Updates to vector storage optimization, rescoring with original vectors

Author: HeidiSteen

articles/search/vector-search-how-to-assign-narrow-data-types.md

@@ -7,18 +7,18 @@ author: heidisteen
 ms.author: heidist
 ms.service: azure-ai-search
 ms.topic: how-to
-ms.date: 11/04/2024
+ms.date: 11/19/2024
 ---
 
-# Assign narrow data types
+# Assign narrow data types to vector fields in Azure AI Search
 
 An easy way to reduce vector size is to store embeddings in a smaller data format. Most embedding models output 32-bit floating point numbers, but if you quantize your vectors, or if your embedding model supports it natively, output might be float16, int16, or int8, which is significantly smaller than float32. You can accommodate these smaller vector sizes by assigning a narrow data type to a vector field. In the vector index, narrow data types consume less storage.
 
 Data types are assigned to fields in an index definition. You can use the Azure portal, the [Search REST APIs](/rest/api/searchservice/indexes/create), or an Azure SDK package that provides the feature.
 
 ## Prerequisites
 
-- An embedding model that output small data formats.
+- An embedding model that output small data formats, such as text-embedding-3 or Cohere V3 embedding models.
 
 ## Supported narrow data types
 
@@ -47,7 +47,7 @@ Data types are assigned to fields in an index definition. You can use the Azure
 
 ## Assign the data type
 
-[Define and build the index](vector-search-how-to-create-index.md). You can use the Azure portal, [Create or Update Index (REST API)](/rest/api/searchservice/indexes/create-or-update), or an Azure SDK package for this step.
+[Define and build an index](vector-search-how-to-create-index.md). You can use the Azure portal, [Create or Update Index (REST API)](/rest/api/searchservice/indexes/create-or-update), or an Azure SDK package for this step.
 
 This field definition uses a narrow data type, `Collection(Edm.Half)`, that can accept a float32 embedding stored as a float16 value. As is true for all vector fields, `dimensions` and `vectorSearchProfile` are set. The specifics of the `vectorSearchProfile` are immaterial to the datatype.
 
@@ -80,12 +80,9 @@ Data types are assigned on new fields when they're created. You can't change the
 
 ## Check results
 
-1. Verify the field content matches the data type. Assuming the vector field is marked as retrievable, use [Search explorer](search-explorer.md) or [Search - POST](/rest/api/searchservice/documents/search-post?) to return vector field content.
+1. Verify the field content matches the data type. Assuming the vector field is marked as `retrievable`, use [Search explorer](search-explorer.md) or [Search - POST](/rest/api/searchservice/documents/search-post?) to return vector field content.
 
-1. To check vector index size, refer to the vector index size column on the Indexes page in the Azure portal or use the [GET Statistics (REST API)](/rest/api/searchservice/indexes/get-statistics) or equivalent Azure SDK method to get the size.
-
-<!-- 
-   Evidence of choosing the wrong data type, for example choosing `int8` for a `float32` embedding, is a field that's indexed as an array of zeros. If you encounter this problem, start over. -->
+1. To check vector index size, refer to the vector index size column on the **Search management > Indexes** page in the [Azure portal](https://portal.azure.com) or use the [GET Statistics (REST API)](/rest/api/searchservice/indexes/get-statistics) or equivalent Azure SDK method to get the size.
 
 > [!NOTE]
-> The field's data type is used to create the physical data structure. If you want to change a data type later, either drop and rebuild the index, or create a second field with the new definition.
+> The field's data type is used to create the physical data structure. If you want to change a data type later, either [drop and rebuild the index](search-howto-reindex.md), or create a second field with the new definition.

articles/search/vector-search-how-to-index-binary-data.md

@@ -9,30 +9,30 @@ ms.service: azure-ai-search
 ms.custom:
   - build-2024
 ms.topic: how-to
-ms.date: 08/05/2024
+ms.date: 11/19/2024
 ---
 
 # Index binary vectors for vector search
 
-Azure AI Search supports a packed binary type of `Collection(Edm.Byte)` for further reducing the storage and memory footprint of vector data. You can use this data type for output from models such as [Cohere's Embed v3 binary embedding models](https://cohere.com/blog/introducing-embed-v3). 
+Azure AI Search supports a packed binary type of `Collection(Edm.Byte)` for further reducing the storage and memory footprint of vector data. You can use this data type for output from models such as [Cohere's Embed v3 binary embedding models](https://cohere.com/blog/introducing-embed-v3) or any other embedding model or process that outputs vectors as binary bytes.
 
