BaseEmbedder¶

Module: neurosurfer.models.embedders.base
Type: Abstract base class

Overview¶

BaseEmbedder defines the contract for embedding backends in Neurosurfer. All embedder implementations inherit from this class so they can be passed interchangeably to RAG pipelines, vector stores, or custom components.

Responsibilities¶

• Manage a shared logger instance
• Load any backend-specific model inside the subclass constructor
• Implement a single embed(query, **kwargs) method that accepts either one string or a list of strings and returns vectors as Python lists

Creating a custom embedder¶

from typing import List, Union
from neurosurfer.models.embedders.base import BaseEmbedder

class MyEmbedder(BaseEmbedder):
    def __init__(self):
        super().__init__()
        self.model = load_something()

    def embed(self, query: Union[str, List[str]], **kwargs) -> Union[List[float], List[List[float]]]:
        if isinstance(query, list):
            return [self._encode(text) for text in query]
        return self._encode(query)

    def _encode(self, text: str) -> List[float]:
        return self.model.encode(text).tolist()

Return plain Python lists so downstream code can serialise or cast to numpy/torch as needed.

See also¶

• SentenceTransformerEmbedder
• LlamaCppEmbedder

mkdocstrings output is temporarily disabled while import hooks are updated.