Semantic search for GEO datasets

[MomirMilutinovic]

Users should be able to query GEO datasets related to a list of PubMed IDs using natural language to find quickly surface the most relevant data.

Create an endpoint /relevant-datasets that:

• Takes as input a list of PubMed IDs and a query string
• Returns a list of GEO datasets related to the PubMed IDs sorted by their relevance to the query

Notes:

• The relevance of a dataset to the user's query should be measured using cosine similarity between the embeddings of the query and dataset metadata
• Use the PubTrends sentence-transformer API to fetch embeddings