Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

.Net: Added implementation of SQLite connector for new memory design #9164

Merged
merged 128 commits into from
Oct 17, 2024
Merged
Show file tree
Hide file tree
Changes from 1 commit
Commits
Show all changes
128 commits
Select commit Hold shift + click to select a range
d7287bb
.Net: ADR for Text Search Abstractions (#8307)
markwallace-microsoft Aug 22, 2024
590745f
.Net: Add TextSearchExtension methods to create KernelPlugins and Ker…
markwallace-microsoft Aug 27, 2024
3c09d91
.Net: Update Bing Search to new Text Search Design (#8343)
markwallace-microsoft Aug 29, 2024
187afca
Merge branch 'main' into feature-vector-search
markwallace-microsoft Aug 29, 2024
16f4fa1
.Net: Update Google Search to new Text Search Design (#8394)
markwallace-microsoft Aug 29, 2024
5af6540
.Net: Add text search concepts demonstrating RAG and Function Calling…
markwallace-microsoft Aug 30, 2024
6d13254
Merge branch 'main' into feature-vector-search
markwallace-microsoft Aug 30, 2024
f806fc2
.Net: Simplify how additional text search filter parameters are speci…
markwallace-microsoft Sep 2, 2024
890c0f8
Merge branch 'main' into feature-vector-search
markwallace-microsoft Sep 2, 2024
cfbac04
Merge branch 'main' into feature-vector-search
markwallace-microsoft Sep 3, 2024
b96d607
Merge branch 'main' into feature-vector-search
markwallace-microsoft Sep 3, 2024
4b8406e
.Net: Add Vector Search ADR document and abstractions (#8494)
westey-m Sep 4, 2024
d02005a
.Net: Add support for advanced search keywords (#8489)
markwallace-microsoft Sep 4, 2024
ac540db
.Net: Use interfaces for mapper abstractions (#8424)
markwallace-microsoft Sep 4, 2024
d745d57
.Net: Add vector search implementation for azure ai search (#8507)
westey-m Sep 5, 2024
7836721
.Net: Add qdrant vector search implementation. (#8508)
westey-m Sep 5, 2024
ec10424
.Net: Add Redis vector search implementation (#8510)
westey-m Sep 5, 2024
d43c5ff
.Net: Unify text and vector search and move all to data namespace. (#…
westey-m Sep 6, 2024
70a98ee
.Net: Add volatile vector search and enforcing collection type. (#8546)
westey-m Sep 6, 2024
ffaa93f
Update adr document with filter changes. (#8590)
westey-m Sep 9, 2024
639694b
Resolve merge conflicts
markwallace-microsoft Sep 11, 2024
c1cd203
Fix formatting
markwallace-microsoft Sep 11, 2024
7bdf519
Merge branch 'main' into feature-vector-search
westey-m Sep 12, 2024
6b92a18
Merge branch 'main' into feature-vector-search
markwallace-microsoft Sep 13, 2024
1fc7e91
.Net: Switch to using interfaces for search instead of query objects.…
westey-m Sep 13, 2024
9c8e438
Merge branch 'main' into feature-vector-search
markwallace-microsoft Sep 13, 2024
031a653
Merge branch 'main' into feature-vector-search
westey-m Sep 17, 2024
971afd6
.Net: Suppress moq warning after taking package updates. (#8838)
westey-m Sep 17, 2024
bf34a51
.Net: Basic implementation of ITextSearch for an IVectorStore impleme…
markwallace-microsoft Sep 17, 2024
434498a
.Net: Add float64 support for redis vector search. (#8847)
westey-m Sep 17, 2024
092a9ac
Merge branch 'main' into feature-vector-search
dmytrostruk Sep 18, 2024
f35d051
.Net: Enhance volatile memory connector to allow collection to be ser…
markwallace-microsoft Sep 18, 2024
8dc7dc3
.Net: Added vector search implementation for Azure CosmosDB for Mongo…
dmytrostruk Sep 18, 2024
fa461ce
.Net: Vector Search Bug fixes (#8890)
westey-m Sep 18, 2024
be531dd
.Net: [Feature branch] Added support for Filter and Offset parameters…
dmytrostruk Sep 19, 2024
