Skip to content

Latest commit

 

History

History
181 lines (124 loc) · 6.26 KB

README.md

File metadata and controls

181 lines (124 loc) · 6.26 KB
Raw

Streamlit-Weaviate Connection

Weaviate Weaviate

This project provides a Streamlit connection for the open-source vector database, Weaviate.

Overview

The Streamlit-Weaviate Connector enables developers to connect to a Weaviate database with the following Python code:

   conn = st.connection(
       "weaviate",
       type=WeaviateConnection,
       url=os.getenv("WEAVIATE_URL"),
       api_key=os.getenv("WEAVIATE_API_KEY"),
   )

Features

  • Hybrid search functionality
  • GraphQL query support
  • Advanced operations using the Weaviate Python client
  • Secrets management for secure connections

🔧 Installation

This project uses poetry for dependency management. To install:

  1. Ensure you have Python >=3.11.0 installed
  2. Install poetry: pip install poetry
  3. Install the project: poetry add git+https://github.com/weaviate/st-weaviate-connection.git

Usage

Check out the demo notebook for basic usage examples.

Before running, set these environment variables:

WEAVIATE_URL=YOUR_WEAVIATE_CLUSTER_URL
WEAVIATE_API_KEY=YOUR_WEAVIATE_API_KEY

🔗 Basic Usage

Connecting to a Weaviate Cloud instance

To connect to a Weaviate Cloud instance, use the following code:

conn = st.connection(
    "weaviate",
    type=WeaviateConnection,
    url=weaviate_url,
    api_key=weaviate_apikey,
    additional_headers={"X-OpenAI-Api-Key": openai_apikey},
)

Where weaviate_url and weaviate_apikey are the URL and API key of your Weaviate Cloud instance, respectively.

Queries

You can use the query method or graphql_query method to query your Weaviate instance.

To perform a hybrid search, use the following method:

df = conn.query(
    collection_name="<COLLECTION NAME>",
    query="<QUERY STRING>",
)

To perform a GraphQL GET query, use the following method:

df = conn.graphql_query(
    """
    {
        Get {
            <YOUR QUERY GOES HERE>
        }
    }
    """
)

Both methods return a Pandas DataFrame.

query method

The query method is a convenience method that was created for the Weaviate connection.

It performs a hybrid search on the Weaviate instance. The method requires collection_name and query as arguments.

  • collection_name : str
    • The name of the collection to query.
  • query : str
    • The query to search for.

It also accepts limit, filters, target_vectors, query_properties, and alpha as optional arguments. Refer to the function docstring for additional information on these arguments.

Notes
  • The query method is a convenience method and is not a direct Weaviate operation. It uses the Weaviate Python client under-the-hood to perform the operation.
  • The query method does not support caching via the cache_ttl parameter. If you need caching, use the graphql_query method.

Advanced Usage

The Weaviate connection uses the Weaviate Python client under-the-hood to interact with the Weaviate instance.

Accordingly, you can use the Weaviate client object directly to perform advanced operations.

We recommend using the following syntax:

with conn.client() as client:
    # Use the client object to perform required operations
    # e.g. 1: Create a collection
    # client.collections.create(...)
    #
    # e.g. 2: Perform retrieval augmented generation & print the generated recommendation
    # collection = client.collections.get(...)
    # response = collection.generate.hybrid(
    #     limit=4,
    #     query="a sweet european red wine",
    #     grouped_task="From these, recommend a wine that would pair well with a steak",
    # )
    #
    # print("## Generated recommendation")
    # print(response.generated)
    ...

This way, the client will also automatically close the connection after the block is executed, ensuring that there are no resource leaks.

See the Weaviate Python client documentation, and the Weaviate documentation for more information on the available operations.

Example notebook

The project includes a demonstration notebook to showcase basic functionalities of the connection (see here demo notebook)

Before you run the Jupyter notebook make sure that you have set the following environment variables

Set environment variables:

WEAVIATE_URL=<YOUR WEAVIATE CLOUD INSTANCE URL>
WEAVIATE_API_KEY=<YOUR WEAVIATE CLOUD INSTANCE API KEY>

📚 Documentation

All connection functionality can be found in the connection.py python file. Documentation about st.connection can be found here.

Demo app(s)

Movie Magic

🎬🍿 Movie Magic is a simple, but fun movie recommendation app built with Streamlit and Weaviate, using st-weaviate-connection Weaviate connection.

The app allows users to search for movies based on search terms using hybrid, semantic, and keyword search options. Then, the app provides recommendations based on the viewing occasion.

Run the app using the following command:

streamlit run demo_app.py

Magic Chat

🔮 Magic Chat searches through Magic The Gathering cards with various search options, such as keyword, semantic and hybrid, and performs retrieval-augmented. The live demo is accessible through Streamlit Community Cloud.

It was built using a previous version (0.0.1) of the Weaviate connection, and the code can be found in the this repository.

💖 Open Source Contribution

Now, you're all set to use the Weaviate Connector for Streamlit. Happy coding!

We encourage contributions. Feel free to suggest improvements, provide feedback, create issues, and submit bug reports!