.dict
Class variables
archived: bool
: Whether the project is archived
archived_by: dict[str, typing.Any] | None
: The user who archived the project
archived_date: str | None
: The date when the project was archived
assignee_type: str | None
: The default assignee when creating issues for this project
avatar_urls: dict[str, typing.Any]
: The URLs of the project's avatars
components: list[typing.Any]
: List of the components contained in the project
deleted: bool
: Whether the project is marked as deleted
deleted_by: dict[str, typing.Any] | None
: The user who marked the project as deleted
deleted_date: str | None
: The date when the project was marked as deleted
description: str | None
: A brief description of the project
email: str | None
: An email address associated with the project
entity_id: str | None
: The unique identifier of the project entity
expand: str | None
: Expand options that include additional project details in the response
favourite: bool
: Whether the project is selected as a favorite
id: str
: The ID of the project
insight: dict[str, typing.Any] | None
: Insights about the project
is_private: bool
: Whether the project is private
issue_type_hierarchy: dict[str, typing.Any] | None
: The issue type hierarchy for the project
issue_types: list[typing.Any]
: List of the issue types available in the project
key: str
: The key of the project
lead: dict[str, typing.Any] | None
: The username of the project lead
name: str
: The name of the project
permissions: dict[str, typing.Any] | None
: User permissions on the project
project_category: dict[str, typing.Any] | None
: The category the project belongs to
project_type_key: str | None
: The project type of the project
properties: dict[str, typing.Any]
: Map of project properties
retention_till_date: str | None
: The date when the project is deleted permanently
roles: dict[str, typing.Any]
: The name and self URL for each role defined in the project
self: str
: The URL of the project details
simplified: bool
: Whether the project is simplified
style: str | None
: The type of the project
url: str | None
: A link to information about this project
uuid: str | None
: Unique ID for next-gen projects
versions: list[typing.Any]
: The versions defined in the project
ProjectsSearchParams(*args, **kwargs)
: Parameters for projects.search operation
Ancestors (in MRO)
- builtins.dict
Class variables
action: str
: The type of the None singleton.
category_id: int
: The type of the None singleton.
expand: str
: The type of the None singleton.
id: list[int]
: The type of the None singleton.
max_results: int
: The type of the None singleton.
order_by: str
: The type of the None singleton.
query: str
: The type of the None singleton.
start_at: int
: The type of the None singleton.
status: list[str]
: The type of the None singleton.
type_key: str
: The type of the None singleton.
Methods
keys(self, /) ‑> list[str]
: Return a set-like object providing a view on the dict's keys.
ProjectsSearchQuery(*args, **kwargs)
: Search query for projects entity.
Ancestors (in MRO)
- builtins.dict
Class variables
filter: airbyte_agent_sdk.connectors.jira.types.ProjectsEqCondition | airbyte_agent_sdk.connectors.jira.types.ProjectsNeqCondition | airbyte_agent_sdk.connectors.jira.types.ProjectsGtCondition | airbyte_agent_sdk.connectors.jira.types.ProjectsGteCondition | airbyte_agent_sdk.connectors.jira.types.ProjectsLtCondition | airbyte_agent_sdk.connectors.jira.types.ProjectsLteCondition | airbyte_agent_sdk.connectors.jira.types.ProjectsInCondition | airbyte_agent_sdk.connectors.jira.types.ProjectsStartswithCondition | airbyte_agent_sdk.connectors.jira.types.ProjectsEndswithCondition | airbyte_agent_sdk.connectors.jira.types.ProjectsFuzzyCondition | airbyte_agent_sdk.connectors.jira.types.ProjectsKeywordCondition | airbyte_agent_sdk.connectors.jira.types.ProjectsContainsCondition | airbyte_agent_sdk.connectors.jira.types.ProjectsArrayContainsCondition | airbyte_agent_sdk.connectors.jira.types.ProjectsNotCondition | airbyte_agent_sdk.connectors.jira.types.ProjectsAndCondition | airbyte_agent_sdk.connectors.jira.types.ProjectsOrCondition | airbyte_agent_sdk.connectors.jira.types.ProjectsAnyCondition
: The type of the None singleton.
sort: list[airbyte_agent_sdk.connectors.jira.types.ProjectsSortFilter]
: The type of the None singleton.