36aa3f2
.Net: Vector search sample: Multi Vector and Paging (#8920)
westey-m Sep 20, 2024
8fa1e8e
.Net: Adding a multi-store vector search sample. (#8909)
westey-m Sep 20, 2024
b5c649f
.Net: [Feature Branch] Added vector search implementation for Azure C…
dmytrostruk Sep 20, 2024
952d679
Merge branch 'main' into feature-vector-search
markwallace-microsoft Sep 20, 2024
6c4aa5d
.Net: Add VectorStoreTextSearch concepts and unit tests (#8891)
markwallace-microsoft Sep 23, 2024
cb00332
Merge branch 'main' into feature-vector-search
westey-m Sep 23, 2024
21f8a27
.Net: [Feature Branch] Added vector search implementation for Weaviat…
dmytrostruk Sep 23, 2024
d6b6019
.Net: Add filtering support for VectoreStoreTextSearch (#8947)
markwallace-microsoft Sep 24, 2024
1c35c01
Merge branch 'main' into feature-vector-search
markwallace-microsoft Sep 24, 2024
6d041b3
.Net: Adding generic data type support to the volatile vector store. …
westey-m Sep 24, 2024
16f787e
.Net: Rename a number of properties as agreed and remove dangerous de…
westey-m Sep 25, 2024
12a85d5
Merge branch 'main' into feature-vector-search
dmytrostruk Sep 26, 2024
66d608c
.Net: Add delegates for the string and result mappers (#9000)
markwallace-microsoft Sep 26, 2024
d08eef3
.Net: Integration tests for Bing and Google text search (#8991)
markwallace-microsoft Sep 27, 2024
21e37a0
.Net: Add DI registration helpers for collections and search (#9007)
westey-m Sep 27, 2024
a6dc6b3
Merge branch 'main' into feature-vector-search
dmytrostruk Sep 30, 2024
2758ff7
Fix after merge
dmytrostruk Sep 30, 2024
da7291b
Implemented database manipulation methods
dmytrostruk Oct 1, 2024
667533e
Merge branch 'main' into feature-vector-search
markwallace-microsoft Oct 1, 2024
7975fed
Merge branch 'main' into feature-vector-search
markwallace-microsoft Oct 1, 2024
6931b79
Merge branch 'main' into feature-vector-search
markwallace-microsoft Oct 1, 2024
5bce499
.Net: Fix issues after vector split merge. (#9050)
westey-m Oct 1, 2024
1046bd3
Merge branch 'feature-vector-search' into sqlite-connector
dmytrostruk Oct 1, 2024
467191e
.Net: Support DI for Text Search Services (#9026)
markwallace-microsoft Oct 1, 2024
a608659
Fixed compilation errors after merge
dmytrostruk Oct 1, 2024
136fc0f
Added creation of virtual table for vectors
dmytrostruk Oct 1, 2024
d6ec1ee
Added deletion of virtual table for vectors
dmytrostruk Oct 1, 2024
2b94b08
.Net: Adapt In-Memory Connector to new Text Search Design (#9046)
markwallace-microsoft Oct 2, 2024
7710e36
Merge branch 'main' into feature-vector-search
markwallace-microsoft Oct 2, 2024
e2e22b6
.Net: Vector search recordreader manual merge (#9059)
westey-m Oct 2, 2024
6b05639
Implemented default mapper
dmytrostruk Oct 2, 2024
c4f3c66
Merge branch 'feature-vector-search' into sqlite-connector
dmytrostruk Oct 2, 2024
3b02888
Implemented Get and Upsert operations
dmytrostruk Oct 3, 2024
d8eb4b5
.Net: Adapt Azure AI Search Connector to new Text Search Design (#9061)
markwallace-microsoft Oct 3, 2024
d98e218
Implemented Delete record operations. Fixed batch operations.
dmytrostruk Oct 3, 2024
098ba75
Exposed virtual table name configuration
dmytrostruk Oct 3, 2024
e668013
Added test for string keys
dmytrostruk Oct 3, 2024
0179733
Added distance function mapping
dmytrostruk Oct 3, 2024
cc8fe45
.Net: Add response container type for vector search results. (#9082)
westey-m Oct 4, 2024
fba0aba
.Net: Use AsynEnumerable search interface for azure ai search. (#9090)
westey-m Oct 4, 2024
c198b09
Vector search and code refactoring
dmytrostruk Oct 4, 2024
