Skip to content

Commit 3c829a0

Browse files
gcf-owl-bot[bot]gcf-owl-bot[bot]
authored
feat: [google-cloud-documentai] Add image block and blob asset in Document proto (#13831)
- [ ] Regenerate this pull request now. PiperOrigin-RevId: 750637959 Source-Link: googleapis/googleapis@37b9fc5 Source-Link: googleapis/googleapis-gen@6481853 Copy-Tag: eyJwIjoicGFja2FnZXMvZ29vZ2xlLWNsb3VkLWRvY3VtZW50YWkvLk93bEJvdC55YW1sIiwiaCI6IjY0ODE4NTMzNmU0MTIxMzQyNGYwMGFjMjg3NjViMTAxMWVhZTA1NjIifQ== --------- Co-authored-by: Owl Bot <gcf-owl-bot[bot]@users.noreply.github.com>
1 parent cbf9782 commit 3c829a0

File tree

7 files changed

+274
-5
lines changed

7 files changed

+274
-5
lines changed

‎packages/google-cloud-documentai/google/cloud/documentai/gapic_version.py‎

Lines changed: 1 addition & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -13,4 +13,4 @@
1313
# See the License for the specific language governing permissions and
1414
# limitations under the License.
1515
#
16-
__version__ = "3.4.0" # {x-release-please-version}
16+
__version__ = "0.0.0" # {x-release-please-version}

‎packages/google-cloud-documentai/google/cloud/documentai_v1/gapic_version.py‎

Lines changed: 1 addition & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -13,4 +13,4 @@
1313
# See the License for the specific language governing permissions and
1414
# limitations under the License.
1515
#
16-
__version__ = "3.4.0" # {x-release-please-version}
16+
__version__ = "0.0.0" # {x-release-please-version}

‎packages/google-cloud-documentai/google/cloud/documentai_v1beta3/gapic_version.py‎

Lines changed: 1 addition & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -13,4 +13,4 @@
1313
# See the License for the specific language governing permissions and
1414
# limitations under the License.
1515
#
16-
__version__ = "3.4.0" # {x-release-please-version}
16+
__version__ = "0.0.0" # {x-release-please-version}

‎packages/google-cloud-documentai/google/cloud/documentai_v1beta3/types/document.py‎

Lines changed: 262 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -67,6 +67,9 @@ class Document(proto.Message):
6767
representations use base64.
6868
6969
This field is a member of `oneof`_ ``source``.
70+
docid (str):
71+
Optional. An internal identifier for
72+
document. Should be loggable (no PII).
7073
mime_type (str):
7174
An IANA published `media type (MIME
7275
type) <https://www.iana.org/assignments/media-types/media-types.xhtml>`__.
@@ -108,6 +111,12 @@ class Document(proto.Message):
108111
Parsed layout of the document.
109112
chunked_document (google.cloud.documentai_v1beta3.types.Document.ChunkedDocument):
110113
Document chunked based on chunking config.
114+
blob_assets (MutableSequence[google.cloud.documentai_v1beta3.types.Document.BlobAsset]):
115+
Optional. The blob assets in this document.
116+
This is used to store the content of the inline
117+
blobs in this document, e.g. image bytes, such
118+
that it can be referenced by other fields in the
119+
document via asset id.
111120
"""
112121

113122
class ShardInfo(proto.Message):
@@ -1819,6 +1828,20 @@ class TextChange(proto.Message):
18191828
message="Document.Provenance",
18201829
)
18211830

1831+
class Annotations(proto.Message):
1832+
r"""Represents the annotation of a block or a chunk.
1833+
1834+
Attributes:
1835+
description (str):
1836+
The description of the content with this
1837+
annotation.
1838+
"""
1839+
1840+
description: str = proto.Field(
1841+
proto.STRING,
1842+
number=1,
1843+
)
1844+
18221845
class DocumentLayout(proto.Message):
18231846
r"""Represents the parsed layout of a document as a collection of
18241847
blocks that the document is divided into.
@@ -1851,11 +1874,17 @@ class DocumentLayoutBlock(proto.Message):
18511874
list_block (google.cloud.documentai_v1beta3.types.Document.DocumentLayout.DocumentLayoutBlock.LayoutListBlock):
18521875
Block consisting of list content/structure.
18531876
1877+
This field is a member of `oneof`_ ``block``.
1878+
image_block (google.cloud.documentai_v1beta3.types.Document.DocumentLayout.DocumentLayoutBlock.LayoutImageBlock):
1879+
Block consisting of image content.
1880+
18541881
This field is a member of `oneof`_ ``block``.
18551882
block_id (str):
18561883
ID of the block.
18571884
page_span (google.cloud.documentai_v1beta3.types.Document.DocumentLayout.DocumentLayoutBlock.LayoutPageSpan):
18581885
Page span of the block.
1886+
bounding_box (google.cloud.documentai_v1beta3.types.BoundingPoly):
1887+
Identifies the bounding box for the block.
18591888
"""
18601889

