Amazon Memorydb does it support return_fields for vector search?

Question

I'm following this search query guide for querying vector:
https://redis.readthedocs.io/en/stable/examples/search_vector_similarity_examples.html

When reaching the part of this:

```
query = (
    Query("@vector:[VECTOR_RANGE $radius $vec]=>{$YIELD_DISTANCE_AS: score}")
     .sort_by("score")
     .return_fields("id", "score")
     .paging(0, 3)
     .dialect(2)
)
```

I get an unknown field of 'id' or score.

This query seems to be working on my local using redis-stack but when using memory db it doesn't work.  But removing the return_fields() work but it returns all the data..

Any idea why?

Sedat Salman · Accepted Answer

Amazon MemoryDB may not support the same vector search features as Redis Stack, such as return_fields for vector search, which can result in unknown field errors for 'id' or 'score'. This difference might be because MemoryDB and Redis Stack support different subsets of Redis commands and functionalities. For detailed MemoryDB capabilities, you should refer to its official documentation or AWS support for more accurate information. If return_fields() is not supported, removing it might default to returning all fields, as you observed. For the most accurate and up-to-date information, check Amazon MemoryDB's official documentation.

watert · Answer

I have the same problem.
Seems like the return fields must all be indexed field.
That's different with the Redis Stack.

Amazon Memorydb does it support return_fields for vector search?

Relevant content