d09e70f
Added offset logic
dmytrostruk Oct 5, 2024
03b89fe
Added filter usage
dmytrostruk Oct 5, 2024
69b9fc4
.Net: Update Qdrant Memory Connector to new Text Search Design (#9076)
markwallace-microsoft Oct 7, 2024
6d32d34
.Net: Add attributes to add to a model which can be converted to a Te…
markwallace-microsoft Oct 7, 2024
805febe
.Net: Change signature of the TextSearchResult constructor (#9153)
markwallace-microsoft Oct 8, 2024
add4461
Merge branch 'feature-vector-search' into sqlite-connector
dmytrostruk Oct 8, 2024
ea335bc
Merge from feature branch
dmytrostruk Oct 8, 2024
22e24a5
Merge branch 'main' into feature-vector-search
dmytrostruk Oct 8, 2024
db95435
Merge branch 'feature-vector-search' into sqlite-connector
dmytrostruk Oct 8, 2024
2d88cfb
Added support for generic data model
dmytrostruk Oct 8, 2024
bb2b457
Added vector store class
dmytrostruk Oct 9, 2024
806829d
Added extension methods
dmytrostruk Oct 9, 2024
4625afb
Small fix
dmytrostruk Oct 9, 2024
46ec215
Small fix
dmytrostruk Oct 9, 2024
b85b0e7
Fixed warnings
dmytrostruk Oct 9, 2024
a8af704
Fixed warning
dmytrostruk Oct 9, 2024
f29e110
Merge branch 'main' into sqlite-connector
dmytrostruk Oct 9, 2024
0dec618
Fix after merge
dmytrostruk Oct 9, 2024
59e9bba
Fixed usings
dmytrostruk Oct 9, 2024
21cf1ac
Added unit tests
dmytrostruk Oct 9, 2024
4687d06
More unit tests
dmytrostruk Oct 9, 2024
52c8a4c
Merge branch 'main' into sqlite-connector
dmytrostruk Oct 10, 2024
7da8523
Merge branch 'main' into sqlite-connector
dmytrostruk Oct 11, 2024
22c9491
Fixes after merge
dmytrostruk Oct 11, 2024
270e9e1
Moved Sqlite unit test project
dmytrostruk Oct 11, 2024
f7cbaf2
Addressed PR feedback
dmytrostruk Oct 11, 2024
9094eaa
Added more comments
dmytrostruk Oct 11, 2024
aa4b4b6
More improvements and fixes
dmytrostruk Oct 11, 2024
79dac79
Merge branch 'main' into sqlite-connector
dmytrostruk Oct 11, 2024
77fe3c2
Updated mapping and fixed tests
dmytrostruk Oct 14, 2024
4ecf78c
Fixed usings
dmytrostruk Oct 14, 2024
a34fc23
Added unit tests for command builder
dmytrostruk Oct 14, 2024
ecf6de9
Added unit tests for record collection class
dmytrostruk Oct 14, 2024
43900fd
Added integration test to get existing record
dmytrostruk Oct 14, 2024
7453e21
Fixed warnings
dmytrostruk Oct 14, 2024
3a797fb
Updated extension methods
dmytrostruk Oct 14, 2024
1b75ccd
Merge branch 'main' into sqlite-connector
dmytrostruk Oct 15, 2024
44e366f
Fixes based on merge from main
dmytrostruk Oct 15, 2024
cbd2902
Added more tests
dmytrostruk Oct 16, 2024
5048f9f
Small fix
dmytrostruk Oct 16, 2024
4458478
Small refactoring
dmytrostruk Oct 16, 2024
c4df0c6
Addressed PR feedback
dmytrostruk Oct 16, 2024
c55e350
Added more tests for collection class
dmytrostruk Oct 16, 2024
d8e72b3
Added unit tests for default record mapper
dmytrostruk Oct 16, 2024
53e3c5a
Added more unit tests
dmytrostruk Oct 16, 2024
9884082
Merge branch 'main' into sqlite-connector
dmytrostruk Oct 16, 2024
e10ccf7
Merge branch 'main' into sqlite-connector
dmytrostruk Oct 17, 2024
a335a61
Addressed PR feedback
dmytrostruk Oct 17, 2024
6510a66
Updated SQLite connector package version suffix
dmytrostruk Oct 17, 2024
c79b8c6
Merge branch 'main' into sqlite-connector
dmytrostruk Oct 17, 2024
5b5399e
Removed kernel builder extensions
dmytrostruk Oct 17, 2024
033ca59
Merge branch 'main' into sqlite-connector
dmytrostruk Oct 17, 2024
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
Prev Previous commit
Next Next commit
.Net: Add qdrant vector search implementation. (#8508)
### Motivation and Context

