Function cosine_similarity

backend/apps/datasource/embedding/utils.py:6–18

(vec_a, vec_b)

def cosine_similarity(vec_a, vec_b):
    if len(vec_a) != len(vec_b):
        raise ValueError("The vector dimension must be the same")

    dot_product = sum(a * b for a, b in zip(vec_a, vec_b))

    norm_a = math.sqrt(sum(a * a for a in vec_a))
    norm_b = math.sqrt(sum(b * b for b in vec_b))

    if norm_a == 0 or norm_b == 0:
        return 0.0

    return dot_product / (norm_a * norm_b)