ProjectsSortFilter(*args, **kwargs)
: Available fields for sorting projects search results.
Ancestors (in MRO)
- builtins.dict
Class variables
archived: Literal['asc', 'desc']
: Whether the project is archived
archived_by: Literal['asc', 'desc']
: The user who archived the project
archived_date: Literal['asc', 'desc']
: The date when the project was archived
assignee_type: Literal['asc', 'desc']
: The default assignee when creating issues for this project
avatar_urls: Literal['asc', 'desc']
: The URLs of the project's avatars
components: Literal['asc', 'desc']
: List of the components contained in the project
deleted: Literal['asc', 'desc']
: Whether the project is marked as deleted
deleted_by: Literal['asc', 'desc']
: The user who marked the project as deleted
deleted_date: Literal['asc', 'desc']
: The date when the project was marked as deleted
description: Literal['asc', 'desc']
: A brief description of the project
email: Literal['asc', 'desc']
: An email address associated with the project
entity_id: Literal['asc', 'desc']
: The unique identifier of the project entity
expand: Literal['asc', 'desc']
: Expand options that include additional project details in the response
favourite: Literal['asc', 'desc']
: Whether the project is selected as a favorite
id: Literal['asc', 'desc']
: The ID of the project
insight: Literal['asc', 'desc']
: Insights about the project
is_private: Literal['asc', 'desc']
: Whether the project is private
issue_type_hierarchy: Literal['asc', 'desc']
: The issue type hierarchy for the project
issue_types: Literal['asc', 'desc']
: List of the issue types available in the project
key: Literal['asc', 'desc']
: The key of the project
lead: Literal['asc', 'desc']
: The username of the project lead
name: Literal['asc', 'desc']
: The name of the project
permissions: Literal['asc', 'desc']
: User permissions on the project
project_category: Literal['asc', 'desc']
: The category the project belongs to
project_type_key: Literal['asc', 'desc']
: The project type of the project
properties: Literal['asc', 'desc']
: Map of project properties
retention_till_date: Literal['asc', 'desc']
: The date when the project is deleted permanently
roles: Literal['asc', 'desc']
: The name and self URL for each role defined in the project
self: Literal['asc', 'desc']
: The URL of the project details
simplified: Literal['asc', 'desc']
: Whether the project is simplified
style: Literal['asc', 'desc']
: The type of the project
url: Literal['asc', 'desc']
: A link to information about this project
uuid: Literal['asc', 'desc']
: Unique ID for next-gen projects
versions: Literal['asc', 'desc']
: The versions defined in the project
ProjectsStartswithCondition(*args, **kwargs)
: Literal case-insensitive prefix match.
Ancestors (in MRO)
- builtins.dict
Class variables
startswith: airbyte_agent_sdk.connectors.jira.types.ProjectsStringFilter
: The type of the None singleton.
ProjectsStringFilter(*args, **kwargs)
: String fields for text search conditions (startswith, endswith, fuzzy, keyword).
Ancestors (in MRO)
- builtins.dict
Class variables
archived: str
: Whether the project is archived
archived_by: str
: The user who archived the project
archived_date: str
: The date when the project was archived
assignee_type: str
: The default assignee when creating issues for this project
avatar_urls: str
: The URLs of the project's avatars
components: str
: List of the components contained in the project
deleted: str
: Whether the project is marked as deleted
deleted_by: str
: The user who marked the project as deleted
deleted_date: str
: The date when the project was marked as deleted
description: str
: A brief description of the project
email: str
: An email address associated with the project
entity_id: str
: The unique identifier of the project entity
expand: str
: Expand options that include additional project details in the response
favourite: str
: Whether the project is selected as a favorite
id: str
: The ID of the project
insight: str
: Insights about the project
is_private: str
: Whether the project is private
issue_type_hierarchy: str
: The issue type hierarchy for the project
issue_types: str
: List of the issue types available in the project
key: str
: The key of the project
lead: str
: The username of the project lead
name: str
: The name of the project
permissions: str
: User permissions on the project
project_category: str