As part of the work on vector storage we have to add vector search
capabilities for each implementation.

### Description

1. Adding vector search for Qdrant.
2. Note that for now, the search interface is implemented directly by
the collection, but in future it should be part of the collection
interface. I'm doing it this way so that each implementation can be
added one by one, and once all have been implemented, we can make the
switch.

### Contribution Checklist

<!-- Before submitting this PR, please make sure: -->

- [x] The code builds clean without any errors or warnings
- [x] The PR follows the [SK Contribution
Guidelines](https://github.com/microsoft/semantic-kernel/blob/main/CONTRIBUTING.md)
and the [pre-submission formatting
script](https://github.com/microsoft/semantic-kernel/blob/main/CONTRIBUTING.md#development-scripts)
raises no violations
- [x] All unit tests pass, and I have added new tests where possible
- [x] I didn't break anyone 😄
  • Loading branch information
westey-m authored Sep 5, 2024
commit 7836721e3b03ece1f5d9b39db03388d4359907e2
2 changes: 1 addition & 1 deletion dotnet/Directory.Packages.props
Original file line number Diff line number Diff line change
Expand Up @@ -94,7 +94,7 @@
<PackageVersion Include="Milvus.Client" Version="2.3.0-preview.1" />
<PackageVersion Include="Testcontainers.Milvus" Version="3.8.0" />
<PackageVersion Include="Microsoft.Data.SqlClient" Version="5.2.1" />
<PackageVersion Include="Qdrant.Client" Version="1.9.0" />
<PackageVersion Include="Qdrant.Client" Version="1.11.0" />
<!-- Symbols -->
<PackageVersion Include="Microsoft.SourceLink.GitHub" Version="8.0.0" />
<!-- Toolset -->
Expand Down
Original file line number Diff line number Diff line change
Expand Up @@ -255,4 +255,61 @@ public virtual Task<IReadOnlyList<RetrievedPoint>> RetrieveAsync(
ShardKeySelector? shardKeySelector = null,
CancellationToken cancellationToken = default)
=> this._qdrantClient.RetrieveAsync(collectionName, ids, withPayload, withVectors, readConsistency, shardKeySelector, cancellationToken);

