search.py

def calc_cosine_similarity(self, vec1, vec2):
  """Calculate cosine similarity between two vectors."""
  v1, v2 = np.array(vec1), np.array(vec2)
  return np.dot(v1, v2) / (np.linalg.norm(v1) * np.linalg.norm(v2))

def search_cache(self, embedding, threshold=0.80):
  """       
  Returns the most similar response if it is above the threshold. Otherwise returns `None`.
  """
  k = 1
  cql = "SELECT * FROM prompts ORDER BY prompt_embedding ANN OF %s LIMIT %s;"
  results = scylla_client.query_data(cql, [embedding, k])
  if len(results) > 0:
    cached_response = results[0]
    similarity = self.calc_cosine_similarity(embedding, cached_response['prompt_embedding'])
    print("Similarity score:", similarity)
    if similarity >= threshold:
      return cached_response['llm_response']
  return None