search

codemonger-io / mumble

Mumble in public through ActivityPub
MIT License
2 stars 1 forks source link

How do we request similarity search? API? What does the API look like? #28

Open kikuomax opened 1 year ago

kikuomax commented 1 year ago

Simply receive a query text? Or receive an embedding vector?

kikuomax commented 1 year ago

I think it is OK to receive a query text. OpenAI's embedding vector needs at least 1,536 × 4 bytes ≒ 6 KB in binary form.

kikuomax commented 1 year ago

We do not want to expose an API to calculate embeddings to prevent abuse.

kikuomax commented 1 year ago

The core of the similarity search will be a Lambda function implemented with flechasdb + flechasdb-s3.

The viewer may directly invoke the Lambda function.