/// <summary>
/// Universally query points.
/// Covers all capabilities of search, recommend, discover, filters.
/// Also enables hybrid and multi-stage queries.
/// </summary>
/// <param name="collectionName">The name of the collection.</param>
/// <param name="query">Query to perform. If missing, returns points ordered by their IDs.</param>
/// <param name="prefetch">Sub-requests to perform first. If present, the query will be performed on the results of the prefetches.</param>
/// <param name="usingVector">Name of the vector to use for querying. If missing, the default vector is used..</param>
/// <param name="filter">Filter conditions - return only those points that satisfy the specified conditions.</param>
/// <param name="scoreThreshold">Return points with scores better than this threshold.</param>
/// <param name="searchParams">Search config.</param>
/// <param name="limit">Max number of results.</param>
/// <param name="offset">Offset of the result.</param>
/// <param name="payloadSelector">Options for specifying which payload to include or not.</param>
/// <param name="vectorsSelector">Options for specifying which vectors to include into the response.</param>
/// <param name="readConsistency">Options for specifying read consistency guarantees.</param>
/// <param name="shardKeySelector">Specify in which shards to look for the points, if not specified - look in all shards.</param>
/// <param name="lookupFrom">The location to use for IDs lookup, if not specified - use the current collection and the 'usingVector' vector</param>
/// <param name="timeout">If set, overrides global timeout setting for this request.</param>
/// <param name="cancellationToken">The token to monitor for cancellation requests. The default value is <see cref="CancellationToken.None" />.
/// </param>
public virtual Task<IReadOnlyList<ScoredPoint>> QueryAsync(
string collectionName,
Query? query = null,
IReadOnlyList<PrefetchQuery>? prefetch = null,
string? usingVector = null,
Filter? filter = null,
float? scoreThreshold = null,
SearchParams? searchParams = null,
ulong limit = 10,
ulong offset = 0,
WithPayloadSelector? payloadSelector = null,
WithVectorsSelector? vectorsSelector = null,
ReadConsistency? readConsistency = null,
ShardKeySelector? shardKeySelector = null,
LookupLocation? lookupFrom = null,
TimeSpan? timeout = null,
CancellationToken cancellationToken = default)
=> this._qdrantClient.QueryAsync(
collectionName,
query,
prefetch,
usingVector,
filter,
scoreThreshold,
searchParams,
limit,
offset,
payloadSelector,
vectorsSelector,
readConsistency,
shardKeySelector,
lookupFrom,
timeout,
cancellationToken);
}
Original file line number Diff line number Diff line change
@@ -0,0 +1,112 @@
// Copyright (c) Microsoft. All rights reserved.

using System;
using System.Collections.Generic;
using Microsoft.SemanticKernel.Data;
using Qdrant.Client.Grpc;

namespace Microsoft.SemanticKernel.Connectors.Qdrant;

/// <summary>
/// Contains mapping helpers to use when searching for documents using Qdrant.
/// </summary>
internal static class QdrantVectorStoreCollectionSearchMapping
{
/// <summary>
/// Build a Qdrant <see cref="Filter"/> from the provided <see cref="VectorSearchFilter"/>.
/// </summary>
/// <param name="basicVectorSearchFilter">The <see cref="VectorSearchFilter"/> to build a Qdrant <see cref="Filter"/> from.</param>
/// <param name="storagePropertyNames">A mapping of data model property names to the names under which they are stored.</param>
/// <returns>The Qdrant <see cref="Filter"/>.</returns>
/// <exception cref="InvalidOperationException">Thrown when the provided filter contains unsupported types, values or unknown properties.</exception>
public static Filter BuildFilter(VectorSearchFilter? basicVectorSearchFilter, Dictionary<string, string> storagePropertyNames)
{
var filter = new Filter();

// Return an empty filter if no filter clauses are provided.
if (basicVectorSearchFilter?.FilterClauses is null)
{
return filter;
}

foreach (var filterClause in basicVectorSearchFilter.FilterClauses)
{
string fieldName;
object filterValue;

// In Qdrant, tag list contains is handled using a keyword match, which is the same as a string equality check.
// We can therefore just extract the field name and value from each clause and handle them the same.
if (filterClause is EqualityFilterClause equalityFilterClause)
{
fieldName = equalityFilterClause.FieldName;
filterValue = equalityFilterClause.Value;
}
else if (filterClause is TagListContainsFilterClause tagListContainsClause)
{
fieldName = tagListContainsClause.FieldName;
filterValue = tagListContainsClause.Value;
}
else
{
throw new InvalidOperationException($"Unsupported filter clause type '{filterClause.GetType().Name}'.");
}