18611890
class LayoutPageSpan(proto.Message):
@@ -2028,6 +2057,74 @@ class LayoutListEntry(proto.Message):
20282057
message="Document.DocumentLayout.DocumentLayoutBlock",
20292058
)
20302059

2060+
class LayoutImageBlock(proto.Message):
2061+
r"""Represents an image type block.
2062+
2063+
This message has `oneof`_ fields (mutually exclusive fields).
2064+
For each oneof, at most one member field can be set at the same time.
2065+
Setting any member of the oneof automatically clears all other
2066+
members.
2067+
2068+
.. _oneof: https://proto-plus-python.readthedocs.io/en/stable/fields.html#oneofs-mutually-exclusive-fields
2069+
2070+
Attributes:
2071+
blob_asset_id (str):
2072+
Optional. Asset id of the inline image. If set, find the
2073+
image content in the blob_assets field.
2074+
2075+
This field is a member of `oneof`_ ``image_source``.
2076+
gcs_uri (str):
2077+
Optional. Google Cloud Storage uri of the
2078+
image.
2079+
2080+
This field is a member of `oneof`_ ``image_source``.
2081+
data_uri (str):
2082+
Optional. Data uri of the image. It is composed of four
2083+
parts: a prefix (data:), a MIME type indicating the type of
2084+
data, an optional base64 token if non-textual, and the data
2085+
itself: data:[][;base64],
2086+
2087+
This field is a member of `oneof`_ ``image_source``.
2088+
mime_type (str):
2089+
Mime type of the image. An IANA published [media type (MIME
2090+
type)]
2091+
(https://www.iana.org/assignments/media-types/media-types.xhtml).
2092+
image_text (str):
2093+
Text extracted from the image using OCR or
2094+
alt text describing the image.
2095+
annotations (google.cloud.documentai_v1beta3.types.Document.Annotations):
2096+
Annotation of the image block.
2097+
"""
2098+
2099+
blob_asset_id: str = proto.Field(
2100+
proto.STRING,
2101+
number=4,
2102+
oneof="image_source",
2103+
)
2104+
gcs_uri: str = proto.Field(
2105+
proto.STRING,
2106+
number=5,
2107+
oneof="image_source",
2108+
)
2109+
data_uri: str = proto.Field(
2110+
proto.STRING,
2111+
number=6,
2112+
oneof="image_source",
2113+
)
2114+
mime_type: str = proto.Field(
2115+
proto.STRING,
2116+
number=1,
2117+
)
2118+
image_text: str = proto.Field(
2119+
proto.STRING,
2120+
number=2,
2121+
)
2122+
annotations: "Document.Annotations" = proto.Field(
2123+
proto.MESSAGE,
2124+
number=3,
2125+
message="Document.Annotations",
2126+
)
2127+
20312128
text_block: "Document.DocumentLayout.DocumentLayoutBlock.LayoutTextBlock" = proto.Field(
20322129
proto.MESSAGE,
20332130
number=2,
@@ -2046,6 +2143,12 @@ class LayoutListEntry(proto.Message):
20462143
oneof="block",
20472144
message="Document.DocumentLayout.DocumentLayoutBlock.LayoutListBlock",
20482145
)
2146+
image_block: "Document.DocumentLayout.DocumentLayoutBlock.LayoutImageBlock" = proto.Field(
2147+
proto.MESSAGE,
2148+
number=7,
2149+
oneof="block",
2150+
message="Document.DocumentLayout.DocumentLayoutBlock.LayoutImageBlock",
2151+
)
20492152
block_id: str = proto.Field(
20502153
proto.STRING,
20512154
number=1,
@@ -2055,6 +2158,11 @@ class LayoutListEntry(proto.Message):
20552158
number=5,
20562159
message="Document.DocumentLayout.DocumentLayoutBlock.LayoutPageSpan",
20572160
)
2161+
bounding_box: geometry.BoundingPoly = proto.Field(
2162+
proto.MESSAGE,
2163+
number=6,
2164+
message=geometry.BoundingPoly,
2165+
)
20582166

