src/app/services/agent/rag/embeddings.py

"""Azure OpenAI embeddings client."""
from __future__ import annotations

from typing import Any

import httpx
import numpy as np

from ....core.config import get_settings
from ....core.errors import LLMError


class AzureEmbeddingClient:
    def __init__(self) -> None:
        self._settings = get_settings()

    def embed(self, texts: list[str]) -> np.ndarray:
        if not texts:
            return np.zeros((0, 0), dtype=np.float32)
        endpoint = self._settings.azure_openai_embeddings_endpoint or self._settings.azure_openai_endpoint
        deployment = self._settings.azure_openai_embeddings_deployment
        if not endpoint or not deployment:
            raise LLMError(
                code="llm_config",
                message="Azure OpenAI embeddings deployment is not configured.",
            )
        url = (
            f"{endpoint}/openai/deployments/{deployment}/embeddings"
            f"?api-version={self._settings.azure_openai_embeddings_api_version}"
        )
        api_key = (
            self._settings.azure_openai_embeddings_api_key
            or self._settings.azure_openai_api_key
        )
        headers = {"api-key": api_key}
        payload = {"input": texts}

        try:
            with httpx.Client(timeout=30.0) as client:
                response = client.post(url, json=payload, headers=headers)
                response.raise_for_status()
                data: dict[str, Any] = response.json()
        except httpx.HTTPStatusError as exc:
            raise LLMError(
                code="llm_http",
                message=f"Embeddings request failed with status {exc.response.status_code}.",
                details={"body": exc.response.text},
            ) from exc
        except httpx.HTTPError as exc:
            raise LLMError(code="llm_http", message="Embeddings request failed.") from exc

        try:
            vectors = [item["embedding"] for item in data["data"]]
        except (KeyError, TypeError) as exc:
            raise LLMError(code="llm_response", message="Invalid embeddings response.") from exc

        return np.asarray(vectors, dtype=np.float32)

    def _normalize_endpoint(self, endpoint: str) -> str:
        cleaned = endpoint.rstrip("/")
        marker = "/openai/"
        if marker in cleaned:
            cleaned = cleaned.split(marker, 1)[0]
        return cleaned