DocumentDB Vector Search with Filtering

Question

Hi all, is it possible to do DocumentDB Vector Search with Filtering?

query = {
        "$search": {
            "vectorSearch": {
                "vector": [0.2, 0.6, 0.9],
                "path": 'vectorEmbedding',
                "similarity": 'euclidean',
                "k": 2,
                "efSearch": 40,
                'filter': {
                    '$and': [
                      {
                        'price': {
                          '$gt': 20
                        }, 
                        'price': {
                          '$lt': 50
                        }
                      }
                    ]
                }, 
            }
        }
    }

results = test_collection.aggregate([query])

for result in results:
    print(result)

It resulted in this error:
OperationFailure: BSON field '$search.vectorSearch.filter' is an unknown field., full error: {'ok': 0.0, 'code': 40415, 'errmsg': "BSON field '$search.vectorSearch.filter' is an unknown field.", 'operationTime': Timestamp(1714105218, 1)}

Answer

I am facing the same problem, and I found a workaround. Just prepend a "$match" statement to your query. This way:

```
mycoll.insert_many([
    {"color": "blue", "vectorField": [0.82421875, -0.6953125]},
    {"color": "red", "vectorField": [0.82421875, -0.6953125]},
    {"color": "red", "vectorField": [0.05883789, -0.02038574]},
    {"color": "red", "vectorField": [0.2, 0.6]},
    {"color": "green", "vectorField": [-0.020385742, 0.32421875]}])

query = { "$search": { "vectorSearch": { "vector": [0.82421875, -0.6953125], "path": 'vectorField', "similarity": 'euclidean', "k": 2}}}
query_filter = {
    "$match": {
        "color":{ "$eq": "red"} 
    }
}

c = mycoll.aggregate([query_filter,query])
print(list(c))

```

I played a bit with this and it seems to work. I'll let mongo experts maybe comment if there is a better way.

DocumentDB Vector Search with Filtering

Relevanter Inhalt