20592167
blocks: MutableSequence[
20602168
"Document.DocumentLayout.DocumentLayoutBlock"
@@ -2088,6 +2196,8 @@ class Chunk(proto.Message):
20882196
Page headers associated with the chunk.
20892197
page_footers (MutableSequence[google.cloud.documentai_v1beta3.types.Document.ChunkedDocument.Chunk.ChunkPageFooter]):
20902198
Page footers associated with the chunk.
2199+
chunk_fields (MutableSequence[google.cloud.documentai_v1beta3.types.Document.ChunkedDocument.Chunk.ChunkField]):
2200+
Chunk fields inside this chunk.
20912201
"""
20922202

20932203
class ChunkPageSpan(proto.Message):
@@ -2149,6 +2259,112 @@ class ChunkPageFooter(proto.Message):
21492259
message="Document.ChunkedDocument.Chunk.ChunkPageSpan",
21502260
)
21512261

2262+
class ImageChunkField(proto.Message):
2263+
r"""The image chunk field in the chunk.
2264+
2265+
This message has `oneof`_ fields (mutually exclusive fields).
2266+
For each oneof, at most one member field can be set at the same time.
2267+
Setting any member of the oneof automatically clears all other
2268+
members.
2269+
2270+
.. _oneof: https://proto-plus-python.readthedocs.io/en/stable/fields.html#oneofs-mutually-exclusive-fields
2271+
2272+
Attributes:
2273+
blob_asset_id (str):
2274+
Optional. Asset id of the inline image. If set, find the
2275+
image content in the blob_assets field.
2276+
2277+
This field is a member of `oneof`_ ``image_source``.
2278+
gcs_uri (str):
2279+
Optional. Google Cloud Storage uri of the
2280+
image.
2281+
2282+
This field is a member of `oneof`_ ``image_source``.
2283+
data_uri (str):
2284+
Optional. Data uri of the image. It is composed of four
2285+
parts: a prefix (data:), a MIME type indicating the type of
2286+
data, an optional base64 token if non-textual, and the data
2287+
itself: data:[][;base64],
2288+
2289+
This field is a member of `oneof`_ ``image_source``.
2290+
annotations (google.cloud.documentai_v1beta3.types.Document.Annotations):
2291+
Annotation of the image chunk field.
2292+
"""
2293+
2294+
blob_asset_id: str = proto.Field(
2295+
proto.STRING,
2296+
number=1,
2297+
oneof="image_source",
2298+
)
2299+
gcs_uri: str = proto.Field(
2300+
proto.STRING,
2301+
number=2,
2302+
oneof="image_source",
2303+
)
2304+
data_uri: str = proto.Field(
2305+
proto.STRING,
2306+
number=3,
2307+
oneof="image_source",
2308+
)
2309+
annotations: "Document.Annotations" = proto.Field(
2310+
proto.MESSAGE,
2311+
number=4,
2312+
message="Document.Annotations",
2313+
)
2314+
2315+
class TableChunkField(proto.Message):
2316+
r"""The table chunk field in the chunk.
2317+
2318+
Attributes:
2319+
annotations (google.cloud.documentai_v1beta3.types.Document.Annotations):
2320+
Annotation of the table chunk field.
2321+
"""
2322+
2323+
annotations: "Document.Annotations" = proto.Field(
2324+
proto.MESSAGE,
2325+
number=1,
2326+
message="Document.Annotations",
2327+
)
2328+
2329+
class ChunkField(proto.Message):
2330+
r"""The chunk field in the chunk. A chunk field could be one of
2331+
the various types (e.g. image, table) supported.
2332+
2333+
This message has `oneof`_ fields (mutually exclusive fields).
2334+
For each oneof, at most one member field can be set at the same time.
2335+
Setting any member of the oneof automatically clears all other
2336+
members.
2337+
2338+
.. _oneof: https://proto-plus-python.readthedocs.io/en/stable/fields.html#oneofs-mutually-exclusive-fields
2339+
2340+
Attributes:
2341+
image_chunk_field (google.cloud.documentai_v1beta3.types.Document.ChunkedDocument.Chunk.ImageChunkField):
2342+
The image chunk field in the chunk.
2343+
2344+
This field is a member of `oneof`_ ``field_type``.
2345+
table_chunk_field (google.cloud.documentai_v1beta3.types.Document.ChunkedDocument.Chunk.TableChunkField):
2346+
The table chunk field in the chunk.
2347+
2348+
This field is a member of `oneof`_ ``field_type``.
2349+
"""
2350+
2351+
image_chunk_field: "Document.ChunkedDocument.Chunk.ImageChunkField" = (
2352+
proto.Field(
2353+
proto.MESSAGE,
2354+
number=1,
2355+
oneof="field_type",
2356+
message="Document.ChunkedDocument.Chunk.ImageChunkField",
2357+
)
2358+
)
2359+
table_chunk_field: "Document.ChunkedDocument.Chunk.TableChunkField" = (
2360+
proto.Field(
2361+
proto.MESSAGE,
2362+
number=2,
2363+
oneof="field_type",
2364+
message="Document.ChunkedDocument.Chunk.TableChunkField",
2365+
)
2366+
)
2367+
21522368
chunk_id: str = proto.Field(
21532369
proto.STRING,
21542370
number=1,
@@ -2180,13 +2396,50 @@ class ChunkPageFooter(proto.Message):
21802396
number=6,
21812397
message="Document.ChunkedDocument.Chunk.ChunkPageFooter",
21822398
)
2399+
chunk_fields: MutableSequence[
2400+
"Document.ChunkedDocument.Chunk.ChunkField"
2401+
] = proto.RepeatedField(
2402+
proto.MESSAGE,
2403+
number=7,
2404+
message="Document.ChunkedDocument.Chunk.ChunkField",
2405+
)
21832406

