Metadata filtering

Vector Databases for Embeddings with Pinecone

James Chapman

Curriculum Manager, DataCamp

Metadata filtering

{
    "genre": "action",
    "year": 2020,
    "color": "blue",
    "fit": "straight",
    "price": 29.99,
    "is_jeans": true,
    "areas": ["London", "Kent", "Bath"]
}

Metadata can be strings, numbers, Booleans, and lists of strings
Metadata filtering: reduces search space and query latency

¹ https://docs.pinecone.io/docs/metadata-filtering

Metadata filtering

index.query(
    vector=[-0.250919762305275, ...],

    filter={

        "genre": {"$eq": "documentary"},
        "year": 2019

    },

    top_k=1
)

¹ https://docs.pinecone.io/docs/metadata-filtering

Metadata filters

$eq - Equal to (number, string, boolean)
$ne - Not equal to (number, string, boolean)
$gt - Greater than (number)
$gte - Greater than or equal to (number)
$lt - Less than (number)
$lte - Less than or equal to (number)
$in - In array (string or number)
$nin - Not in array (string or number)

¹ https://docs.pinecone.io/docs/metadata-filtering

Metadata filtering - greater than

index.query(
    vector=[-0.250919762305275, ...],

    filter={

        "year": {"$gt": 2019},

    },

    top_k=1,

    include_metadatas=True

)

{'matches': [{'id': '1', 'score': 0.0478537641,
              'values': [],
              'metadata': {'genre': 'action', 'year': 2020}}],
 'namespace': '',
 'usage': {'read_units': 5}}

Let's practice!

Vector Databases for Embeddings with Pinecone