Vector Retriever

What's a Vector Retriever?

Vector Retriever is the most commonly used retriever type for document-based applications. Vector Retriever retrieves documents and information from vector databases using semantic similarity search.

Best For:

• Document search and retrieval

• Semantic similarity matching

• Large-scale text corpora

• Unstructured data search

Key Features:

• Embedding-based similarity search

• Support for multiple vector databases (Chroma, Elasticsearch, Redis)

• Metadata filtering and scoring

• Configurable similarity thresholds

Use Cases:

• Document Q&A systems

• Content recommendation engines

• Semantic search applications

• Knowledge base retrieval

Prerequisites

This example specifically requires completion of all setup steps listed on the Prerequisites page.

You should be familiar with these concepts:

1. Vector Data Store

2. Embedding Model (EM) Invoker

Installation

Linux, macOS, or Windows WSL

# you can use a Conda environment
pip install --extra-index-url https://oauth2accesstoken:$(gcloud auth print-access-token)@glsdk.gdplabs.id/gen-ai-internal/simple/ "gllm-retrieval"

Windows Powershell

Windows Command Prompt

What it does

The Vector Retriever is a component that retrieves relevant documents from a vector database based on semantic similarity to a query. It provides a standardized interface for document retrieval operations in Gen AI applications.

Inputs

• Query: A text string representing the search query

• Data Store: A vector data store instance (e.g., Chroma, Elasticsearch, Redis)

• Top-k: Maximum number of documents to retrieve (optional, defaults to system default)

• Retrieval Parameters: Additional parameters for fine-tuning the search (optional)

Outputs

• List of Chunks: A list of Chunk objects containing the retrieved documents with their metadata

Save and Retrieve Data

import asyncio
from gllm_datastore.vector_data_store.chroma_vector_data_store import ChromaVectorDataStore
from gllm_retrieval.retriever.vector_retriever import BasicVectorRetriever

# Initialize Chroma data store in-memory
vector_store = ChromaVectorDataStore(
    collection_name="documents", 
    embedding=OpenAIEMInvoker(model_name="text-embedding-3-small")
)

# Initialize the vector retriever
retriever = BasicVectorRetriever(vector_store)
    
# Perform a basic retrieval
query = "What is machine learning?"
results = await retriever.retrieve(query)

Instead of relying solely on a string for semantic queries, we can also apply metadata filtering through the retrieval_params parameter following the retrieval params provided by that specific data store. For example, if we are using ChromaVectorDataStore, the retrieval parameter can be used as follows:

retrieval_params = {
    "filter": {
        "$and": [
            {"type": "document"},
        ]
    },
    "where_document": {"$contains": {"text": "AI"}},
}
query = "What is machine learning?"
results = await retriever.retrieve(query, retrieval_params=retrieval_params)