Class SemanticSimilarityExampleSelector<V>

Class that selects examples based on semantic similarity. It extends the BaseExampleSelector class.

Example

const exampleSelector = await SemanticSimilarityExampleSelector.fromExamples(
  [
    { input: "happy", output: "sad" },
    { input: "tall", output: "short" },
    { input: "energetic", output: "lethargic" },
    { input: "sunny", output: "gloomy" },
    { input: "windy", output: "calm" },
  ],
  new OpenAIEmbeddings(),
  HNSWLib,
  { k: 1 },
);
const dynamicPrompt = new FewShotPromptTemplate({
  exampleSelector,
  examplePrompt: PromptTemplate.fromTemplate(
    "Input: {input}\nOutput: {output}",
  ),
  prefix: "Give the antonym of every input",
  suffix: "Input: {adjective}\nOutput:",
  inputVariables: ["adjective"],
});
console.log(await dynamicPrompt.format({ adjective: "rainy" }));

Type Parameters

Hierarchy (view full)

Constructors

constructor

Properties

exampleKeys? inputKeys? vectorStoreRetriever

Methods

addExample selectExamples toJSON toJSONNotImplemented fromExamples

Constructors

constructor

Properties

OptionalexampleKeys

exampleKeys?: string[]

OptionalinputKeys

inputKeys?: string[]

vectorStoreRetriever

vectorStoreRetriever: VectorStoreRetrieverInterface<V>

Methods

addExample

selectExamples

  • selectExamples<T>(inputVariables): Promise<Example[]>

  • Method that selects which examples to use based on semantic similarity. It performs a similarity search in the vectorStore using the input variables and returns the examples with the highest similarity.

    Type Parameters

    • T

    Parameters

    • inputVariables: Record<string, T>

      The input variables used for the similarity search.

    Returns Promise<Example[]>

    Promise that resolves with an array of the selected examples.

toJSON

toJSONNotImplemented

StaticfromExamples

  • fromExamples<C>(examples, embeddings, vectorStoreCls, options?): Promise<SemanticSimilarityExampleSelector<VectorStoreInterface>>

  • Static method that creates a new instance of SemanticSimilarityExampleSelector. It takes a list of examples, an instance of Embeddings, a VectorStore class, and an options object as parameters. It converts the examples to strings, creates a VectorStore from the strings and the embeddings, and returns a new SemanticSimilarityExampleSelector with the created VectorStore and the options provided.

    Type Parameters

    Parameters

    • examples: Record<string, string>[]

      The list of examples to be used.

    • embeddings: Embeddings

      The instance of Embeddings to be used.

    • vectorStoreCls: C

      The VectorStore class to be used.

    • options: {
          inputKeys?: string[];
          k?: number;
      } & Parameters<C["fromTexts"]>[3] = {}

      The options object for the SemanticSimilarityExampleSelector.

    Returns Promise<SemanticSimilarityExampleSelector<VectorStoreInterface>>

    Promise that resolves with a new instance of SemanticSimilarityExampleSelector.

Settings

Member Visibility

On This Page

Constructors
Properties
Methods