Skip to main content

Retrievers

info

Conceptual Guide

The concept of a "retriever" within a language or framework, particularly in blockchain contexts, refers to a mechanism designed to extract or fetch data from a designated source. In the realm of blockchain, this could involve retrieving transaction details, block information, or the states of smart contracts from the blockchain's ledger.

Reasons for Using a Retriever:​

  • Data Accessibility: Provides a gateway for accessing data stored on the blockchain, crucial for applications needing to present this information to users or leverage it for further processing.

  • Efficiency: Optimizes the process of fetching data, reducing latency and enhancing the performance of blockchain applications.

  • Abstraction: Simplifies querying the blockchain by hiding its underlying complexity, offering developers a more straightforward API.

  • Integration: Enables the seamless incorporation of blockchain data into other applications or services, broadening potential use cases and functionalities.

  • Security: Allows applications to access blockchain data safely without direct ledger interactions, minimizing exposure to security risks.

How To​

The implementation of a retriever varies depending on the blockchain platform and the specific data requirements. However, the general process involves the following steps:

You need use a embedder, can you ollama, huggingface ..

    llm, err := ollama.New(ollama.WithModel("llama2"))

    if err != nil {
        log.Fatal(err)
    }

    embedder, err := embeddings.NewEmbedder(llm)
    if err != nil {
        log.Fatal(err)
    }

After it chose a storage vector like pinecone, postgres, Qdrant, in example I'll use qdrant

    url, err := url.Parse("http://localhost:6333")
    if err != nil {
        log.Fatal(err)
    }

    store, err := qdrant.New(
        qdrant.WithURL(*url),
        qdrant.WithCollectionName("youtube_transcript"),
        qdrant.WithEmbedder(embedder),
    )
    if err != nil {
        log.Fatal(err)
    }


Now Create a retriever

    searchQuery := "how to make a cake"
    optionsVector := []vectorstores.Option{
        vectorstores.WithScoreThreshold(0.80), // use for precision, when you want to get only the most relevant documents
        //vectorstores.WithNameSpace(""),            // use for set a namespace in the storage
        //vectorstores.WithFilters(map[string]interface{}{"language": "en"}), // use for filter the documents
        //vectorstores.WithEmbedder(embedder), // use when you want add documents or doing similarity search
        //vectorstores.WithDeduplicater(vectorstores.NewSimpleDeduplicater()), //  This is useful to prevent wasting time on creating an embedding
    }

    retriever := vectorstores.ToRetriever(store, 10, optionsVector...)
    // search
    resDocs, err := retriever.GetRelevantDocuments(context.Background(), searchQuery)

    if err != nil {
        log.Fatal(err)
    }

This is a simple example of how to use a retriever, you can use it in a lot of ways, like a chatbot, a search engine, a recommendation system, etc.