 There are three steps to configuring an index for binary vectors:
 
 > [!div class="checklist"]
 > + Add a vector search algorithm that specifies Hamming distance for binary vector comparison
 > + Add a vector profile that points to the algorithm
-> + Add the vector profile to your binary field definition
+> + Add a vector field of type `Collection(Edm.Byte)` and assign the Hamming distance
 
-This article assumes you're familiar with [creating an index in Azure AI Search](search-how-to-create-search-index.md). It uses the REST APIs to illustrate each step, but you could also add a binary field to an index in the Azure portal.
+This article assumes you're familiar with [creating an index in Azure AI Search](search-how-to-create-search-index.md) and [adding vector fields](vector-search-how-to-create-index.md). It uses the REST APIs to illustrate each step, but you could also add a binary field to an index in the Azure portal or Azure SDK.
 
-Binary data types are generally available starting with API version 2024-07-01 and are assigned to fields using the [Create Index](/rest/api/searchservice/indexes/create) or [Create Or Update Index](/rest/api/searchservice/indexes/create-or-update) APIs.
+The binary data type is generally available starting with API version 2024-07-01 and is assigned to fields using the [Create Index](/rest/api/searchservice/indexes/create) or [Create Or Update Index](/rest/api/searchservice/indexes/create-or-update) APIs.
 
 > [!TIP]
-> If you're investigating binary vector support for its smaller footprint, you might also consider the vector quantization and storage reduction features in Azure AI Search. Inputs are float32 or float16 embeddings. Output is stored data in a much smaller format. For more information, see [Assign narrow data types](vector-search-how-to-assign-narrow-data-types.md).
+> If you're investigating binary vector support for its smaller footprint, you might also consider the vector quantization and storage reduction features in Azure AI Search. Inputs are float32 or float16 embeddings. Output is stored data in a much smaller format. For more information, see [Compress using binary or scalar quantization](vector-search-how-to-quantization.md) and [Assign narrow data types](vector-search-how-to-assign-narrow-data-types.md).
 
 ## Prerequisites
 
-+ Binary vectors, with 1 bit per dimension, packaged in uint8 values with 8 bits per value. These can be obtained by using models that directly generate "packaged binary" vectors, or by quantizing vectors into binary vectors client-side during indexing and searching.
++ Binary vectors, with 1 bit per dimension, packaged in uint8 values with 8 bits per value. These can be obtained by using models that directly generate *packaged binary* vectors, or by quantizing vectors into binary vectors client-side during indexing and searching.
 
 ## Limitations
 

articles/search/vector-search-how-to-quantization.md

@@ -1,38 +1,44 @@
 ---
-title: Quantize vector fields
+title: Compress vectors using quantization
 titleSuffix: Azure AI Search
 description: Configure built-in scalar or quantization for compressing vectors on disk and in memory.
 
 author: heidisteen
 ms.author: heidist
 ms.service: azure-ai-search
 ms.topic: how-to
-ms.date: 11/04/2024
+ms.date: 11/19/2024
 ---
 
-# Use scalar or binary quantization to compress vector size
+# Compress vectors using scalar or binary quantization
 