: The category the project belongs to
project_type_key: str
: The project type of the project
properties: str
: Map of project properties
retention_till_date: str
: The date when the project is deleted permanently
roles: str
: The name and self URL for each role defined in the project
self: str
: The URL of the project details
simplified: str
: Whether the project is simplified
style: str
: The type of the project
url: str
: A link to information about this project
uuid: str
: Unique ID for next-gen projects
versions: str
: The versions defined in the project
UsersAndCondition(*args, **kwargs)
: dict() -> new empty dictionary
dict(mapping) -> new dictionary initialized from a mapping object's
(key, value) pairs
dict(iterable) -> new dictionary initialized as if via:
d = {}
for k, v in iterable:
d[k] = v
dict(**kwargs) -> new dictionary initialized with the name=value pairs
in the keyword argument list. For example: dict(one=1, two=2)
Ancestors (in MRO)
- builtins.dict
Class variables
and: list[airbyte_agent_sdk.connectors.jira.types.UsersEqCondition | airbyte_agent_sdk.connectors.jira.types.UsersNeqCondition | airbyte_agent_sdk.connectors.jira.types.UsersGtCondition | airbyte_agent_sdk.connectors.jira.types.UsersGteCondition | airbyte_agent_sdk.connectors.jira.types.UsersLtCondition | airbyte_agent_sdk.connectors.jira.types.UsersLteCondition | airbyte_agent_sdk.connectors.jira.types.UsersInCondition | airbyte_agent_sdk.connectors.jira.types.UsersStartswithCondition | airbyte_agent_sdk.connectors.jira.types.UsersEndswithCondition | airbyte_agent_sdk.connectors.jira.types.UsersFuzzyCondition | airbyte_agent_sdk.connectors.jira.types.UsersKeywordCondition | airbyte_agent_sdk.connectors.jira.types.UsersContainsCondition | airbyte_agent_sdk.connectors.jira.types.UsersArrayContainsCondition | airbyte_agent_sdk.connectors.jira.types.UsersNotCondition | airbyte_agent_sdk.connectors.jira.types.UsersAndCondition | airbyte_agent_sdk.connectors.jira.types.UsersOrCondition | airbyte_agent_sdk.connectors.jira.types.UsersAnyCondition]
: The type of the None singleton.
UsersAnyCondition(*args, **kwargs)
: dict() -> new empty dictionary
dict(mapping) -> new dictionary initialized from a mapping object's
(key, value) pairs
dict(iterable) -> new dictionary initialized as if via:
d = {}
for k, v in iterable:
d[k] = v
dict(**kwargs) -> new dictionary initialized with the name=value pairs
in the keyword argument list. For example: dict(one=1, two=2)
Ancestors (in MRO)
- builtins.dict
Class variables
any: airbyte_agent_sdk.connectors.jira.types.UsersAnyValueFilter
: The type of the None singleton.
UsersAnyValueFilter(*args, **kwargs)
: Available fields with Any value type. Used for 'contains' and 'any' conditions.
Ancestors (in MRO)
- builtins.dict
Class variables
account_id: Any
: The account ID of the user, uniquely identifying the user across all Atlassian products
account_type: Any
: The user account type (atlassian, app, or customer)
active: Any
: Indicates whether the user is active
application_roles: Any
: The application roles assigned to the user
avatar_urls: Any
: The avatars of the user
display_name: Any
: The display name of the user
email_address: Any
: The email address of the user
expand: Any
: Options to include additional user details in the response
groups: Any
: The groups to which the user belongs
key: Any
: Deprecated property
locale: Any
: The locale of the user
name: Any
: Deprecated property
self: Any
: The URL of the user
time_zone: Any
: The time zone specified in the user's profile
UsersArrayContainsCondition(*args, **kwargs)
: Exact membership test on an array field. Example: {"array_contains": {"tags": "premium"}}
Ancestors (in MRO)
- builtins.dict
Class variables
array_contains: airbyte_agent_sdk.connectors.jira.types.UsersAnyValueFilter
: The type of the None singleton.
UsersContainsCondition(*args, **kwargs)
: Case-insensitive substring match on a scalar field. Example: {"contains": {"subject": "billing"}}
Ancestors (in MRO)
- builtins.dict
Class variables
contains: airbyte_agent_sdk.connectors.jira.types.UsersAnyValueFilter
: The type of the None singleton.
UsersEndswithCondition(*args, **kwargs)
: Literal case-insensitive suffix match.
Ancestors (in MRO)
- builtins.dict
Class variables
endswith: airbyte_agent_sdk.connectors.jira.types.UsersStringFilter
: The type of the None singleton.
