Vector Retriever

What's a Vector Retriever?

Vector Retriever is the most commonly used retriever type for document-based applications. It retrieves documents from a data store with vector capability 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 data stores with vector capability (Chroma, Elasticsearch, Redis, etc.)

• Metadata filtering and scoring

• Configurable similarity thresholds and batch queries

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. Data Store and the vector capability

2. EM Invoker for embeddings

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 retrieves relevant documents from a data store with vector capability based on semantic similarity to a query. It provides a standardized interface for document retrieval operations in Gen AI applications.

Usage

Use VectorRetriever with a data store that has vector capability registered:

from gllm_datastore.data_store import ChromaDataStore
from gllm_datastore.data_store.chroma.data_store import ChromaClientType
from gllm_inference.em_invoker.openai_em_invoker import OpenAIEMInvoker
from gllm_retrieval.retriever import VectorRetriever

# Data store with vector capability
em_invoker = OpenAIEMInvoker(model_name="text-embedding-3-small")
data_store = ChromaDataStore(
    collection_name="documents",
    client_type=ChromaClientType.MEMORY,
).with_vector(em_invoker=em_invoker)

retriever = VectorRetriever(data_store=data_store)

# Single query
query = "What is machine learning?"
results = await retriever.retrieve(query, top_k=10)

# Single query with filters and threshold
from gllm_datastore.core.filters import filter as F
results = await retriever.retrieve(
    "What is machine learning?",
    query_filter=F.eq("metadata.category", "AI"),
    top_k=10,
    threshold=0.8
)

# Batch queries
batch_results = await retriever.retrieve(["query 1", "query 2"], top_k=10)

Implementation notes: Filter syntax depends on the data store backend. See Query filters and the backend documentation for supported operators and field names.

The previous vector retriever implementation (BasicVectorRetriever) is deprecated. See Vector Retriever (Legacy) only if you still use the legacy vector data store API.