-Quantization is recommended for reducing vector size because it lowers both memory and disk storage requirements for float16 and float32 embeddings. To offset the effects of a smaller index, you can add oversampling and reranking over uncompressed vectors.
-
-Quantization applies to vector fields receiving float-type vectors. In the examples in this article, the field's data type is `Collection(Edm.Single)` for incoming float32 embeddings, but float16 is also supported. When the vectors are received on a field with compression configured, the engine automatically performs quantization to reduce the footprint of the vector data in memory and on disk.
-
-Two types of quantization are supported:
-
-- Scalar quantization compresses float values into narrower data types. AI Search currently supports int8, which is 8 bits, reducing vector index size fourfold.
-
-- Binary quantization converts floats into binary bits, which takes up 1 bit. This results in up to 28 times reduced vector index size.
+Azure AI Search supports scalar and binary quantization for reducing the size of vectors in a search index. Quantization is recommended for reducing vector size because it lowers both memory and disk storage consumption for float16 and float32 embeddings. To offset the effects of a smaller index, you can add oversampling and reranking over uncompressed vectors.
 
 To use built-in quantization, follow these steps:
 
 > [!div class="checklist"]
-> - Use [Create Index](/rest/api/searchservice/indexes/create) or [Create Or Update Index](/rest/api/searchservice/indexes/create-or-update) to specify vector compression
-> - Add `vectorSearch.compressions` to a search index
+> - Add [vector fields and a `vectorSearch` configuration](vector-search-how-to-create-index.md) to an index
+> - Add `vectorSearch.compressions`
 > - Add a `scalarQuantization` or `binaryQuantization` configuration and give it a name
 > - Set optional properties to mitigate the effects of lossy indexing
 > - Create a new vector profile that uses the named configuration
 > - Create a new vector field having the new vector profile
 > - Load the index with float32 or float16 data that's quantized during indexing with the configuration you defined
-> - Optionally, [query quantized data](#) using the oversampling parameter if you want to override the default
+> - Optionally, [query quantized data](#query-a-quantized-vector-field-using-oversampling) using the oversampling parameter if you want to override the default
+
+## Prerequisites
+
+- [Vector fields in a search index](vector-search-how-to-create-index.md) with a `vectorSearch` configuration, using the HNSW algorithm and a new vector profile.
+
+## Supported quantization techniques
+
+Quantization applies to vector fields receiving float-type vectors. In the examples in this article, the field's data type is `Collection(Edm.Single)` for incoming float32 embeddings, but float16 is also supported. When the vectors are received on a field with compression configured, the engine automatically performs quantization to reduce the footprint of the vector data in memory and on disk.
+
+Two types of quantization are supported:
+
+- Scalar quantization compresses float values into narrower data types. AI Search currently supports int8, which is 8 bits, reducing vector index size fourfold.
+
+- Binary quantization converts floats into binary bits, which takes up 1 bit. This results in up to 28 times reduced vector index size.
 
 ## Add "compressions" to a search index
 
@@ -76,7 +82,7 @@ POST https://[servicename].search.windows.net/indexes?api-version=2024-07-01
 
 **Key points**:
 
-- `kind` must be set to `scalarQuantization` or `binaryQuantization`
+- `kind` must be set to `scalarQuantization` or `binaryQuantization`.
 
 - `rerankWithOriginalVectors` uses the original, uncompressed vectors to recalculate similarity and rerank the top results returned by the initial search query. The uncompressed vectors exist in the search index even if `stored` is false. This property is optional. Default is true.
 
@@ -227,3 +233,17 @@ POST https://[service-name].search.windows.net/indexes/demo-index/docs/search?ap
 - Applies to vector fields that undergo vector compression, per the vector profile assignment.
 
 - Overrides the `defaultOversampling` value or introduces oversampling at query time, even if the index's compression configuration didn't specify oversampling or reranking options.
+
+<!-- 
+RESCORE WITH ORIGINAL VECTORS -- NEEDS AN H2 or H3
+It's used to rescore search results obtained used compressed vectors.
+
+Rescore with original vectors
+After the initial query, rescore results using uncompressed vectors
+ 
+For "enableRescoring", we provide true or false options. if it's true, the query will first retrieve using compressed vectors, then rescore results using uncompressed vectors.
+
+Step one: Vector query executes using the compressed vectors.
+Step two: Query returns the top oversampling k-matches.
+Step three: Oversampling k-matches are rescored using the uncompressed vectors, adjusting the scores and ranking so that more relevant matches appear first.
+ -->