ContextQMD
Back to Llama Index

Redis Vector Store

docs/examples/vector_stores/RedisIndexDemo.ipynb

0.14.2111.2 KB
Original Source

<a href="https://colab.research.google.com/github/run-llama/llama_index/blob/main/docs/examples/vector_stores/RedisIndexDemo.ipynb" target="_parent"></a>

Redis Vector Store

In this notebook we are going to show a quick demo of using the RedisVectorStore.

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

python
%pip install -U llama-index llama-index-vector-stores-redis llama-index-embeddings-cohere llama-index-embeddings-openai
python
import os
import getpass
import sys
import logging
import textwrap
import warnings

warnings.filterwarnings("ignore")

# Uncomment to see debug logs
logging.basicConfig(stream=sys.stdout, level=logging.INFO)

from llama_index.core import VectorStoreIndex, SimpleDirectoryReader
from llama_index.vector_stores.redis import RedisVectorStore

Start Redis

The easiest way to start Redis is using the Redis Stack docker image or quickly signing up for a FREE Redis Cloud instance.

To follow every step of this tutorial, launch the image as follows:

bash
docker run --name redis-vecdb -d -p 6379:6379 -p 8001:8001 redis/redis-stack:latest

This will also launch the RedisInsight UI on port 8001 which you can view at http://localhost:8001.

Setup OpenAI

Lets first begin by adding the openai api key. This will allow us to access openai for embeddings and to use chatgpt.

python
oai_api_key = getpass.getpass("OpenAI API Key:")
os.environ["OPENAI_API_KEY"] = oai_api_key

Download Data

python
!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'

Read in a dataset

Here we will use a set of Paul Graham essays to provide the text to turn into embeddings, store in a RedisVectorStore and query to find context for our LLM QnA loop.

python
# load documents
documents = SimpleDirectoryReader("./data/paul_graham").load_data()
print(
    "Document ID:",
    documents[0].id_,
    "Document Filename:",
    documents[0].metadata["file_name"],
)

Initialize the default Redis Vector Store

Now we have our documents prepared, we can initialize the Redis Vector Store with default settings. This will allow us to store our vectors in Redis and create an index for real-time search.

python
from llama_index.core import StorageContext
from redis import Redis

# create a Redis client connection
redis_client = Redis.from_url("redis://localhost:6379")

# create the vector store wrapper
vector_store = RedisVectorStore(redis_client=redis_client, overwrite=True)

# load storage context
storage_context = StorageContext.from_defaults(vector_store=vector_store)

# build and load index from documents and storage context
index = VectorStoreIndex.from_documents(
    documents, storage_context=storage_context
)
# index = VectorStoreIndex.from_vector_store(vector_store=vector_store)

Query the default vector store

Now that we have our data stored in the index, we can ask questions against the index.

The index will use the data as the knowledge base for an LLM. The default setting for as_query_engine() utilizes OpenAI embeddings and GPT as the language model. Therefore, an OpenAI key is required unless you opt for a customized or local language model.

Below we will test searches against out index and then full RAG with an LLM.

python
query_engine = index.as_query_engine()
retriever = index.as_retriever()
python
result_nodes = retriever.retrieve("What did the author learn?")
for node in result_nodes:
    print(node)
python
response = query_engine.query("What did the author learn?")
print(textwrap.fill(str(response), 100))
python
result_nodes = retriever.retrieve("What was a hard moment for the author?")
for node in result_nodes:
    print(node)
python
response = query_engine.query("What was a hard moment for the author?")
print(textwrap.fill(str(response), 100))
python
index.vector_store.delete_index()

Use a custom index schema

In most use cases, you need the ability to customize the underling index configuration and specification. For example, this is handy in order to define specific metadata filters you wish to enable.

With Redis, this is as simple as defining an index schema object (from file or dict) and passing it through to the vector store client wrapper.

For this example, we will:

  1. switch the embedding model to Cohere
  2. add an additional metadata field for the document updated_at timestamp
  3. index the existing file_name metadata field
python
from llama_index.core.settings import Settings
from llama_index.embeddings.cohere import CohereEmbedding

# set up Cohere Key
co_api_key = getpass.getpass("Cohere API Key:")
os.environ["CO_API_KEY"] = co_api_key

# set llamaindex to use Cohere embeddings
Settings.embed_model = CohereEmbedding()
python
from redisvl.schema import IndexSchema


custom_schema = IndexSchema.from_dict(
    {
        # customize basic index specs
        "index": {
            "name": "paul_graham",
            "prefix": "essay",
            "key_separator": ":",
        },
        # customize fields that are indexed
        "fields": [
            # required fields for llamaindex
            {"type": "tag", "name": "id"},
            {"type": "tag", "name": "doc_id"},
            {"type": "text", "name": "text"},
            # custom metadata fields
            {"type": "numeric", "name": "updated_at"},
            {"type": "tag", "name": "file_name"},
            # custom vector field definition for cohere embeddings
            {
                "type": "vector",
                "name": "vector",
                "attrs": {
                    "dims": 1024,
                    "algorithm": "hnsw",
                    "distance_metric": "cosine",
                },
            },
        ],
    }
)
python
custom_schema.index
python
custom_schema.fields

