Embedder typedef
Embeds a batch of texts into vectors.
rag_kit does not talk to any embedding model itself. You supply this function and bind it to whatever produces vectors: an HTTP call to OpenAI or Ollama, a local ONNX model, or a deterministic fake in tests.
Implementations receive the whole batch in one call, so an HTTP-backed embedder can send a single request instead of one request per text. The returned list must contain exactly one vector per input text, in input order, and all vectors must have the same length.
Implementation
typedef Embedder = Future<List<List<double>>> Function(List<String> texts);