backend/models/embedder.py

"""MiniCPM-Embedding client — runs on Modal GPU (NF4 4-bit)."""

from typing import List

from config import settings
from models.modal_client import get_embedder


class MiniCPMEmbedder:
    def __init__(self):
        self._remote = get_embedder()

    def embed_documents(self, texts: List[str]) -> List[List[float]]:
        if not texts:
            return []
        batch_size = settings.EMBED_BATCH_SIZE
        if len(texts) <= batch_size:
            return self._remote.embed_documents.remote(texts)

        vectors: List[List[float]] = []
        for start in range(0, len(texts), batch_size):
            batch = texts[start : start + batch_size]
            vectors.extend(self._remote.embed_documents.remote(batch))
        return vectors

    def embed_query(self, query: str) -> List[float]:
        return self._remote.embed_query.remote(query)

    def get_embedding_dim(self) -> int:
        return 2304