instantiate a redis vector store with custom index schema and vector schema

[sjadlakha]

I am trying to instantiate the redis vector store for my pre existing index in redis.
It's vector field is named description_vector and there are other text fields that need to be retrieved.
I am landing up with the error :

langchain_community/vectorstores/redis/schema.py
ValueError("No content_vector field found")

I am looking for the correct way to instantiate the vector store so that all the default values for "content" and "content_vectors" are overrriden.

[dosubot[bot]]

To instantiate a Redis vector store with a custom index schema and vector schema, you need to override the default "content" and "content_vectors" values by specifying your custom vector field name "description_vector". Here's how you can do it:

1. Define your custom vector schema with the desired field name and other parameters.
2. Pass this custom vector schema to the from_texts method when creating the Redis instance.

Here's an example code snippet:

from langchain_community.vectorstores.redis.base import Redis

# Define your custom vector schema
custom_vector_schema = {
    "name": "description_vector",  # Your custom vector field name
    "algorithm": "FLAT",
    "dims": 1536,  # Ensure this matches your embedding dimensions
    "distance_metric": "COSINE",
    "datatype": "FLOAT32",
}

# Create the Redis instance with the custom vector schema
rds = Redis.from_texts(
    texts,  # a list of strings
    metadata,  # a list of metadata dicts
    embeddings,  # an Embeddings object
    vector_schema=custom_vector_schema,  # Pass the custom vector schema here
    redis_url="redis://localhost:6379",
)

This setup should help you override the default settings and use your custom vector field "description_vector" ^[1].

To continue talking to Dosu, mention @dosu.

[sjadlakha]

@dosu
I already have the documents loaded in my redis db
I want to now instantiate the vector store to work as the retriever.
Please provide a solution to this requirement

[dosubot[bot]]

To connect to your existing Redis index and use it as a retriever, you can use the from_existing_index method of the Redis class. This method allows you to specify your existing index name and schema without creating a new index. Here's how you can do it:

from langchain_community.vectorstores.redis.base import Redis
from langchain.embeddings import Embeddings

# Assuming you have an Embeddings object
embeddings = Embeddings(...)

# Connect to the existing Redis index
rds = Redis.from_existing_index(
    embedding=embeddings,
    index_name="your_existing_index_name",  # Replace with your actual index name
    schema={
        "name": "description_vector",  # Your custom vector field name
        # Add other fields as necessary
    },
    redis_url="redis://localhost:6379",
)

# Now you can use `rds` as a retriever

This approach connects to your existing index and allows you to use it for retrieval operations ^[1].

To continue talking to Dosu, mention @dosu.