// Map each type of filter value to the appropriate Qdrant match type.
var match = filterValue switch
{
string stringValue => new Match { Keyword = stringValue },
int intValue => new Match { Integer = intValue },
long longValue => new Match { Integer = longValue },
bool boolValue => new Match { Boolean = boolValue },
_ => throw new InvalidOperationException($"Unsupported filter value type '{filterValue.GetType().Name}'.")
};

// Get the storage name for the field.
if (!storagePropertyNames.TryGetValue(fieldName, out var storagePropertyName))
{
throw new InvalidOperationException($"Property name '{fieldName}' provided as part of the filter clause is not a valid property name.");
}

filter.Must.Add(new Condition() { Field = new FieldCondition() { Key = storagePropertyName, Match = match } });
}

return filter;
}

/// <summary>
/// Map the given <see cref="ScoredPoint"/> to a <see cref="VectorSearchResult{TRecord}"/>.
/// </summary>
/// <typeparam name="TRecord">The type of the record to map to.</typeparam>
/// <param name="point">The point to map to a <see cref="VectorSearchResult{TRecord}"/>.</param>
/// <param name="mapper">The mapper to perform the main mapping operation with.</param>
/// <param name="includeVectors">A value indicating whether to include vectors in the mapped result.</param>
/// <param name="databaseSystemName">The name of the database system the operation is being run on.</param>
/// <param name="collectionName">The name of the collection the operation is being run on.</param>
/// <param name="operationName">The type of database operation being run.</param>
/// <returns>The mapped <see cref="VectorSearchResult{TRecord}"/>.</returns>
public static VectorSearchResult<TRecord> MapScoredPointToVectorSearchResult<TRecord>(ScoredPoint point, IVectorStoreRecordMapper<TRecord, PointStruct> mapper, bool includeVectors, string databaseSystemName, string collectionName, string operationName)
where TRecord : class
{
// Since the mapper doesn't know about scored points, we need to convert the scored point to a point struct first.
var pointStruct = new PointStruct
{
Id = point.Id,
Vectors = point.Vectors,
Payload = { }
};

foreach (KeyValuePair<string, Value> payloadEntry in point.Payload)
{
pointStruct.Payload.Add(payloadEntry.Key, payloadEntry.Value);
}

// Do the mapping with error handling.
return new VectorSearchResult<TRecord>(
VectorStoreErrorHandler.RunModelConversion(
databaseSystemName,
collectionName,
operationName,
() => mapper.MapFromStorageToDataModel(pointStruct, new() { IncludeVectors = includeVectors })),
point.Score);
}
}
Original file line number Diff line number Diff line change
Expand Up @@ -18,7 +18,7 @@ namespace Microsoft.SemanticKernel.Connectors.Qdrant;
/// </summary>
/// <typeparam name="TRecord">The data model to use for adding, updating and retrieving data from storage.</typeparam>
#pragma warning disable CA1711 // Identifiers should not have incorrect suffix
public sealed class QdrantVectorStoreRecordCollection<TRecord> : IVectorStoreRecordCollection<ulong, TRecord>, IVectorStoreRecordCollection<Guid, TRecord>
public sealed class QdrantVectorStoreRecordCollection<TRecord> : IVectorStoreRecordCollection<ulong, TRecord>, IVectorStoreRecordCollection<Guid, TRecord>, IVectorSearch<TRecord>
#pragma warning restore CA1711 // Identifiers should not have incorrect suffix
where TRecord : class
{
Expand Down Expand Up @@ -56,6 +56,9 @@ public sealed class QdrantVectorStoreRecordCollection<TRecord> : IVectorStoreRec
/// <summary>A dictionary that maps from a property name to the configured name that should be used when storing it.</summary>
private readonly Dictionary<string, string> _storagePropertyNames = new();

/// <summary>The name of the first vector field for the collections that this class is used with.</summary>
private readonly string? _firstVectorPropertyName = null;