Learn more about schema and index design with redis.

python
from datetime import datetime


def date_to_timestamp(date_string: str) -> int:
    date_format: str = "%Y-%m-%d"
    return int(datetime.strptime(date_string, date_format).timestamp())


# iterate through documents and add new field
for document in documents:
    document.metadata["updated_at"] = date_to_timestamp(
        document.metadata["last_modified_date"]
    )
python
vector_store = RedisVectorStore(
    schema=custom_schema,  # provide customized schema
    redis_client=redis_client,
    overwrite=True,
)

storage_context = StorageContext.from_defaults(vector_store=vector_store)

# build and load index from documents and storage context
index = VectorStoreIndex.from_documents(
    documents, storage_context=storage_context
)

Query the vector store and filter on metadata

Now that we have additional metadata indexed in Redis, let's try some queries with filters.

python
from llama_index.core.vector_stores import (
    MetadataFilters,
    MetadataFilter,
    ExactMatchFilter,
)

retriever = index.as_retriever(
    similarity_top_k=3,
    filters=MetadataFilters(
        filters=[
            ExactMatchFilter(key="file_name", value="paul_graham_essay.txt"),
            MetadataFilter(
                key="updated_at",
                value=date_to_timestamp("2023-01-01"),
                operator=">=",
            ),
            MetadataFilter(
                key="text",
                value="learn",
                operator="text_match",
            ),
        ],
        condition="and",
    ),
)
python
result_nodes = retriever.retrieve("What did the author learn?")

for node in result_nodes:
    print(node)

Restoring from an existing index in Redis

Restoring from an index requires a Redis connection client (or URL), overwrite=False, and passing in the same schema object used before. (This can be offloaded to a YAML file for convenience using .to_yaml())

python
custom_schema.to_yaml("paul_graham.yaml")
python
vector_store = RedisVectorStore(
    schema=IndexSchema.from_yaml("paul_graham.yaml"),
    redis_client=redis_client,
)
index = VectorStoreIndex.from_vector_store(vector_store=vector_store)

In the near future -- we will implement a convenience method to load just using an index name:

python
RedisVectorStore.from_existing_index(index_name="paul_graham", redis_client=redis_client)

Deleting documents or index completely

Sometimes it may be useful to delete documents or the entire index. This can be done using the delete and delete_index methods.

python
document_id = documents[0].doc_id
document_id
python
print("Number of documents before deleting", redis_client.dbsize())
vector_store.delete(document_id)
print("Number of documents after deleting", redis_client.dbsize())

However, the Redis index still exists (with no associated documents) for continuous upsert.

python
vector_store.index_exists()
python
# now lets delete the index entirely
# this will delete all the documents and the index
vector_store.delete_index()
python
print("Number of documents after deleting", redis_client.dbsize())

Troubleshooting

If you get an empty query result, there a couple of issues to check:

Schema

Unlike other vector stores, Redis expects users to explicitly define the schema for the index. This is for a few reasons:

  1. Redis is used for many use cases, including real-time vector search, but also for standard document storage/retrieval, caching, messaging, pub/sub, session mangement, and more. Not all attributes on records need to be indexed for search. This is partially an efficiency thing, and partially an attempt to minimize user foot guns.
  2. All index schemas, when using Redis & LlamaIndex, must include the following fields id, doc_id, text, and vector, at a minimum.

Instantiate your RedisVectorStore with the default schema (assumes OpenAI embeddings), or with a custom schema (see above).

Prefix issues

Redis expects all records to have a key prefix that segments the keyspace into "partitions" for potentially different applications, use cases, and clients.

Make sure that the chosen prefix, as part of the index schema, is consistent across your code (tied to a specific index).

To see what prefix your index was created with, you can run FT.INFO <name of your index> in the Redis CLI and look under index_definition => prefixes.

Data vs Index

Redis treats the records in the dataset and the index as different entities. This allows you more flexibility in performing updates, upserts, and index schema migrations.

If you have an existing index and want to make sure it's dropped, you can run FT.DROPINDEX <name of your index> in the Redis CLI. Note that this will not drop your actual data unless you pass DD

Empty queries when using metadata

If you add metadata to the index after it has already been created and then try to query over that metadata, your queries will come back empty.

Redis indexes fields upon index creation only (similar to how it indexes the prefixes, above).