From 4cb84340eb2908f2e67b44123fe0af08cfb89f9b Mon Sep 17 00:00:00 2001
From: Jonathan Buttner <56361221+jonathan-buttner@users.noreply.github.com>
Date: Thu, 8 Feb 2024 12:20:45 -0500
Subject: [PATCH] [ML] Adding text embedding byte representation (#2411)

* Adding text embedding byte representation

* Removing semi-colon
---
 output/schema/schema.json                 | 59 ++++++++++++++++++++++-
 output/typescript/types.ts                |  7 +++
 specification/inference/_types/Results.ts | 16 +++++-
 3 files changed, 79 insertions(+), 3 deletions(-)

diff --git a/output/schema/schema.json b/output/schema/schema.json
index 4e94e19611..d8f4dd8778 100644
--- a/output/schema/schema.json
+++ b/output/schema/schema.json
@@ -127831,6 +127831,25 @@
       },
       "specLocation": "indices/validate_query/IndicesValidateQueryResponse.ts#L23-L30"
     },
+    {
+      "description": "Text Embedding results containing bytes are represented as Dense\nVectors of bytes.",
+      "kind": "type_alias",
+      "name": {
+        "name": "DenseByteVector",
+        "namespace": "inference._types"
+      },
+      "specLocation": "inference/_types/Results.ts#L39-L43",
+      "type": {
+        "kind": "array_of",
+        "value": {
+          "kind": "instance_of",
+          "type": {
+            "name": "byte",
+            "namespace": "_types"
+          }
+        }
+      }
+    },
     {
       "description": "Text Embedding results are represented as Dense Vectors\nof floats.",
       "kind": "type_alias",
@@ -127858,6 +127877,20 @@
         "namespace": "inference._types"
       },
       "properties": [
+        {
+          "name": "text_embedding_bytes",
+          "required": false,
+          "type": {
+            "kind": "array_of",
+            "value": {
+              "kind": "instance_of",
+              "type": {
+                "name": "TextEmbeddingByteResult",
+                "namespace": "inference._types"
+              }
+            }
+          }
+        },
         {
           "name": "text_embedding",
           "required": false,
@@ -127887,7 +127920,7 @@
           }
         }
       ],
-      "specLocation": "inference/_types/Results.ts#L46-L53",
+      "specLocation": "inference/_types/Results.ts#L59-L67",
       "variants": {
         "kind": "container"
       }
@@ -128066,6 +128099,28 @@
       },
       "specLocation": "inference/_types/TaskType.ts#L20-L26"
     },
+    {
+      "description": "The text embedding result object for byte representation",
+      "kind": "interface",
+      "name": {
+        "name": "TextEmbeddingByteResult",
+        "namespace": "inference._types"
+      },
+      "properties": [
+        {
+          "name": "embedding",
+          "required": true,
+          "type": {
+            "kind": "instance_of",
+            "type": {
+              "name": "DenseByteVector",
+              "namespace": "inference._types"
+            }
+          }
+        }
+      ],
+      "specLocation": "inference/_types/Results.ts#L45-L50"
+    },
     {
       "description": "The text embedding result object",
       "kind": "interface",
@@ -128086,7 +128141,7 @@
           }
         }
       ],
-      "specLocation": "inference/_types/Results.ts#L39-L44"
+      "specLocation": "inference/_types/Results.ts#L52-L57"
     },
     {
       "attachedBehaviors": [
diff --git a/output/typescript/types.ts b/output/typescript/types.ts
index 9498f1d415..4332c5c735 100644
--- a/output/typescript/types.ts
+++ b/output/typescript/types.ts
@@ -11464,9 +11464,12 @@ export interface IndicesValidateQueryResponse {
   error?: string
 }
 
+export type InferenceDenseByteVector = byte[]
+
 export type InferenceDenseVector = float[]
 
 export interface InferenceInferenceResult {
+  text_embedding_bytes?: InferenceTextEmbeddingByteResult[]
   text_embedding?: InferenceTextEmbeddingResult[]
   sparse_embedding?: InferenceSparseEmbeddingResult[]
 }
@@ -11494,6 +11497,10 @@ export type InferenceTaskSettings = any
 
 export type InferenceTaskType = 'sparse_embedding' | 'text_embedding'
 
+export interface InferenceTextEmbeddingByteResult {
+  embedding: InferenceDenseByteVector
+}
+
 export interface InferenceTextEmbeddingResult {
   embedding: InferenceDenseVector
 }
diff --git a/specification/inference/_types/Results.ts b/specification/inference/_types/Results.ts
index e16d49ee9b..b7cfd981a4 100644
--- a/specification/inference/_types/Results.ts
+++ b/specification/inference/_types/Results.ts
@@ -17,7 +17,7 @@
  * under the License.
  */
 
-import { float } from '@_types/Numeric'
+import { float, byte } from '@_types/Numeric'
 import { Dictionary } from '@spec_utils/Dictionary'
 
 /**
@@ -36,6 +36,19 @@ export class SparseEmbeddingResult {
   embedding: SparseVector
 }
 
+/**
+ * Text Embedding results containing bytes are represented as Dense
+ * Vectors of bytes.
+ */
+export type DenseByteVector = Array<byte>
+
+/**
+ * The text embedding result object for byte representation
+ */
+export class TextEmbeddingByteResult {
+  embedding: DenseByteVector
+}
+
 /**
  * The text embedding result object
  */
@@ -48,6 +61,7 @@ export class TextEmbeddingResult {
  * @variants container
  */
 export class InferenceResult {
+  text_embedding_bytes?: Array<TextEmbeddingByteResult>
   text_embedding?: Array<TextEmbeddingResult>
   sparse_embedding?: Array<SparseEmbeddingResult>
 }