add knowledge data

This commit is contained in:
Jicheng Lu 2024-08-06 17:43:11 -05:00
parent 93646a01ae
commit 6ab7c02545
12 changed files with 146 additions and 5 deletions

View file

@ -14,5 +14,6 @@ public interface IKnowledgeService
#region List
Task<KnowledgeCollectionInfo> GetKnowledgeCollectionInfo(string collectionName);
Task<UuidPagedItems<KnowledgeCollectionData>> GetKnowledgeCollectionData(KnowledgeFilter filter);
#endregion
}

View file

@ -0,0 +1,9 @@
namespace BotSharp.Abstraction.Knowledges.Models;
public class KnowledgeCollectionData
{
public string Id { get; set; }
public string Text { get; set; }
public string Answer { get; set; }
public float[]? Vector { get; set; }
}

View file

@ -0,0 +1,10 @@
namespace BotSharp.Abstraction.Knowledges.Models;
public class KnowledgeFilter : UuidPagination
{
[JsonPropertyName("collection_name")]
public string CollectionName { get; set; }
[JsonPropertyName("with_vector")]
public bool WithVector { get; set; }
}

View file

@ -0,0 +1,15 @@
namespace BotSharp.Abstraction.Utilities;
public class UuidPagination : Pagination
{
[JsonPropertyName("start_id")]
public string? StartId { get; set; }
}
public class UuidPagedItems<T> : PagedItems<T>
{
public new ulong Count { get; set; }
[JsonPropertyName("next_id")]
public string? NextId { get; set; }
}

View file

@ -6,6 +6,7 @@ public interface IVectorDb
{
Task<List<string>> GetCollections();
Task<KnowledgeCollectionInfo> GetCollectionInfo(string collectionName);
Task<UuidPagedItems<KnowledgeCollectionData>> GetCollectionData(KnowledgeFilter filter);
Task CreateCollection(string collectionName, int dim);
Task<bool> Upsert(string collectionName, string id, float[] vector, string text, Dictionary<string, string>? payload = null);
Task<List<string>> Search(string collectionName, float[] vector, string returnFieldName, int limit = 5, float confidence = 0.5f);

View file

@ -92,11 +92,25 @@ public class KnowledgeBaseController : ControllerBase
return Ok(new { count = files.Count, size });
}
[HttpGet("/knowledge/info")]
[HttpGet("/knowledge/collection/info")]
public async Task<KnowledgeCollectionInfoViewModel> GetKnowledgeCollectionInfo([FromQuery] string collectionName)
{
var info = await _knowledgeService.GetKnowledgeCollectionInfo(collectionName);
return KnowledgeCollectionInfoViewModel.ToViewModel(info);
}
[HttpPost("/knowledge/collection/data")]
public async Task<UuidPagedItems<KnowledgeCollectionDataViewModel>> GetKnowledgeCollectionData([FromBody] KnowledgeFilter filter)
{
var data = await _knowledgeService.GetKnowledgeCollectionData(filter);
var items = data.Items?.Select(x => KnowledgeCollectionDataViewModel.ToViewModel(x))?
.ToList() ?? new List<KnowledgeCollectionDataViewModel>();
return new UuidPagedItems<KnowledgeCollectionDataViewModel>
{
Count = data.Count,
NextId = data.NextId,
Items = items
};
}
}

View file

@ -0,0 +1,31 @@
using BotSharp.Abstraction.Knowledges.Models;
using System.Text.Json.Serialization;
namespace BotSharp.OpenAPI.ViewModels.Knowledges;
public class KnowledgeCollectionDataViewModel
{
[JsonPropertyName("id")]
public string Id { get; set; }
[JsonPropertyName("text")]
public string Text { get; set; }
[JsonPropertyName("answer")]
public string Answer { get; set; }
[JsonPropertyName("vector")]
[JsonIgnore(Condition = JsonIgnoreCondition.WhenWritingNull)]
public float[]? Vector { get; set; }
public static KnowledgeCollectionDataViewModel ToViewModel(KnowledgeCollectionData data)
{
return new KnowledgeCollectionDataViewModel
{
Id = data.Id,
Text = data.Text,
Answer = data.Answer,
Vector = data.Vector
};
}
}