UsersEqCondition(*args, **kwargs)
: Equal to: field equals value.
Ancestors (in MRO)
- builtins.dict
Class variables
eq: airbyte_agent_sdk.connectors.jira.types.UsersSearchFilter
: The type of the None singleton.
UsersFuzzyCondition(*args, **kwargs)
: Ordered word text match (case-insensitive).
Ancestors (in MRO)
- builtins.dict
Class variables
fuzzy: airbyte_agent_sdk.connectors.jira.types.UsersStringFilter
: The type of the None singleton.
UsersGetParams(*args, **kwargs)
: Parameters for users.get operation
Ancestors (in MRO)
- builtins.dict
Class variables
account_id: str
: The type of the None singleton.
expand: str
: The type of the None singleton.
UsersGtCondition(*args, **kwargs)
: Greater than: field > value.
Ancestors (in MRO)
- builtins.dict
Class variables
gt: airbyte_agent_sdk.connectors.jira.types.UsersSearchFilter
: The type of the None singleton.
UsersGteCondition(*args, **kwargs)
: Greater than or equal: field >= value.
Ancestors (in MRO)
- builtins.dict
Class variables
gte: airbyte_agent_sdk.connectors.jira.types.UsersSearchFilter
: The type of the None singleton.
UsersInCondition(*args, **kwargs)
: dict() -> new empty dictionary
dict(mapping) -> new dictionary initialized from a mapping object's
(key, value) pairs
dict(iterable) -> new dictionary initialized as if via:
d = {}
for k, v in iterable:
d[k] = v
dict(**kwargs) -> new dictionary initialized with the name=value pairs
in the keyword argument list. For example: dict(one=1, two=2)
Ancestors (in MRO)
- builtins.dict
Class variables
in: airbyte_agent_sdk.connectors.jira.types.UsersInFilter
: The type of the None singleton.
UsersInFilter(*args, **kwargs)
: Available fields for 'in' condition (values are lists).
Ancestors (in MRO)
- builtins.dict
Class variables
account_id: list[str]
: The account ID of the user, uniquely identifying the user across all Atlassian products
account_type: list[str]
: The user account type (atlassian, app, or customer)
active: list[bool]
: Indicates whether the user is active
application_roles: list[dict[str, typing.Any]]
: The application roles assigned to the user
avatar_urls: list[dict[str, typing.Any]]
: The avatars of the user
display_name: list[str]
: The display name of the user
email_address: list[str]
: The email address of the user
expand: list[str]
: Options to include additional user details in the response
groups: list[dict[str, typing.Any]]
: The groups to which the user belongs
key: list[str]
: Deprecated property
locale: list[str]
: The locale of the user
name: list[str]
: Deprecated property
self: list[str]
: The URL of the user
time_zone: list[str]
: The time zone specified in the user's profile
UsersKeywordCondition(*args, **kwargs)
: Keyword text match (any word present).
Ancestors (in MRO)
- builtins.dict
Class variables
keyword: airbyte_agent_sdk.connectors.jira.types.UsersStringFilter
: The type of the None singleton.
UsersListParams(*args, **kwargs)
: Parameters for users.list operation
Ancestors (in MRO)
- builtins.dict
Class variables
max_results: int
: The type of the None singleton.
start_at: int
: The type of the None singleton.
UsersLtCondition(*args, **kwargs)
: Less than: field < value.
Ancestors (in MRO)
- builtins.dict
Class variables
lt: airbyte_agent_sdk.connectors.jira.types.UsersSearchFilter
: The type of the None singleton.
UsersLteCondition(*args, **kwargs)
: Less than or equal: field <= value.
Ancestors (in MRO)
- builtins.dict
Class variables
lte: airbyte_agent_sdk.connectors.jira.types.UsersSearchFilter
: The type of the None singleton.
UsersNeqCondition(*args, **kwargs)
: Not equal to: field does not equal value.
Ancestors (in MRO)
- builtins.dict
Class variables
neq: airbyte_agent_sdk.connectors.jira.types.UsersSearchFilter
: The type of the None singleton.
