File size: 1,182 Bytes
ef93aa3
2412cb7
ef93aa3
 
2412cb7
 
 
 
 
 
 
 
ef93aa3
2412cb7
 
ef93aa3
2412cb7
 
ef93aa3
2412cb7
 
ef93aa3
 
 
 
 
2412cb7
ef93aa3
 
 
 
2412cb7
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
import lancedb # type: ignore
import os
import gradio as gr # type: ignore
from sentence_transformers import SentenceTransformer, CrossEncoder # type: ignore


db = lancedb.connect(".lancedb")

TABLE = db.open_table(os.getenv("TABLE_NAME"))
VECTOR_COLUMN = os.getenv("VECTOR_COLUMN", "vector")
TEXT_COLUMN = os.getenv("TEXT_COLUMN", "text")
BATCH_SIZE = int(os.getenv("BATCH_SIZE", 32))
RERANKER = os.getenv("RERANKER", "cross-encoder/ms-marco-MiniLM-L-6-v2")

retriever = SentenceTransformer(os.getenv("EMB_MODEL"))
reranker = CrossEncoder(RERANKER)


def retrieve(query, k, rerank_factor=3):
    query_vec = retriever.encode(query)
    try:
        documents = (
            TABLE.search(query_vec, vector_column_name=VECTOR_COLUMN)
            .limit(k * rerank_factor)
            .to_list()
        )
        documents = [doc[TEXT_COLUMN] for doc in documents]
        scores = reranker.predict([(query, doc) for doc in documents])
        best_scores_and_documents = sorted(zip(scores, documents), reverse=True)[:k]
        best_documents = [doc[1] for doc in best_scores_and_documents]
        return best_documents

    except Exception as e:
        raise gr.Error(str(e))