21842407
chunks: MutableSequence["Document.ChunkedDocument.Chunk"] = proto.RepeatedField(
21852408
proto.MESSAGE,
21862409
number=1,
21872410
message="Document.ChunkedDocument.Chunk",
21882411
)
21892412

2413+
class BlobAsset(proto.Message):
2414+
r"""Represents a blob asset. It's used to store the content of
2415+
the inline blob in this document, e.g. image bytes, such that it
2416+
can be referenced by other fields in the document via asset id.
2417+
2418+
Attributes:
2419+
asset_id (str):
2420+
Optional. The id of the blob asset.
2421+
content (bytes):
2422+
Optional. The content of the blob asset, e.g.
2423+
image bytes.
2424+
mime_type (str):
2425+
The mime type of the blob asset. An IANA published `media
2426+
type (MIME
2427+
type) <https://www.iana.org/assignments/media-types/media-types.xhtml>`__.
2428+
"""
2429+
2430+
asset_id: str = proto.Field(
2431+
proto.STRING,
2432+
number=1,
2433+
)
2434+
content: bytes = proto.Field(
2435+
proto.BYTES,
2436+
number=2,
2437+
)
2438+
mime_type: str = proto.Field(
2439+
proto.STRING,
2440+
number=3,
2441+
)
2442+
21902443
uri: str = proto.Field(
21912444
proto.STRING,
21922445
number=1,
@@ -2197,6 +2450,10 @@ class ChunkPageFooter(proto.Message):
21972450
number=2,
21982451
oneof="source",
21992452
)
2453+
docid: str = proto.Field(
2454+
proto.STRING,
2455+
number=15,
2456+
)
22002457
mime_type: str = proto.Field(
22012458
proto.STRING,
22022459
number=3,
@@ -2255,6 +2512,11 @@ class ChunkPageFooter(proto.Message):
22552512
number=18,
22562513
message=ChunkedDocument,
22572514
)
2515+
blob_assets: MutableSequence[BlobAsset] = proto.RepeatedField(
2516+
proto.MESSAGE,
2517+
number=19,
2518+
message=BlobAsset,
2519+
)
22582520

22592521

22602522
class RevisionRef(proto.Message):

‎packages/google-cloud-documentai/google/cloud/documentai_v1beta3/types/document_processor_service.py‎

Lines changed: 7 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -144,6 +144,9 @@ class LayoutConfig(proto.Message):
144144
return_bounding_boxes (bool):
145145
Optional. Whether to include bounding boxes
146146
in layout parser processor response.
147+
enable_image_annotation (bool):
148+
Optional. Whether to include image
149+
annotations in layout parser response.
147150
enable_llm_layout_parsing (bool):
148151
Optional. Whether to refine PDF layout using
149152
LLM.
@@ -201,6 +204,10 @@ class ChunkingConfig(proto.Message):
201204
proto.BOOL,
202205
number=3,
203206
)
207+
enable_image_annotation: bool = proto.Field(
208+
proto.BOOL,
209+
number=4,
210+
)
204211
enable_llm_layout_parsing: bool = proto.Field(
205212
proto.BOOL,
206213
number=5,

0 commit comments

Comments
 (0)