Class EmbeddingService

Service for generating embeddings using AI Engine

Index

Constructors

Methods

embedGoldenQueries embedQueries embedQuery

Constructors

constructor

new EmbeddingService(modelName?: string): EmbeddingService
Create an EmbeddingService with the specified embedding model
Parameters
- modelName: string = 'text-embedding-3-small'
  Name of the embedding model to use (default: 'text-embedding-3-small')
Returns EmbeddingService
- Defined in QueryGen/src/vectors/EmbeddingService.ts:22

Methods

embedGoldenQueries

embedGoldenQueries(goldenQueries: GoldenQuery[]): Promise<EmbeddedGoldenQuery[]>
Embed all golden queries for few-shot learning

Golden queries are example queries that serve as few-shot learning examples. This method embeds all fields of all golden queries for similarity search. Note: name field is excluded from embeddings as it's not available during query generation.
Parameters
- goldenQueries: GoldenQuery[]
  Array of golden queries to embed
Returns Promise<EmbeddedGoldenQuery[]>
Array of golden queries with their embeddings
- Defined in QueryGen/src/vectors/EmbeddingService.ts:68

embedQueries

embedQueries(
    queries: {
        description: string;
        technicalDescription: string;
        userQuestion: string;
    }[],
): Promise<QueryEmbeddings[]>
Embed multiple queries in batch
Parameters
- queries: { description: string; technicalDescription: string; userQuestion: string }[]
  Array of queries to embed
Returns Promise<QueryEmbeddings[]>
Array of embeddings corresponding to input queries
- Defined in QueryGen/src/vectors/EmbeddingService.ts:93

embedQuery

embedQuery(
    query: {
        description: string;
        technicalDescription: string;
        userQuestion: string;
    },
): Promise<QueryEmbeddings>
Embed a single query by generating embeddings for all its fields

Generates separate embeddings for:
- userQuestion: Natural language question
- description: High-level description
- technicalDescription: Technical implementation details
Parameters
- query: { description: string; technicalDescription: string; userQuestion: string }
  Query to embed (partial structure with text fields)
Returns Promise<QueryEmbeddings>
Embeddings for all fields
- Defined in QueryGen/src/vectors/EmbeddingService.ts:37