Class StandardRAGPipeline

Standard RAG pipeline implementation.

Orchestrates the complete RAG workflow:

  1. Ingest: Chunk documents and store with embeddings
  2. Retrieve: Find relevant documents using vector search
  3. Generate: Create responses augmented with retrieved context

Example

const pipeline = new StandardRAGPipeline({
  llmProvider: openAIProvider,
  vectorStore: new InMemoryVectorStore(openAIProvider, 'text-embedding-3-small'),
  chunker: new RecursiveTextChunker({ chunkSize: 1000, chunkOverlap: 200 }),
  embeddingModel: 'text-embedding-3-small',
  generationModel: 'gpt-4-turbo',
  defaultSearchLimit: 5,
  defaultMinScore: 0.7
});

// Ingest documents
await pipeline.ingest([
  { id: '1', content: 'Product documentation...' },
  { id: '2', content: 'FAQ content...' }
], { chunk: true });

// Query with RAG
const result = await pipeline.query('How do I reset my password?');
console.log(result.response);

Implements

  • RAGPipeline

Constructors

constructor

Methods

ingest retrieve query clear getStatistics

Properties

config

Constructors

constructor

Methods

ingest

  • ingest(documents: Document[], options?: IngestOptions): Promise<IngestResult>

    Ingest documents into the pipeline

    Parameters

    • documents: Document[]

      Documents to ingest

    • options: IngestOptions = {}

      Ingestion options

    Returns Promise<IngestResult>

    Ingestion result with statistics

retrieve

  • retrieve(
        query: string,
        options?: RetrievalOptions,
    ): Promise<VectorSearchResult[]>

    Retrieve relevant documents for a query

    Parameters

    • query: string

      Search query

    • options: RetrievalOptions = {}

      Retrieval options

    Returns Promise<VectorSearchResult[]>

    Matching documents with scores

query

  • query(
        query: string,
        options?: RetrievalOptions,
        systemPrompt?: string,
    ): Promise<RAGResult>

    Query with retrieval-augmented generation

    Parameters

    • query: string

      User query

    • options: RetrievalOptions = {}

      Retrieval options

    • OptionalsystemPrompt: string

      Optional system prompt for generation

    Returns Promise<RAGResult>

    Generated response with context and token usage

clear

getStatistics

  • getStatistics(): Promise<RAGPipelineStatistics>

    Get statistics about the pipeline

    Returns Promise<RAGPipelineStatistics>

    Pipeline statistics

Properties

Readonlyconfig

config: RAGPipelineConfig

Configuration for this pipeline

Settings

Member Visibility

On This Page

Constructors
Methods
Properties