Added a common retrieval context with text and image implementations

This commit is contained in:
Thorsten Sommer 2025-02-10 16:12:09 +01:00
parent 53dd6a6144
commit 55f55dab86
Signed by: tsommer
GPG Key ID: 371BBA77A02C0108
3 changed files with 116 additions and 0 deletions

View File

@ -0,0 +1,45 @@
namespace AIStudio.Tools.RAG;
/// <summary>
/// The common interface for any retrieval context.
/// </summary>
public interface IRetrievalContext
{
/// <summary>
/// The name of the data source.
/// </summary>
/// <remarks>
/// Depending on the configuration, the AI is selecting the appropriate data source.
/// In order to inform the user about where the information is coming from, the data
/// source name is necessary.
/// </remarks>
public string DataSourceName { get; init; }
/// <summary>
/// The category of the content, like e.g., text, audio, image, etc.
/// </summary>
public RetrievalContentCategory Category { get; init; }
/// <summary>
/// What type of content is being retrieved? Like e.g., a project proposal, spreadsheet, art, etc.
/// </summary>
public RetrievalContentType Type { get; init; }
/// <summary>
/// The path to the content, e.g., a URL, a file path, a path in a graph database, etc.
/// </summary>
public string Path { get; init; }
/// <summary>
/// Links to related content, e.g., links to Wikipedia articles, links to sources, etc.
/// </summary>
/// <remarks>
/// Why would you need links for retrieval? You are right that not all retrieval
/// contexts need links. But think about a web search feature, where we want to
/// query a search engine and get back a list of links to the most relevant
/// matches. Think about a continuous web crawler that is constantly looking for
/// new information and adding it to the knowledge base. In these cases, links
/// are essential.
/// </remarks>
public IReadOnlyList<string> Links { get; init; }
}

View File

@ -0,0 +1,33 @@
using AIStudio.Chat;
namespace AIStudio.Tools.RAG;
public sealed class RetrievalImageContext : IRetrievalContext
{
#region Implementation of IRetrievalContext
public required string DataSourceName { get; init; }
public required RetrievalContentCategory Category { get; init; }
public required RetrievalContentType Type { get; init; }
public required string Path { get; init; }
public IReadOnlyList<string> Links { get; init; } = [];
#endregion
/// <summary>
/// The type of the image source.
/// </summary>
/// <remarks>
/// Is the image source a URL, a local file path, a base64 string, etc.?
/// </remarks>
public required ContentImageSource SourceType { get; init; }
/// <summary>
/// The image source.
/// </summary>
public required string Source { get; set; }
}

View File

@ -0,0 +1,38 @@
namespace AIStudio.Tools.RAG;
/// <summary>
/// The retrieval context for text data.
/// </summary>
public sealed class RetrievalTextContext : IRetrievalContext
{
#region Implementation of IRetrievalContext
public required string DataSourceName { get; init; }
public required RetrievalContentCategory Category { get; init; }
public required RetrievalContentType Type { get; init; }
public required string Path { get; init; }
public IReadOnlyList<string> Links { get; init; } = [];
#endregion
/// <summary>
/// The text content.
/// </summary>
/// <remarks>
/// Should contain the matched text and some small context around it.
/// </remarks>
public required string MatchedText { get; set; }
/// <summary>
/// The surrounding content of the matched text.
/// </summary>
/// <remarks>
/// Might give the user some context about the matched text.
/// For example, one sentence or paragraph before and after the matched text.
/// </remarks>
public IReadOnlyList<string> SurroundingContent { get; set; } = [];
}