gistfile1.txt

schema passage {
   document passage {
     field doc_id type long {} #duplicated for every passage extracted from doc
     field title type string {} #duplicated for every passage from doc
     field passage type string {}
     field embedding type tensor<float>(x[768]) {}
    }
    rank-profile hybrid {
       inputs {
         query(q): tensor<float>(x[768])
       }
       first-phase {
         expression: 0.12*bm25(title) + 0.5*bm25(passage) + 11.3*closeness(field, embedding) 
    }
}

vespa query \
    'yql=select title, doc_id from passage where {targetHits:100}nearestNeighbor(embedding,q) or userQuery() limit 0 | all(group(doc_id) max(10) each(output(count()) max(2) each(output(summary()))))' \
    'query=hybrid is the way' \
    'type=weakAnd' \
    'ranking=hybrid' \
    'input.query(q)=[2...........]' \

#Search for passages using hybrid retriveal, group by document id and list at max 10 unique doc_ids ordered by max relevance (as assigned by hybrid),
display 2 best scoring passages from doc per unique doc_id.