Weaviate Vector Store - Hybrid Search#

If you’re opening this Notebook on colab, you will probably need to install LlamaIndex 🦙.

!pip install llama-index

import logging
import sys

logging.basicConfig(stream=sys.stdout, level=logging.INFO)
logging.getLogger().addHandler(logging.StreamHandler(stream=sys.stdout))

Creating a Weaviate Client#

import weaviate

resource_owner_config = weaviate.AuthClientPassword(
    username="<username>",
    password="<password>",
)

# Connect to cloud instance
# client = weaviate.Client("https://<cluster-id>.semi.network/", auth_client_secret=resource_owner_config)

# Connect to local instance
client = weaviate.Client("http://localhost:8080")

from llama_index import VectorStoreIndex, SimpleDirectoryReader
from llama_index.vector_stores import WeaviateVectorStore
from llama_index.response.notebook_utils import display_response

Download Data#

!mkdir -p 'data/paul_graham/'
!wget 'https://raw.githubusercontent.com/run-llama/llama_index/main/docs/examples/data/paul_graham/paul_graham_essay.txt' -O 'data/paul_graham/paul_graham_essay.txt'

Load documents#

# load documents
documents = SimpleDirectoryReader("./data/paul_graham/").load_data()

Build the VectorStoreIndex with WeaviateVectorStore#

from llama_index.storage.storage_context import StorageContext


vector_store = WeaviateVectorStore(weaviate_client=client)
storage_context = StorageContext.from_defaults(vector_store=vector_store)
index = VectorStoreIndex.from_documents(
    documents, storage_context=storage_context
)

# NOTE: you may also choose to define a index_name manually.
# index_name = "test_prefix"
# vector_store = WeaviateVectorStore(weaviate_client=client, index_name=index_name)

Query Index with Default Vector Search#

# set Logging to DEBUG for more detailed outputs
query_engine = index.as_query_engine(similarity_top_k=2)
response = query_engine.query("What did the author do growing up?")

display_response(response)

Query Index with Hybrid Search#

Use hybrid search with bm25 and vector.
alpha parameter determines weighting (alpha = 0 -> bm25, alpha=1 -> vector search).

By default, `alpha=0.75` is used (very similar to vector search)#

# set Logging to DEBUG for more detailed outputs
query_engine = index.as_query_engine(
    vector_store_query_mode="hybrid", similarity_top_k=2
)
response = query_engine.query(
    "What did the author do growing up?",
)

display_response(response)

Set `alpha=0.` to favor bm25#

# set Logging to DEBUG for more detailed outputs
query_engine = index.as_query_engine(
    vector_store_query_mode="hybrid", similarity_top_k=2, alpha=0.0
)
response = query_engine.query(
    "What did the author do growing up?",
)

display_response(response)