View file

@ -32,6 +32,11 @@ public class MemVectorDatabase : IVectorDb
};
}
public Task<UuidPagedItems<KnowledgeCollectionData>> GetCollectionData(KnowledgeFilter filter)
{
throw new NotImplementedException();
}
public async Task<List<string>> Search(string collectionName, float[] vector, string returnFieldName, int limit = 5, float confidence = 0.5f)
{
if (!_vectors.ContainsKey(collectionName))

View file

@ -15,4 +15,18 @@ public partial class KnowledgeService
return new KnowledgeCollectionInfo();
}
}
public async Task<UuidPagedItems<KnowledgeCollectionData>> GetKnowledgeCollectionData(KnowledgeFilter filter)
{
try
{
var db = GetVectorDb();
return await db.GetCollectionData(filter);
}
catch (Exception ex)
{
_logger.LogWarning($"Error when getting knowledge collectio data. {ex.Message}\r\n{ex.InnerException}");
return new UuidPagedItems<KnowledgeCollectionData>();
}
}
}

View file

@ -1,4 +1,5 @@
using BotSharp.Abstraction.Knowledges.Models;
using BotSharp.Abstraction.Utilities;
using BotSharp.Abstraction.VectorStorage;
using System;
using System.Collections.Generic;
@ -18,6 +19,11 @@ public class FaissDb : IVectorDb
throw new NotImplementedException();
}
public Task<UuidPagedItems<KnowledgeCollectionData>> GetCollectionData(KnowledgeFilter filter)
{
throw new NotImplementedException();
}
public Task<List<string>> GetCollections()
{
throw new NotImplementedException();

View file

@ -1,4 +1,4 @@
using BotSharp.Abstraction.Knowledges.Models;
using BotSharp.Abstraction.Utilities;
using Qdrant.Client;
using Qdrant.Client.Grpc;
@ -41,7 +41,12 @@ public class QdrantDb : IVectorDb
public async Task<KnowledgeCollectionInfo> GetCollectionInfo(string collectionName)
{
var info = await GetClient().GetCollectionInfoAsync(collectionName);
var client = GetClient();
var exists = await client.CollectionExistsAsync(collectionName);
if (!exists) return new KnowledgeCollectionInfo();
var info = await client.GetCollectionInfoAsync(collectionName);
return new KnowledgeCollectionInfo
{
DataCount = info.PointsCount,
@ -49,6 +54,35 @@ public class QdrantDb : IVectorDb
};
}
public async Task<UuidPagedItems<KnowledgeCollectionData>> GetCollectionData(KnowledgeFilter filter)
{
var client = GetClient();
var exists = await client.CollectionExistsAsync(filter.CollectionName);
if (!exists)
{
return new UuidPagedItems<KnowledgeCollectionData>();
}
var totalPointCount = await client.CountAsync(filter.CollectionName);
var response = await client.ScrollAsync(filter.CollectionName, limit: (uint)filter.Size,
offset: !string.IsNullOrWhiteSpace(filter.StartId) ? new PointId { Uuid = filter.StartId } : 0,
vectorsSelector: filter.WithVector);
var points = response?.Result?.Select(x => new KnowledgeCollectionData
{
Id = x.Id?.Uuid ?? string.Empty,
Text = x.Payload.ContainsKey(KnowledgePayloadName.Text) ? x.Payload[KnowledgePayloadName.Text].StringValue : string.Empty,
Answer = x.Payload.ContainsKey(KnowledgePayloadName.Answer) ? x.Payload[KnowledgePayloadName.Answer].StringValue : string.Empty,
Vector = filter.WithVector ? x.Vectors?.Vector?.Data?.ToArray() : null
})?.ToList() ?? new List<KnowledgeCollectionData>();
return new UuidPagedItems<KnowledgeCollectionData>
{
Count = totalPointCount,
NextId = response?.NextPageOffset?.Uuid,
Items = points
};
}
public async Task CreateCollection(string collectionName, int dim)
{
var collections = await GetCollections();
@ -81,7 +115,7 @@ public class QdrantDb : IVectorDb
},
Vectors = vector,
Payload =
Payload =
{
{ KnowledgePayloadName.Text, text }
}

View file

@ -3,4 +3,5 @@ global using System.Collections.Generic;
global using System.Linq;
global using System.Threading.Tasks;
global using BotSharp.Abstraction.VectorStorage;
global using BotSharp.Abstraction.Knowledges.Enums;
global using BotSharp.Abstraction.Knowledges.Enums;
global using BotSharp.Abstraction.Knowledges.Models;