/// <summary>
/// Initializes a new instance of the <see cref="QdrantVectorStoreRecordCollection{TRecord}"/> class.
/// </summary>
Expand Down Expand Up @@ -95,6 +98,10 @@ internal QdrantVectorStoreRecordCollection(MockableQdrantClient qdrantClient, st

// Build a map of property names to storage names.
this._storagePropertyNames = VectorStoreRecordPropertyReader.BuildPropertyNameToStorageNameMap(properties);
if (properties.VectorProperties.Count > 0)
{
this._firstVectorPropertyName = this._storagePropertyNames[properties.VectorProperties.First().DataModelPropertyName];
}

// Assign Mapper.
if (this._options.PointStructCustomMapper is not null)
Expand Down Expand Up @@ -432,6 +439,93 @@ private async IAsyncEnumerable<TRecord> GetBatchByPointIdAsync<TKey>(
}
}

/// <inheritdoc />
public async IAsyncEnumerable<VectorSearchResult<TRecord>> SearchAsync(VectorSearchQuery vectorQuery, [EnumeratorCancellation] CancellationToken cancellationToken = default)
{
if (this._firstVectorPropertyName is null)
{
throw new InvalidOperationException("The collection does not have any vector fields, so vector search is not possible.");
}

if (vectorQuery is VectorizedSearchQuery<ReadOnlyMemory<float>> floatVectorQuery)
{
var internalOptions = floatVectorQuery.SearchOptions ?? Data.VectorSearchOptions.Default;

// Build filter object.
var filter = QdrantVectorStoreCollectionSearchMapping.BuildFilter(internalOptions.VectorSearchFilter, this._storagePropertyNames);

// Specify the vector name if named vectors are used.
string? vectorName = null;
if (this._options.HasNamedVectors)
{
vectorName = this.ResolveVectorFieldName(internalOptions.VectorFieldName);
}

// Specify whether to include vectors in the search results.
var vectorsSelector = new WithVectorsSelector();
vectorsSelector.Enable = internalOptions.IncludeVectors;

var query = new Query
{
Nearest = new VectorInput(floatVectorQuery.Vector.ToArray()),
};

// Execute Search.
var points = await this.RunOperationAsync(
"Query",
() => this._qdrantClient.QueryAsync(
this.CollectionName,
query: query,
usingVector: vectorName,
filter: filter,
limit: (ulong)internalOptions.Limit,
offset: (ulong)internalOptions.Offset,
vectorsSelector: vectorsSelector,
cancellationToken: cancellationToken)).ConfigureAwait(false);

// Map to data model and return results.
foreach (var point in points)
{
yield return QdrantVectorStoreCollectionSearchMapping.MapScoredPointToVectorSearchResult(
point,
this._mapper,
internalOptions.IncludeVectors,
DatabaseName,
this._collectionName,
"Query");
}

yield break;
}

throw new NotSupportedException($"A {nameof(VectorSearchQuery)} of type {vectorQuery.QueryType} is not supported by the Qdrant connector.");
}

/// <summary>
/// Resolve the vector field name to use for a search by using the storage name for the field name from options
/// if available, and falling back to the first vector field name if not.
/// </summary>
/// <param name="optionsVectorFieldName">The vector field name provided via options.</param>
/// <returns>The resolved vector field name.</returns>
/// <exception cref="InvalidOperationException">Thrown if the provided field name is not a valid field name.</exception>
private string ResolveVectorFieldName(string? optionsVectorFieldName)
{
string? vectorFieldName;
if (optionsVectorFieldName is not null)
{
if (!this._storagePropertyNames.TryGetValue(optionsVectorFieldName, out vectorFieldName))
{
throw new InvalidOperationException($"The collection does not have a vector field named '{optionsVectorFieldName}'.");
}
}
else
{
vectorFieldName = this._firstVectorPropertyName;
}

return vectorFieldName!;
}

/// <summary>
/// Run the given operation and wrap any <see cref="RpcException"/> with <see cref="VectorStoreOperationException"/>."/>
/// </summary>
Expand Down
Loading
Loading