UsersNotCondition(*args, **kwargs)
: dict() -> new empty dictionary
dict(mapping) -> new dictionary initialized from a mapping object's
(key, value) pairs
dict(iterable) -> new dictionary initialized as if via:
d = {}
for k, v in iterable:
d[k] = v
dict(**kwargs) -> new dictionary initialized with the name=value pairs
in the keyword argument list. For example: dict(one=1, two=2)
Ancestors (in MRO)
- builtins.dict
Class variables
not: airbyte_agent_sdk.connectors.jira.types.UsersEqCondition | airbyte_agent_sdk.connectors.jira.types.UsersNeqCondition | airbyte_agent_sdk.connectors.jira.types.UsersGtCondition | airbyte_agent_sdk.connectors.jira.types.UsersGteCondition | airbyte_agent_sdk.connectors.jira.types.UsersLtCondition | airbyte_agent_sdk.connectors.jira.types.UsersLteCondition | airbyte_agent_sdk.connectors.jira.types.UsersInCondition | airbyte_agent_sdk.connectors.jira.types.UsersStartswithCondition | airbyte_agent_sdk.connectors.jira.types.UsersEndswithCondition | airbyte_agent_sdk.connectors.jira.types.UsersFuzzyCondition | airbyte_agent_sdk.connectors.jira.types.UsersKeywordCondition | airbyte_agent_sdk.connectors.jira.types.UsersContainsCondition | airbyte_agent_sdk.connectors.jira.types.UsersArrayContainsCondition | airbyte_agent_sdk.connectors.jira.types.UsersNotCondition | airbyte_agent_sdk.connectors.jira.types.UsersAndCondition | airbyte_agent_sdk.connectors.jira.types.UsersOrCondition | airbyte_agent_sdk.connectors.jira.types.UsersAnyCondition
: The type of the None singleton.
UsersOrCondition(*args, **kwargs)
: dict() -> new empty dictionary
dict(mapping) -> new dictionary initialized from a mapping object's
(key, value) pairs
dict(iterable) -> new dictionary initialized as if via:
d = {}
for k, v in iterable:
d[k] = v
dict(**kwargs) -> new dictionary initialized with the name=value pairs
in the keyword argument list. For example: dict(one=1, two=2)
Ancestors (in MRO)
- builtins.dict
Class variables
or: list[airbyte_agent_sdk.connectors.jira.types.UsersEqCondition | airbyte_agent_sdk.connectors.jira.types.UsersNeqCondition | airbyte_agent_sdk.connectors.jira.types.UsersGtCondition | airbyte_agent_sdk.connectors.jira.types.UsersGteCondition | airbyte_agent_sdk.connectors.jira.types.UsersLtCondition | airbyte_agent_sdk.connectors.jira.types.UsersLteCondition | airbyte_agent_sdk.connectors.jira.types.UsersInCondition | airbyte_agent_sdk.connectors.jira.types.UsersStartswithCondition | airbyte_agent_sdk.connectors.jira.types.UsersEndswithCondition | airbyte_agent_sdk.connectors.jira.types.UsersFuzzyCondition | airbyte_agent_sdk.connectors.jira.types.UsersKeywordCondition | airbyte_agent_sdk.connectors.jira.types.UsersContainsCondition | airbyte_agent_sdk.connectors.jira.types.UsersArrayContainsCondition | airbyte_agent_sdk.connectors.jira.types.UsersNotCondition | airbyte_agent_sdk.connectors.jira.types.UsersAndCondition | airbyte_agent_sdk.connectors.jira.types.UsersOrCondition | airbyte_agent_sdk.connectors.jira.types.UsersAnyCondition]
: The type of the None singleton.
UsersSearchFilter(*args, **kwargs)
: Available fields for filtering users search queries.
Ancestors (in MRO)
- builtins.dict
Class variables
account_id: str
: The account ID of the user, uniquely identifying the user across all Atlassian products
account_type: str | None
: The user account type (atlassian, app, or customer)
active: bool
: Indicates whether the user is active
application_roles: dict[str, typing.Any] | None
: The application roles assigned to the user
avatar_urls: dict[str, typing.Any]
: The avatars of the user
display_name: str | None
: The display name of the user
email_address: str | None
: The email address of the user
expand: str | None
: Options to include additional user details in the response
groups: dict[str, typing.Any] | None
: The groups to which the user belongs
key: str | None
: Deprecated property
locale: str | None
: The locale of the user
name: str | None
: Deprecated property
self: str
: The URL of the user
time_zone: str | None
: The time zone specified in the user's profile
UsersSearchParams(*args, **kwargs)
: Parameters for users.search operation
Ancestors (in MRO)
- builtins.dict
Class variables
account_id: str
: The type of the None singleton.
max_results: int
: The type of the None singleton.
property: str
: The type of the None singleton.
query: str
: The type of the None singleton.
start_at: int
: The type of the None singleton.
