Skip to content
Tracon

TraconKnowledgeOptions

Namespace Tracon · Assembly Tracon.Core.dll

Options for the knowledge base (semantic search).

public sealed class TraconKnowledgeOptions
Inheritance
object ← TraconKnowledgeOptions
Inherited Members
object.GetType(), object.ToString(), object.Equals(object?), object.Equals(object?, object?), object.ReferenceEquals(object?, object?), object.GetHashCode()

Reads values from the Tracon:Knowledge configuration section. It does not require a separate Use... call, for the same reason as scheduling and quotas. It becomes functional only when IVectorSearchStore and IEmbeddingGenerator are registered.

public TraconKnowledgeOptions()

The configuration section name.

public const string SectionName = "Tracon:Knowledge"

string

The overlap length between consecutive chunks, in characters.

public int ChunkOverlap { get; set; }

int

The chunk length in characters.

public int ChunkSize { get; set; }

int

The embedding dimensions. They are fixed with the schema. Changing them invalidates all embeddings.

public int Dimensions { get; set; }

int

The name of the provider the registered embedding generator calls, as a tenant egress policy names it (for example azure-openai).

public string? EmbeddingProvider { get; set; }

string?

Every embedding call — a search_knowledge query, a knowledge search and a document upload — checks the tenant’s egress policy against this name before the text leaves the process, and refuses a tenant that has its own credential for this provider: the generator was built with the setup key and cannot use the tenant’s.

When it is not set, the provider name the generator reports is used. That name does not always match a policy name: the OpenAI and the Azure OpenAI generators both report openai. When neither name is known, a tenant with an egress policy is refused, and a tenant without one is not.

The maximum number of results returned by the search_knowledge tool.

public int MaxResults { get; set; }

int