feat: Add image block and blob asset in Document proto

Google APIs · copybara-github · commit 37b9fc529cb8 · 2025-04-23T10:19:51.000-07:00
PiperOrigin-RevId: 750637959
diff --git a/google/cloud/documentai/v1beta3/document.proto b/google/cloud/documentai/v1beta3/document.proto
@@ -897,6 +897,12 @@ message Document {
     repeated Provenance provenance = 3 [deprecated = true];
   }
 
+  // Represents the annotation of a block or a chunk.
+  message Annotations {
+    // The description of the content with this annotation.
+    string description = 1;
+  }
+
   // Represents the parsed layout of a document as a collection of blocks that
   // the document is divided into.
   message DocumentLayout {
@@ -975,6 +981,37 @@ message Document {
         repeated DocumentLayoutBlock blocks = 1;
       }
 
+      // Represents an image type block.
+      message LayoutImageBlock {
+        // Source of the image.
+        oneof image_source {
+          // Optional. Asset id of the inline image. If set, find the image
+          // content in the blob_assets field.
+          string blob_asset_id = 4 [(google.api.field_behavior) = OPTIONAL];
+
+          // Optional. Google Cloud Storage uri of the image.
+          string gcs_uri = 5 [(google.api.field_behavior) = OPTIONAL];
+
+          // Optional. Data uri of the image.
+          // It is composed of four parts: a prefix (data:), a MIME type
+          // indicating the type of data, an optional base64 token if
+          // non-textual, and the data itself:
+          // data:[<mediatype>][;base64],<data>
+          string data_uri = 6 [(google.api.field_behavior) = OPTIONAL];
+        }
+
+        // Mime type of the image. An IANA published [media type (MIME type)]
+        // (https://www.iana.org/assignments/media-types/media-types.xhtml).
+        string mime_type = 1;
+
+        // Text extracted from the image using OCR or alt text describing the
+        // image.
+        string image_text = 2;
+
+        // Annotation of the image block.
+        Annotations annotations = 3;
+      }
+
       oneof block {
         // Block consisting of text content.
         LayoutTextBlock text_block = 2;
@@ -984,13 +1021,19 @@ message Document {
 
         // Block consisting of list content/structure.
         LayoutListBlock list_block = 4;
+
+        // Block consisting of image content.
+        LayoutImageBlock image_block = 7;
       }
 
       // ID of the block.
       string block_id = 1;
 
       // Page span of the block.
       LayoutPageSpan page_span = 5;
+
+      // Identifies the bounding box for the block.
+      BoundingPoly bounding_box = 6;
     }
 
     // List of blocks in the document.
@@ -1028,6 +1071,48 @@ message Document {
         ChunkPageSpan page_span = 2;
       }
 
+      // The image chunk field in the chunk.
+      message ImageChunkField {
+        // Source of the image.
+        oneof image_source {
+          // Optional. Asset id of the inline image. If set, find the image
+          // content in the blob_assets field.
+          string blob_asset_id = 1 [(google.api.field_behavior) = OPTIONAL];
+
+          // Optional. Google Cloud Storage uri of the image.
+          string gcs_uri = 2 [(google.api.field_behavior) = OPTIONAL];
+
+          // Optional. Data uri of the image.
+          // It is composed of four parts: a prefix (data:), a MIME type
+          // indicating the type of data, an optional base64 token if
+          // non-textual, and the data itself:
+          // data:[<mediatype>][;base64],<data>
+          string data_uri = 3 [(google.api.field_behavior) = OPTIONAL];
+        }
+
+        // Annotation of the image chunk field.
+        Annotations annotations = 4;
+      }
+
+      // The table chunk field in the chunk.
+      message TableChunkField {
+        // Annotation of the table chunk field.
+        Annotations annotations = 1;
+      }
+
+      // The chunk field in the chunk. A chunk field could be one of the various
+      // types (e.g. image, table) supported.
+      message ChunkField {
+        // The type of the chunk field.
+        oneof field_type {
+          // The image chunk field in the chunk.
+          ImageChunkField image_chunk_field = 1;
+
+          // The table chunk field in the chunk.
+          TableChunkField table_chunk_field = 2;
+        }
+      }
+
       // ID of the chunk.
       string chunk_id = 1;
 
@@ -1045,12 +1130,31 @@ message Document {
 
       // Page footers associated with the chunk.
       repeated ChunkPageFooter page_footers = 6;
+
+      // Chunk fields inside this chunk.
+      repeated ChunkField chunk_fields = 7;
     }
 
     // List of chunks.
     repeated Chunk chunks = 1;
   }
 
+  // Represents a blob asset. It's used to store the content of the inline blob
+  // in this document, e.g. image bytes, such that it can be referenced by
+  // other fields in the document via asset id.
+  message BlobAsset {
+    // Optional. The id of the blob asset.
+    string asset_id = 1 [(google.api.field_behavior) = OPTIONAL];
+
+    // Optional. The content of the blob asset, e.g. image bytes.
+    bytes content = 2 [(google.api.field_behavior) = OPTIONAL];
+
+    // The mime type of the blob asset.
+    // An IANA published [media type (MIME
+    // type)](https://www.iana.org/assignments/media-types/media-types.xhtml).
+    string mime_type = 3;
+  }
+
   // Original source document from the user.
   oneof source {
     // Optional. Currently supports Google Cloud Storage URI of the form
@@ -1065,6 +1169,9 @@ message Document {
     bytes content = 2 [(google.api.field_behavior) = OPTIONAL];
   }
 
+  // Optional. An internal identifier for document. Should be loggable (no PII).
+  string docid = 15 [(google.api.field_behavior) = OPTIONAL];
+
   // An IANA published [media type (MIME
   // type)](https://www.iana.org/assignments/media-types/media-types.xhtml).
   string mime_type = 3;
@@ -1110,6 +1217,11 @@ message Document {
 
   // Document chunked based on chunking config.
   ChunkedDocument chunked_document = 18;
+
+  // Optional. The blob assets in this document. This is used to store the
+  // content of the inline blobs in this document, e.g. image bytes, such that
+  // it can be referenced by other fields in the document via asset id.
+  repeated BlobAsset blob_assets = 19 [(google.api.field_behavior) = OPTIONAL];
 }
 
 // The revision reference specifies which revision on the document to read.