UsersSearchQuery(*args, **kwargs)
: Search query for users entity.
Ancestors (in MRO)
- builtins.dict
Class variables
filter: airbyte_agent_sdk.connectors.jira.types.UsersEqCondition | airbyte_agent_sdk.connectors.jira.types.UsersNeqCondition | airbyte_agent_sdk.connectors.jira.types.UsersGtCondition | airbyte_agent_sdk.connectors.jira.types.UsersGteCondition | airbyte_agent_sdk.connectors.jira.types.UsersLtCondition | airbyte_agent_sdk.connectors.jira.types.UsersLteCondition | airbyte_agent_sdk.connectors.jira.types.UsersInCondition | airbyte_agent_sdk.connectors.jira.types.UsersStartswithCondition | airbyte_agent_sdk.connectors.jira.types.UsersEndswithCondition | airbyte_agent_sdk.connectors.jira.types.UsersFuzzyCondition | airbyte_agent_sdk.connectors.jira.types.UsersKeywordCondition | airbyte_agent_sdk.connectors.jira.types.UsersContainsCondition | airbyte_agent_sdk.connectors.jira.types.UsersArrayContainsCondition | airbyte_agent_sdk.connectors.jira.types.UsersNotCondition | airbyte_agent_sdk.connectors.jira.types.UsersAndCondition | airbyte_agent_sdk.connectors.jira.types.UsersOrCondition | airbyte_agent_sdk.connectors.jira.types.UsersAnyCondition
: The type of the None singleton.
sort: list[airbyte_agent_sdk.connectors.jira.types.UsersSortFilter]
: The type of the None singleton.
UsersSortFilter(*args, **kwargs)
: Available fields for sorting users search results.
Ancestors (in MRO)
- builtins.dict
Class variables
account_id: Literal['asc', 'desc']
: The account ID of the user, uniquely identifying the user across all Atlassian products
account_type: Literal['asc', 'desc']
: The user account type (atlassian, app, or customer)
active: Literal['asc', 'desc']
: Indicates whether the user is active
application_roles: Literal['asc', 'desc']
: The application roles assigned to the user
avatar_urls: Literal['asc', 'desc']
: The avatars of the user
display_name: Literal['asc', 'desc']
: The display name of the user
email_address: Literal['asc', 'desc']
: The email address of the user
expand: Literal['asc', 'desc']
: Options to include additional user details in the response
groups: Literal['asc', 'desc']
: The groups to which the user belongs
key: Literal['asc', 'desc']
: Deprecated property
locale: Literal['asc', 'desc']
: The locale of the user
name: Literal['asc', 'desc']
: Deprecated property
self: Literal['asc', 'desc']
: The URL of the user
time_zone: Literal['asc', 'desc']
: The time zone specified in the user's profile
UsersStartswithCondition(*args, **kwargs)
: Literal case-insensitive prefix match.
Ancestors (in MRO)
- builtins.dict
Class variables
startswith: airbyte_agent_sdk.connectors.jira.types.UsersStringFilter
: The type of the None singleton.
UsersStringFilter(*args, **kwargs)
: String fields for text search conditions (startswith, endswith, fuzzy, keyword).
Ancestors (in MRO)
- builtins.dict
Class variables
account_id: str
: The account ID of the user, uniquely identifying the user across all Atlassian products
account_type: str
: The user account type (atlassian, app, or customer)
active: str
: Indicates whether the user is active
application_roles: str
: The application roles assigned to the user
avatar_urls: str
: The avatars of the user
display_name: str
: The display name of the user
email_address: str
: The email address of the user
expand: str
: Options to include additional user details in the response
groups: str
: The groups to which the user belongs
key: str
: Deprecated property
locale: str
: The locale of the user
name: str
: Deprecated property
self: str
: The URL of the user
time_zone: str
: The time zone specified in the user's profile