app/main.py

from fastapi import Depends, FastAPI, HTTPException, status

from app.config import get_settings
from app.embedding import embed_text, embed_texts
from app.schemas import (
    BatchEmbedRequest,
    BatchEmbedResponse,
    EmbedRequest,
    EmbedResponse,
    HealthResponse,
)
from app.security import require_api_key

app = FastAPI(title="ACCESS SBERT Embedding Service")


def _validate_text(text: str, max_length: int) -> str:
    trimmed = text.strip()

    if not trimmed:
        raise HTTPException(
            status_code=status.HTTP_422_UNPROCESSABLE_ENTITY,
            detail="Text is required",
        )

    if len(trimmed) > max_length:
        raise HTTPException(
            status_code=status.HTTP_422_UNPROCESSABLE_ENTITY,
            detail="Text too long",
        )

    return trimmed


@app.get("/health", response_model=HealthResponse)
def health() -> dict[str, int | str]:
    settings = get_settings()
    return {
        "status": "ok",
        "model": settings.model_name,
        "dimension": settings.dimension,
    }


@app.post("/embed", response_model=EmbedResponse)
def embed(
    request: EmbedRequest,
    _: None = Depends(require_api_key),
) -> dict[str, int | list[float] | str]:
    settings = get_settings()
    text = _validate_text(request.text, settings.max_text_length)
    vector = embed_text(text)

    return {
        "embedding": vector,
        "dimension": len(vector),
        "model": settings.model_name,
    }


@app.post("/embed/batch", response_model=BatchEmbedResponse)
def embed_batch(
    request: BatchEmbedRequest,
    _: None = Depends(require_api_key),
) -> dict[str, int | list[list[float]] | str]:
    settings = get_settings()

    if len(request.texts) > settings.max_batch_size:
        raise HTTPException(
            status_code=status.HTTP_422_UNPROCESSABLE_ENTITY,
            detail="Too many texts",
        )

    texts = [
        _validate_text(text, settings.max_text_length)
        for text in request.texts
    ]
    vectors = embed_texts(texts)

    return {
        "embeddings": vectors,
        "dimension": len(vectors[0]) if vectors else 0,
        "model": settings.model_name,
    }