import lancedb
import os
import gradio as gr
from sentence_transformers import SentenceTransformer, CrossEncoder
from dotenv import load_dotenv

load_dotenv()

DB_PATH = os.getenv("DB_PATH", ".lancedb")

db = lancedb.connect(DB_PATH)

TABLE_NAME = os.getenv("TABLE_NAME")

if not TABLE_NAME:
  raise ValueError("TABLE_NAME environment variable is not set")

TABLE = db.open_table(TABLE_NAME)
VECTOR_COLUMN = os.getenv("VECTOR_COLUMN", "vector")
TEXT_COLUMN = os.getenv("TEXT_COLUMN", "text")
BATCH_SIZE = int(os.getenv("BATCH_SIZE", 32))

retriever = SentenceTransformer(os.getenv("EMB_MODEL"))

def retrieve(query, k):
    query_vec = retriever.encode(query)
    try:
        documents = TABLE.search(query_vec, vector_column_name=VECTOR_COLUMN).limit(k).to_list()
        documents = [doc[TEXT_COLUMN] for doc in documents]

        return documents

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

def reranking(query: str, docs: list[str], k: int) -> list[str]:
  model_name = 'BAAI/bge-reranker-v2-m3'
  # model_name = 'cross-encoder/ms-marco-MiniLM-L-6-v2'
  model = CrossEncoder(model_name, max_length=512)

  # Prepare the list of lists (query, document) for the model
  pairs = [[query, curr] for curr in docs]
  scores = model.predict(pairs)
  scored_pairs = zip(scores, docs)
  sorted_pairs = sorted(scored_pairs, reverse=True)
  return [pair[1] for pair in sorted_pairs][:k]